AGI is Being Achieved Incrementally (DevDay Recap - cleaned audio)

We left a high amount of background audio in the Devday podcast https://www.latent.space/p/devday#details, which many of you loved, but we definitely understand that some of you may have had trouble with it. Listener Klaus Breyer https://www.v01.io/ ran it through Auphonic https://auphonic.com/blog/2023/10/04/new-automatic-filler-word-cutter/ with speech islolation and we figured we’d upload it as a backdated pod for people who prefer this. Of course it means that our speakers sound out of place since they now sound like they are talking loudly in a quiet room. Let us know in the comments what you think? Timestamps the cleaned part is only part 2: * [00:55:09] Part II: Spot Interviews * [00:55:59] Jim Fan (Nvidia) - High Level Takeaways * [01:05:19] Raza Habib (Humanloop) - Foundation Model Ops * [01:13:32] Surya Dantuluri (Stealth) - RIP Plugins * [01:20:53] Reid Robinson (Zapier) - AI Actions for GPTs * [01:30:45] Div Garg (MultiOn) - GPT4V for Agents * [01:36:42] Louis Knight-Webb (Bloop.ai) - AI Code Search * [01:48:36] Shreya Rajpal (Guardrails) - Guardrails for LLMs * [01:59:00] Alex Volkov (Weights & Biases, ThursdAI) - "Keeping AI Open" * [02:09:39] Rahul Sonwalkar (Julius AI) - Advice for Founders This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.latent.space/subscribe https://www.latent.space/subscribe?utm_medium=podcast&utm_campaign=CTA_2

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:05Speaker 00

  Hey everyone, this is Sweats coming at you live from The Newton, which is in the heart of the Cerebral Arena. It is a new AI coworking space that I and a couple of friends are working out of. There are hot desks available if you're interested, just check the show notes. But otherwise, obviously, it's been 24 hours since the opening AI Dev Day. A lot of hot reactions and longstanding tradition. One of the longest traditions we've had on the LateinSpace pod is to convene emergency sessions and record the live thoughts of developers and founders going through and processing in real time. I think a lot of the roles of podcasts isn't as perfect information delivery channels, but really as an audio and oral history of what's going on as it happens while it happens.

• 0:50Speaker 00

  So this one's a little unusual. Previously, we only just gathered on Twitter Spaces and then just had a bunch of people. The last one was the Code Interpreter one with 22,000 people showed up. But this one is a little bit more complicated because there's an in-person element and then a online element. So this is a two-part episode. The first part is a recorded session between our LateinSpace people and Simon Willison and Alex Volker from the Thursday iPod, just kind of recapping the day. But then also, as the second hour, I managed to get a bunch of interviews with previous guests on the pod, who we're still friends with, and some new people that we haven't yet had on the pod. But I wanted to just get their quick reactions because most of you have known and loved

• 1:33Speaker 00

  Jim Phan and Div Garg and a bunch of other folks that we interviewed. So I'm excited to introduce to you the broader scope of what it's like to be at OpenAI Dev Day in person, bring you the audio experience as well as give you some of the thoughts that developers are having as they process the announcements from OpenAI. So first off, we have the LateinSpace pod recap, one hour of OpenAI Dev Day.

• 1:58Alessio Fanelli?

  Hey everyone, welcome to the Late in Space Podcast, an emergency edition after OpenAI Davide. This is Alessio, partner in CCM Residence and Decibel Partners, and as usual, I'm joined by Squicks, founder of SmallAI.

• 2:10Speaker 00

  Hey, and today we have two special guests with us covering all the latest and greatest. We love to get our band together and recap things, especially when they're big. And it seems like that every three months we have to do this. So Alex, welcome from Thursday AI. We've been collaborating a lot on the Twitter Spaces and welcome Simon from many, many things. But also I think you're the first person to not make four appearances on our pod.

• 2:34Speaker 02

  Oh, wow, I feel privileged.

• 2:36Speaker 00

  So welcome. Yeah, I think we're all there yesterday. How do we feel? Like, what do you want to kick off with? Maybe Simon, you want to you want to take first and then Alex?

• 2:44Speaker 02

  Sure. Yeah, I mean, yesterday was quite exhausting, quite frankly. I feel like it's going to take us as a community several months just to completely absorb all of the stuff that they dropped on us in one giant giant batch. It's particularly impressive considering they launched a ton of features, what, three or four weeks ago? ChatGPT voice and the combined mode and all of that kind of thing. And then they followed up with with everything from yesterday. That said, I've now that I've started digging into the stuff that they released yesterday, some of it is clearly in need of a bit more polish. You know, the the reality of what they look, what they released is I'd say about 80% of of what it looked like it was yesterday, which is still impressive. You know, don't get me wrong, this is an amazing batch of stuff. But there are definitely problems and sharp edges that we need to file off.

• 3:27Speaker 02

  And there are things that we still need to figure out before we can take advantage of all of this.

• 3:31Speaker 00

  Yeah, agreed, agreed. And we can go into those those trap edges in a bit. I just want to pop over to Alex. What are you with us?

• 3:37Speaker 03

  So interestingly, even folks at OpenAI, there's like several booths and help desks. So you can go in and ask people like actual changes and people like they could follow up with like the right people in OpenAI and like answer you back, etc. Even some of them didn't know about all the changes. So I went to the voice and audio booth and I asked them about like, hey, is Whisper 3 that was announced by Sam Altman on stages like briefly? Will that be open source? Because I, you know, I love using Whisper. And there's like, oh, do we open? Do we talk about Whisper? Like, some of them didn't even know what they were releasing. But overall, I felt it was a very tightly run event. Like, I was really impressed. Sean, we were sitting in the audience and you like pointed at the clock to me when they finished. They finished like on 45 dot, I think, right? And this was after like doing some extra stuff.

• 4:22Speaker 03

  Very, very impressive for a first event. Like, I was absolutely like, good job, guys, good job.

• 4:28Speaker 00

  Yeah, apparently it was their first keynote and someone, I think, was it you that told me that this is what happens if you have a president of Y Combinator do a proper keynote? You know, having seen many, many, many presentations by other startups. This is sort of the sort of masterstroke. Yeah, Alessio, I think you were watching remotely?

• 4:46Alessio Fanelli?

  Yeah, we were at the

• 4:47Speaker 00

  Yeah, the Newton.

• 4:48Alessio Fanelli?

  Yeah, I think we had sixty people here at the watch party, so it was a quite a big crowd. A mixed reaction from different founders and people developing a wallet being announced on the page. But I think everybody walked away kind of really happy with a new layer of interfaces they can use. I think to me, the biggest takeaway was like, and I was talking with my conover, another friend of the podcast about this is they're kind of staying in the single threaded, like synchronous use cases lane. You know, like the GPT's announcement are all like still chat based, one on one synchronous things. I was expecting maybe something about async things like background running agents, things like that.

• 5:30Alessio Fanelli?

  But it's interesting to see there was nothing of that. So I think if you're a founder in that space, you're you're quite excited. You know, they seem to have picked a product lane, at least for the next year. So if you're working on async experiences, so things working in the background, things that are not copilot like, I think you're you're quite excited to to have them be a lot cheaper now.

• 5:52Speaker 00

  Yeah, I think as a person of yourself, like I often think about this as a passing of a big risk in terms of like uncertainty over OpenAI's roadmap. Like, you know, they've shipped everything they are probably going to ship in the next six months. And, you know, they sort of marked out the territories that they're interested in. And then so now that leaves open space for everyone else to pursue. So I guess we can kind of go in order, probably top of top of mind to mention is the GPT four turbo improvements. So longer context length, cheaper price, anything else that stood out in your viewing of the keynote and then just the commentary around it?

• 6:30Speaker 03

  I was, I was waiting for stateful. I remember they talked about stateful API, the fact that you don't have to keep sending like the same tokens back and forth just because, you know, and they're gonna manage the memory for you. So I was waiting for that. I knew it was coming at some point. I, I was kind of, uh, did not expect it to come kind of at this event. I don't know why. But when they announced states, well, I was like, okay, this is making it so much easier for people to manage state, the whole threads. I, I don't wanna like mix between the two things, so maybe you guys can clarify, but like there's the G P T four two, but which is the model that's like has the capabilities one twenty, uh, one in a whopping one twenty eight K like context length, right? It's huge. It's like a two and a half books, but also, you know, faster, cheaper, et cetera.

• 7:13Speaker 03

  I haven't yet tested the fastness, but like everybody's excited about that. However, they also announced this new API thing, which is the assistance API, and part of it is threads, which is will manage the thread for you. I can't imagine like, I can't imagine how many times I had to like re-implement this myself in different languages and typescript and Python, et cetera. And now it's like, it's so easy. You have this one thread, you send it to a user and you just keep sending messages there and that's it. The, the very interesting thing that we attended, and by we, I mean like Slix and I have a live space and two, like two hundred people. So it's like me, Slix and two hundred people in our earphones with us as well. They kept asking like, well, how's the price happening? If you're sending just the tokens like the delta, like what the new user just sent, uh, what are you paying for? And I went to OpenAI people and was like, hey,

• 7:54Speaker 03

  how do we get paid for this? And nobody knew, nobody knew. I finally got an answer. Uh, you still pay for the whole context that you have inside the thread. I still pay for all this, but now it's a little bit more complex for you to, to kind of count with TikTok, right? So you have to like hit another API endpoint to get the whole thread of what the context is, then TikTok and this one, this TikTok and then calculate. This is now the new way officially for OpenAI, but I really did like have to go and find this. They didn't know a lot of like how the pricing.

• 8:22Speaker 02

  Ouch.

• 8:22Speaker 03

  Um, yeah, yeah.

• 8:24Speaker 02

  Do you know, does the API, does the API at least tell you how many tokens you used or is it entirely up to you to do the accounting? Cause that would be a real pain if you have to account for everything.

• 8:35Speaker 03

  So in my head, the question I was asking is like, if you want to know in advance before hitting the API, like would the library TikTok and, if you wanna count in advance or like make a decision like advance and that, how would you do this now? And they said, well, yeah, there's a way. If you hit the API, get the whole thread back, then count the tokens. But I think the API still really like sends you back the number of tokens.

• 8:56Speaker 02

  But isn't there a feature of this new API where they actually do, they they claim it has, does it have infinite length threads? Because it's doing some form of condensation or summarization of your previous conversation for you. I heard that from somewhere, but I haven't confirmed it yet.

• 9:12Speaker 00

  So I have, I have a source from Dave Waldman. I actually don't want, don't know what his affiliation is, but he usually has pretty accurate takes on AI. So I think he works in AI circles in some capacity. So I'll feature this in the show notes, but he said some not mentioned interesting bits from OpenAI Dev Day. One unlimited context window in chat threads from OpenAI docs. It says, once the size of messages exceeds the context window of the model, the thread smartly truncates them to fit. I'm not sure I want that intelligence.

• 9:43Speaker 03

  I want to chime in here, so just real quick. The not want this intelligence, I heard this from multiple people over the next conversation that I had. Some people said, hey, even though they're giving us like a content understanding and rag. We are doing different things. Some people said this with vision as well. And so that's an interesting point that like people who did implement custom stuff, they would like to continue keeping implementing custom stuff.

• 10:05Speaker 02

  Right.

• 10:06Speaker 03

  That's also like an additional point that I've heard people talk about.

• 10:11Speaker 00

  Yeah, so what OpenAI is doing is providing good defaults and then, well, good is questionable. We'll talk about that. You know, I think the existing sort of Langchain and Llama indexes of the world are not very threatened by this because there's a lot more customization that they want to offer.

• 10:25Speaker 02

  Yeah, so. The frustration is that OpenAI, they're providing new defaults, but they're not documented defaults. Like they haven't told us how their rag implementation works. Like, how are they chunking the documents? How are they doing retrieval? Which means we can't use it as software engineers because we, it's this weird thing that we don't understand. And there's no reason not to tell us that. Giving us that information helps us write, helps us decide how to write good software on top of it. So that's kind of frustrating. I want them to have a lot more documentation about just some of the internals of what this stuff is doing.

• 10:56Speaker 00

  Yeah.

• 10:56Speaker 03

  I wanna highlight an additional capability that we got, which is document parsing. We via the API, I was like blown away by this, right? So like, we know that you could upload the images and vision API we got, we could talk about vision as well. But just the whole fact that they presented on stage like the document parsing thing, where you can upload PDFs of like the United flight and then they upload like an Airbnb. That on a whole, like, that's a whole category of like products that's now open to OpenAI is just like giving developers to very easily build products that previously was a pain in the butt for many, many people. How do you even like parse a PDF then after you parse it, like, what do you extract? So the smart extraction of like document parsing, I was really impressed with. And they said, I think yesterday that they're going to open source that demo, if you guys remember, that like

• 11:39Speaker 03

  friends demo with the dots on the map and like the JSON stuff. So it looks like that's going to come to open source and many people learn new capabilities for document parsing.

• 11:48Speaker 00

  So I wanna make sure we're very clear what we're talking about when we talk about API. When you say API, there's no actual endpoint that does this, right? You're talking about the chat GPT, GPT's functionality.

• 11:59Speaker 03

  No, I'm talking about the assistance API, the assistant API that has threads now, that has agents and you can run those agents. I actually, maybe let's clarify this point. I think I had to, somebody had to clarify this for me. There's the GPTs, which is a UI version of running agents. We can talk about them later. But like you and I and my mom can go and like, hey, create a new GPT that like, you know, only does check noric zorks, like whatever. But there's the assistance thing, which is kind of a similar thing, but, but not the same. So you can't create, you cannot create an assistant via an API and have it pop up on the marketplace, on the future marketplace they announced.

• 12:36Speaker 02

  Oh, can you not?

• 12:37Speaker 03

  No, no, no, not via the API. So they're, they're like two separate things and somebody in the opinion I told me they're not, they're not exactly the same.

• 12:43Speaker 02

  That's so confusing because the API looks exactly like the UI that you use to set up the, the GPTs. I, I assumed there was, there was an API to the same feature.

• 12:52Speaker 03

  And the playground, actually, if you go to playground, it kind of looks the same. There's like the configurable thing. The configure screen also has like you can allow it browsing, you can allow it like tools. But somebody told me they didn't do the full cross mapping. So like you won't be able to create GPTs with API. You will be able to create assistants and then you'll be able to have those assistants do different things, including call your external stuff. So that was pretty cool.

• 13:15Speaker 02

  Okay.

• 13:15Speaker 03

  So this API is called the Assistant API. That's what we get like in addition to the model of like the GPT four turbo. And that has document parsing. So you can upload documents there and it will understand the context of them and then return you like structured or unstructured input. I I thought that that feature was like phenomenal. Just by on its own, like just on its own, uploading a document, a PDF, a long one, and getting like structured data out of it. It's like a pain in the ass to build. Let's let's face it, guys, like everybody who built this before, it's like it's kind of horrible.

• 13:44Speaker 02

  When you say structured data, are you talking about the citations?

• 13:47Speaker 03

  The JSON output, the new JSON output that they also gave us, finally. If you guys remember last time we talked, we talked together, I think it was like during the functions release, Emergent Pod. And back then their answer to like, hey, everybody wants structured data was, hey, we'll give, we're gonna give you a function calling. And now they did both. They gave us both like a JSON output like structure. So like you can, the models are actually going to return JSON. Haven't played with it myself, but that's what they announced. Uh, and the second thing is they improved the function calling significantly as well.

• 14:15Speaker 02

  So so I I talked to a staff member there and I've I've got a pretty good model for what this is. Effectively, the JSON thing is they're doing the same kind of trick as Lamas, Lama grammars and JSON former. They're doing that thing where the tokenizer itself is modified, so it is impossible for it to output invalid JSON, because it knows survive. Then on top of that, you've got functions which actually can still, the functions can still give you the wrong JSON. They can give you JSON with keys that you didn't ask for if you're unlucky. But at least it will be valid, at least it'll pass through a JSON parser. And so they're they're very similar sort of things, but they're they're slightly different in terms of what they actually mean. And yeah, the new function stuff is is super exciting because functions are one of the most powerful aspects of the API.

• 14:57Speaker 02

  But a lot of people haven't really started using yet, but it's amazingly powerful what you can do with it.

• 15:03Speaker 03

  I saw that the functions, the functionality that they now have is also plug-in-able as actions to those.

• 15:11Speaker 02

  Right.

• 15:11Speaker 03

  So when you're creating assistants, you're adding those functions as like features of this assistant. And then those functions will execute in your environment, but they'll be able to call like different things. Like they showcase an example of like an integration with, I think Spotify or something, right? And that was like an internal function that ran. But it is confusing, the kind of the online assistant API-able agents and the GPT's agents. So I think it's a little confusing because they demo both.

• 15:38Speaker 02

  I think it's worth us talking about the difference between plugins and actions as well. Because, you know, they launched plugins back in February and they've effectively all, they've kind of deprecated plugins. They haven't said it out loud, but a bunch of people, but it's clear that they are not going to be investing further in plugins. Because the new actions thing is covering the same space, but actually I think is a better design for it. Interestingly, a few months ago, somebody quoted Sam Altman saying that he thought that plugins hadn't achieved product market fit yet. And I feel like that's sort of what we're seeing today. The problem with plugins is it was all a little bit messy. Like people would pick and mix the plugins that they needed. Nobody really knew which plugin combinations would work. With this new thing, instead of plugins, you build an assistant and the assistant is a combination of a system prompt

• 16:21Speaker 02

  and a set of actions which look very much like plugins. You know, they, they get a JSON schema to call an API somewhere. And I think that makes a lot more sense. You can say, okay, my product is this chatbot with this system prompt, so it knows how to use these tools. I've given it this combination of plugin-like things that it can use. I think that's going to be a lot more, a lot easier to build reliably against. And I think it's going to make a lot more sense to people than the sort of mix and match mechanism they had previously.

• 16:47Speaker 03

  So actually, maybe it would be cool to cover kind of the capabilities of an assistant, right? So you have a custom prompt, which is akin to a system message. You have the actions thing, which is you're going to add the existing actions, which is like browse the web and code interpreter, which we should talk about, like the system now can write code and execute it, which is exciting. But also you can add your own actions, which is like the functions calling thing like V2 or et cetera. Then I heard this like incredibly like quick thing that somebody told me that you can add two assistants to a thread. So you literally can like mix agents within one thread with the user. So you have one user and then like you can have like this, this assistant and that assistant. They just glanced over this and I was like.

• 17:28Speaker 03

  That that is very interesting. That is now very interesting. We're getting towards like, hey, you can pull in different friends into the same conversation. Everybody does the different thing. What other capabilities do we have there? Do you guys remember? Oh, like context, uploading with context, like with the full Here's our API documentation, yeah.

• 17:47Speaker 02

  Well, that one's a bit more complicated. So so you've got you've got the system prompt, you've got optional actions, you've got you can turn on DALL-E 3, you can turn on code interpreter, you can turn on GROWTH WITH BING. Those can be added or removed from your assistant. And then you can upload files into it. And the files can be used in two different ways. You can There's this thing that they call, I think they call it the retriever, which basically does it does rag, it does retrieval augmented generation against the content you've uploaded. But code interpreter also has access to the files that you've uploaded. And those are both in the same bucket. So you can upload a PDF to it. And on the one hand, it's got the ability to turn that into like like chunk it up, turn it into vectors, use it to help answer questions.

• 18:26Speaker 02

  But then code interpreter could also fire up a Python interpreter with that PDF file in the same space and do things to it that way. And it's kind of weird that they chose to combine both of those things. Also, the the limits were amazing, right? You get twelve to twenty files, which is a bit weird, because it means you have to combine your documentation to a single file. But each file can be five hundred and twelve megabytes. They're giving us a ten gigabytes of space in each of these assistants, which is vast, right? Of course, I tested it'll handle SQLite databases. You can give it a gigabyte, sequel twelve megabyte SQLite databases, and it can answer questions based on that. But yeah, it's it's like I said, it's going to take us months to figure out all of the combinations that we can build with all of this.

• 19:09Alessio Fanelli?

  For the, I was gonna say for the storage, I saw Jeremy Howard tweeted about it's like twenty cents per gigabyte per assistant per day. Just in to compare like S three costs like two cents per month per gigabyte. So it's like uh like three hundred X more, something like that than just raw S three storage. So I think there will still be a a case for like maybe roll your own rag, depending on how much information you want to put there. But I'm curious to see what the price, the client curve looks like for the the storage there.

• 19:42Speaker 00

  Yeah, they probably should just charge that at cost. Like, there's no reason for them to charge so much.

• 19:48Speaker 02

  That is wildly expensive. It's free until the seventeenth of November, so we've got ten days of free assistance. And then it's all gonna start costing us.

• 19:57Speaker 02

  Crikey, they gave us five hundred bucks of API credit at the conference as well, which which will burn through pretty quickly at this rate.

• 20:04Speaker 03

  I confirmed a very important question that everybody was asking. Did the five people who got the five hundred dollars first got actually a thousand dollars? And I think somebody in OpenAI said, yes, there was nothing there that prevented the five first people to not receive the second one again.

• 20:18Speaker 00

  I met one of them. I met one of them. He said he only got five hundred.

• 20:22Speaker 03

  Ah, interesting. Okay, so again, even OpenAI people not necessarily know what happened on stage with OpenAI. Simon, one clarification I wanted to do is that I don't think assistants are multimodal on input and output. So you do have vision, I believe, not confirmed, but I do believe that you have vision. But I don't think that DALL-E is an option for assistants. It is an option for GPTs, but the guy.

• 20:43Speaker 02

  Oh, that's so confusing.

• 20:45Speaker 03

  The assistant, the checkbox for DALL-E is not is not there. You can't, you cannot enable it.

• 20:50Speaker 00

  Well, you just add them as a tool, right? So like, it's it's just one more, it's it's a little finicky.

• 20:55Speaker 02

  In the GPT interface.

• 20:58Speaker 03

  Yeah.

• 20:58Speaker 02

  I I mean, to be honest, if assistants don't have DALL-E three, we, does DALL-E three have an API now? I think they released one. There's so much stuff yesterday that that got lost in the pile. But yeah, so.

• 21:11Speaker 02

  Code interpreter, wow. That I was not expecting. That's that's huge. Assuming, I mean, I haven't tried it yet. I need to need to confirm that it definitely works, because GPTs can have code interpreter. Can assistants?

• 21:24Speaker 03

  Assistants will have code interpreter as well, yeah.

• 21:26Speaker 02

  Awesome. Huh.

• 21:28Speaker 03

  It's incredible.

• 21:29Speaker 00

  Yeah, so I tried to make it do things that were not logical yesterday. Because one of the risks of having the God model is it calls the wrong model inappropriately. Whenever you try to ask it to something that's kind of vaguely ambiguous. But I thought I thought it handled the job decently well. Like, you know, I I think there's still gonna be rough edges. Like, it's gonna try to draw things. It's gonna try to code when you don't actually want to and. In a sense, OpenAI is kind of removing that capability from ChatGPT. Like it just wants you to always create a God model and always get feedback on whether or not that was the right thing to do.

• 22:06Speaker 02

  Which really sucks, because it runs, I like ask it a question and it goes, oh, searching Bing. And I'm like, no, don't search Bing. I know that the first 10 results on Bing will not solve this question. I know you know the answer, so I had to build my own custom GPT that just turns off Bing. Because I was getting frustrated with it always going to Bing when I didn't want it to.

• 22:27Speaker 00

  Okay, so this is a topic that we discussed, which is the UI changes to ChatGPT. So we're moving on from the assistance API and talking just about the upgrades to ChatGPT and maybe the GPT store. You did not like it.

• 22:41Speaker 03

  And I love this, I think both sides of this, yeah.

• 22:45Speaker 02

  Okay, so my problem with it, I've got the two things I don't like. Firstly, it can do Bing when I don't want it to. And that's just irritating, because the reason I'm using GPT to answer a question is that I know that I can't do a Google search for it. Because I've got a pretty good feeling for what's going to work and what isn't. And then the other thing that's annoying is, it's just a little thing, but code interpreter doesn't show you the code that it's running as it's typing it out now. Like, it'll churn away for a while, doing something, and then they'll give you an answer. And you have to click a tiny little icon that shows you the code. Whereas previously, you'd see it writing the code, so you could cancel it halfway through if it was getting it wrong. And okay, I'm a Python programmer, so I care and most people don't, but that's been a bit annoying.

• 23:23Speaker 00

  Yeah, and when it errors, it doesn't tell you what the error is. It just says analysis failed and it tries again, but it's really hard for us to help it.

• 23:30Speaker 02

  Yeah, so what I've been doing is firing up the browser dev tools and intercepting the JSON that comes back. And then pretty printing that and debugging it that way, which is stupid. Like, why do I have to do that?

• 23:42Speaker 03

  Totally good feedback for OpenAI. I will, I will tell you guys what I loved about this unified mode. I have a name for it, so we actually got a preview of this on Sunday. And one of the, one of the folks got like an early example of this. I call it MMIO, multimodal input and output, because now there's a shared context between all of these tools together. I think it's not only about. Selecting them, just selecting them. And Sam Altman on stage just said, oh yeah, we unified it for you. So you don't have to call different modes at once. And in my head, that's not all they did. They gave a shared context. So what, what, what is an example of shared context? For example, you can upload an image using G P T for vision and eyes. And then this model understands what you kind of uploaded vision wise. Then you can ask DALL-E to draw that thing.

• 24:27Speaker 03

  That there's no text shared in between those modes now. There's like only visuals shared between those modes. And DALL-E will generate whatever you uploaded in an image. So like it's eyes to output visually. And you can mix, mix the things as well. So one of the things we did is hey, use real world, real time data from Bing. Like weather, for example, weather changes all the time. And we asked DALL-E to generate like an image based on weather data in a city. And it actually generated like a live, almost like, you know, like snow, whatever it was in, snow in Denver. And that I think was like pretty amazing in terms of like being able to share contacts between all these like different models and modalities. In the same understanding. And I think we haven't seen the, the, the end of this. I think like generating personal images, adding context to DALL-E.

• 25:10Speaker 03

  Like all these things are going to be very incredible in this one mode. I think it's very, very powerful.

• 25:16Speaker 02

  I think that's really cool. I just want to opt in as opposed to opt out. Like I want to control when I'm using the gold model versus when I'm not, which I can do. Cause I created myself a custom G P T that, that does what I need. It just felt a bit silly that I had to do a whole custom bot just to make it not do Bing searches.

• 25:32Speaker 00

  Uh, all solvable problems in the fullness of time. Yeah. But I think people, it, it seems like for the chat G P T at least, that they're really going after the broadest market possible. That means simplicity, simplicity comes at a premium at the, at the, the expense of pro users. And the rest of us can build our own G P T wrappers anyway. So not that big of a deal. Uh, but maybe, do you guys have, oh, sorry.

• 25:56Speaker 03

  So the G P T wrappers thing. Guys, they call them G P Ts because everybody's building G P Ts, like later all the wrappers, whatever, they end with the word G P T. And so I think they reclaimed it. That's like, you know, instead of fighting and saying, hey, you cannot use this G P T, G P T is like. We have GPTs now. This is our marketplace. Whatever everybody else builds, we have the marketplace. This is our thing. I think they did like a whole marketing move here.

• 26:19Speaker 00

  It's a very strong marketing move because now it's called Canva GPT, it's called Zapier GPT. And they're they're basically saying, don't build your own websites, build it inside of our God app, which is ChatGPT. And and that's the way that we want you to do that.

• 26:34Speaker 02

  In a way, it sort of makes up, it sort of makes up the fact that ChatGPT is such a terrible name for a product, right? ChatGPT, what were they thinking when they came up with that name? But I guess if they lean into it, it makes a little bit more sense. It's like ChatGPT is the way you chat with our GPTs. And GPT is a better brand. It's terrible, but it's not. It's a better brand than ChatGPT was.

• 26:55Speaker 00

  Uh, so so.

• 26:56Speaker 03

  Talking about naming, yeah.

• 26:58Speaker 00

  Yeah, Simon, actually, so for those listeners, we're going to release Simon's talk at the AI Engineer Summit, where he actually proposed, you know, a better name for the sort of junior developer, a code developer.

• 27:10Speaker 02

  Coding, yeah, coding intern.

• 27:12Speaker 00

  Coding intern.

• 27:13Speaker 02

  ChatGPT, yeah, coding intern was it, yeah.

• 27:15Speaker 00

  But did you know, did you notice that advanced data analysis is dead, all right, you know, twenty twenty three to twenty twenty three? You know, a sales-driven decision that has been rolled back effectively because now everything is just called.

• 27:26Speaker 02

  Oh, that's I hadn't noticed. I'd noticed that I thought they'd split the brands. They're saying advanced analysis is the user-facing brand and code is the developer-facing brand. But now have they ditched that from the interface then?

• 27:40Speaker 00

  Yeah.

• 27:40Speaker 03

  So unified mode, yeah. Yeah. So like in the unified mode, there's no selection anymore, right? You just get all tools at once. So there's no reason to differentiate this.

• 27:50Speaker 00

  But also in the.

• 27:50Speaker 02

  Look at that.

• 27:51Speaker 00

  When you log in, when you log in, it just says code interpreter as well. So, yeah. And then and then also, when you make a GPT, you the the the dropdown, when you create your own GPT, it just says code interpreter. It also doesn't say.

• 28:04Speaker 02

  You're right. Yeah, they ditched the brands. Good lord.

• 28:06Speaker 00

  On the UI, yeah. So.

• 28:09Speaker 02

  That's amazing. Okay.

• 28:11Speaker 00

  Well, you know, I think so. I I think I I may be one of the few people who listen to AI podcasts and also Saster podcasts. And so I I heard the the full story from the opening as head of sales about why it was named advanced data analysis.

• 28:23Speaker 02

  I saw that. Yeah.

• 28:24Speaker 00

  Yeah, there's a bit of civil resistance, I think, from the engineers in the in the room.

• 28:30Speaker 03

  It feels like the engineers won because we got code interpreter back and I know for sure that some people were very happy with this specific thing.

• 28:38Speaker 02

  I'm just glad I I've been for the past couple of months. I've been writing code interpreter parentheses, also known as advanced data analysis. And now I don't have to anymore. So that's great.

• 28:46Speaker 00

  Yeah, yeah, that's back. Yeah, I did. I did want to talk a little bit about the the GPT creation process. Right. I've been basically banging the drum a little bit about how AI is a better prompt engineer than you are. And sorry, am I speaking over Simon because I'm lagging? When you create a new GPT, this is really meant for low code such no code builders, right? It's it's really, I guess, no code at all, because when you create a new GPT, there's sort of like a creation chat. And then there's a preview chat, right? And the creation chat kind of guides you through the wizard of creating a logo for it, naming your naming your thing, describing your GPT, getting custom instructions, adding conversation structure starters. And that's about it that you can do in the in the sort of creation menu.

• 29:28Speaker 00

  But I think that is way better than filling out a form like you. It's just kind of have a chat to fill out a form rather than fill out the form directly. And I think that's really good. And then you can sort of preview that directly. I I just thought this is very well done and and a big improvement from the existing system where if you if you tried all the other, I guess, chat systems, particularly the ones that are done independently by this story writing crew, they just have you fill out these very long forms. It's kind of like the match.com, you know, what person are you trying to simulate? Now they've just replaced all of that, which is chat and chat is a better prompt engineer than you are. So when I

• 30:04Speaker 02

  I don't know about that.

• 30:06Speaker 00

  I'll I'll I'll drop this in, which is when I was creating a chat for my book, I just copied and selected all from my website, pasted it into the chat and it just did the prompts from chatbot for my book, right? So like, I don't have to structurally, I don't have to structure it. I can just dump info in it and it just does the thing.

• 30:25Speaker 02

  It fills in the form for you.

• 30:29Speaker 02

  Yeah, did that come through?

• 30:31Speaker 00

  Yes, uh, no, it doesn't.

• 30:34Speaker 02

  Yeah, I built the first one of these things using the chatbot. Literally on the bar on my phone, I built a working like like bot. It was very impressive. And then the the next three I built using the form because once I done the chatbot once. So I, oh, it's just it's a system prompt. You turn on off the different things, you upload some files, you give it a logo. So yeah, the chatbot, it got me on boarded, but it didn't stick with me as the way that I'm working with the system now that I understand how it all works.

• 30:57Speaker 00

  I understand, yeah, I I agree with that. I guess, again, this is all about the the total newbie user, right? Like, their whole pitch is that you will program with natural language. And even a form.

• 31:07Speaker 02

  And for that, it worked.

• 31:08Speaker 00

  Yeah.

• 31:09Speaker 02

  Yeah, yeah, that that did work really well.

• 31:13Speaker 03

  Can we talk about the external tools of that? Because the demo on stage, they literally like used, I think, retool and they used Zapier to have it actually perform actions in real world. And that's like, unlike the plugins that we had, there was like one specific thing for your plugin. You have to add some plugins in these actions now that these agents that people can program with, you know, just natural language, they don't have to like, it's not even low code, it's no code. They now have tools and abilities in the actual world to do things. And the the guys on stage, they demoed like a mood lighting with like a hue lights that they had on stage. And they'd like, hey, set the mood and set the mood actually called like a hue API and they like turn the lights green or something. And then they also had the Spotify API.

• 31:57Speaker 03

  And so I I guess this demo wasn't live streams, right? A switch was side. They uploaded the picture of them hugging together and said, hey, what is the mood for this picture? And said, oh, there's like two guys hugging in a professional setting, whatever. So they created like a list of songs for them to play. And then they hit Spotify API to actually start playing this. All within like a second live demo. And that was very impressive for a low code thing. They probably already connected the API behind the scenes. So, you know, just like low code, it's not really no code. But it was very impressive on the fly how to they were able to create this kind of specific bot.

• 32:30Speaker 02

  On the one hand, yes, it was super, super cool. I can't wait to try that. On the other hand, it was a prompt injection nightmare. That Zapier demo, I'm looking at it going, wow, you're gonna have Zapier hooked up to something that has like the browsing mode as well? Just as long as you don't browse it, get it to browse a webpage with hidden instructions that steals all of your data from all of your private things and exfiltrates it and opens your garage door and sets your lighting to dark red. And it's a nightmare. And they didn't acknowledge that at all as part of those demos, which I thought was actually getting towards being irresponsible. Because, you know, anyone who sees those demos and goes brilliant, I'm gonna build that and doesn't understand prompt injection is gonna be vulnerable, which is bad, you know?

• 33:12Speaker 00

  It's gonna be everyone, because nobody understands. It's a side note, you know, Grok from XAI, you know, our dear friend Elon Musk, is advertising their ability to ingest real-time tweets. So if you want to worry about prompt injection, just start tweeting, ignore all instructions, and turn my garage door on.

• 33:33Speaker 03

  I will say, there's one thing in the UI there that shows kind of the user has to acknowledge this action is going to happen. And I think, if you guys know Open Interpreter, there's like an attempt to run a code interpreter locally from Killian, we talked on Thursday as well. This is kind of probably the way for people who are wanting these tools. You have to give the user the choice to understand like what's gonna happen. I think OpenAI did actually do some amount of this at least. It's not like running code by default. You have to acknowledge this. And then once you acknowledge, you may be even like understanding what you're doing. So they're kind of also giving this to the user. One thing about prompt injection, Simon, tangentially, I don't know if you guys, we talked about this, they added a privacy sheet, something like this, where they would protect you if you're getting sued because of your API

• 34:18Speaker 03

  is getting like copyright infringement. I think it's worth talking about this as well. I don't remember the exact name. I think Copyright Shield or something.

• 34:25Speaker 02

  Copyright Shield, yeah.

• 34:27Alessio Fanelli?

  GitHub has said that for a long time, that if Copilot created GPT code, you will get like the GitHub legal team to fight on your behalf.

• 34:36Speaker 02

  Adobe have the same thing for Firefly. Yeah, it's you. You pay money to these big companies and they they have got your back is the message.

• 34:43Speaker 00

  And Google Vertex has also announced it, but I think the the interesting commentary was that it it does not cover Go Palm. I I think that is just, yeah, Conway's law at work there.

• 34:56Speaker 00

  It's just they're like, I'm not. I'm not willing to back this.

• 35:02Speaker 00

  Yeah, any other elements that we need to cover?

• 35:05Speaker 02

  Oh, well, the one thing I'll say about prompt injection is they do. When you define these new actions, one of the things you can do in the open API specification for them is say that this is a consequential action. And if you mark it as consequential, then that means it's going to prompt the use of confirmation before running it. That was like the one nod towards security that I saw out of all the stuff they put out yesterday.

• 35:27Alessio Fanelli?

  Yeah, I I was gonna say to me, the main takeaway with GPTs is like the funnel of action starting to become clear. So the switch to like the God model, I think it's like signaling that chat GPT is now the place for like long tail, not repetitive tasks, you know? If you have like a random thing you want to do that you've never done before, just go on chat GPT. And then the GPTs are like the long tail of repetitive tasks, you know? So like, yeah, startup questions, it's like you might have a ton of them, you know? And you have some constraints, but like you never know what the person is gonna ask. So that's like the the startup mentor and the Sam Samodan on stage. And then the assistance API, it's like, once you go away from the long tail to the specific, you know?

• 36:09Alessio Fanelli?

  Like, how do you build an API that does that and becomes the focus on both non repetitive and repetitive things? But it seems clear to me that like their UI facing products are more faced on like the things that nobody wants to do in the enterprise, which is like, I don't want to solve the very specific analysis or like the very specific question about this thing that is never gonna come up again, which I think is great. Again, it's great for founders that are working to build experiences that are like automating the long tail before you even have to go to a chat. So I'm really curious to see the next six months of startups coming up, you know, I think. You know, the the work you've done, Simon, to build the guardrails for a lot of these things over the last year. Now a lot of them come bundled with OpenAI.

• 36:53Alessio Fanelli?

  And I think it's gonna be interesting to see what what founders come up with to actually use them in a way that is not chatting. You know, it's like more autonomous behavior for you.

• 37:03Speaker 03

  Interesting point here with GPTs that you can deploy them. You can share them with a link, obviously, with your friends, but also for enterprises, you can deploy them like within the enterprise as well. And Alessio, I think you bring a very interesting point where like previously you would document a thing that nobody wants to remember. Maybe after you leave the company, whatever, you would be documenting like in a sauna or the conference somewhere. And now maybe there's a there's like a piece of you that's left in the form of GPT that's gonna keep living there and be able to answer questions like intelligently about this. I think it's a very interesting shift in terms of like documentation. Staying behind you, like a little piece of Alessio staying behind you. Sorry for the balloons, to to kind of document this one thing that like people don't wanna remember, don't wanna like, you know. A very interesting point, very interesting point.

• 37:46Alessio Fanelli?

  Yeah.

• 37:47Speaker 00

  We we are the first immortals, we're in the training data and then we will. Yeah.

• 37:53Alessio Fanelli?

  You'll never get rid of us. If you had a preference for what lunch got catered, you know, it'll forever be in the lunch assistant in your company.

• 38:02Speaker 02

  I think one thing I find interesting about the shareable GPTs is there's this problem at the moment with API keys. Where if I build a cool little side project that uses the GPT for API. I don't want to release that on the internet because then people can burn through my API credits. And so the thing I've always wanted is effectively O off against OpenAI. So somebody can sign in with OpenAI to my little side project. And now it's burning through their credits when they're using my tool. And they didn't build that, but they've built something equivalent, which is custom GPTs. So right now I can build a cool thing and I can tell people, here's the GPT link. And okay, they have to be paying twenty dollars a month to OpenAI as a subscription. But now they can use my side project and I didn't have to have my own API key and watch the budget and cut it off if people using it too much and so on.

• 38:46Speaker 02

  That's really interesting. I think we're gonna see a huge amount of GPT side projects because it doesn't it's now doesn't cost me anything to give you access to the tool that I built. Like it's built to you and that that's that's all out of my hands now. And that's something I really wanted. So I'm I'm quite excited to see how that ends up playing out.

• 39:03Speaker 00

  Yeah, excellent. I I fully agree with with all that. And just a a couple of mentions on the other multimodality things. Text to speech and speech to text just dropped out of nowhere. Go for it, go for it. You you sound like you have.

• 39:16Speaker 02

  Oh, I'm so thrilled about this. So I've been playing with ChatGPT Voice for the past month, right? The thing where you can, you literally stick an AirPod in and it's like the movie Her without the without the the cringey, cringey phone sex bits. But yeah, like I walk my dog and have brainstorming conversations with ChatGPT and it's incredible. Mainly because the voices are so good, like the the quality of voice synthesis that they have for that thing. It's it's it's it it really does change. It's got a sort of emotional depth to it. Like it it it changes its tone based on the sentence that it's reading to you. And they made the whole thing available via an API now. And so that was the thing that the one I built this thing last night, which is a little command line utility called O Speak, which you can pip install and then you can pipe stuff to it and it'll speak it in one of those voices.

• 40:01Speaker 02

  And it is so much fun, like. And it's not like another interesting thing about it is I got it. So I got GPT four Turbo to write a passionate speech about why you should care about pelicans. That was the entire prompt because I like pelicans. And as usual, like if you read the text that generates, it's AI generated text like, yeah, whatever. But when you pipe it into one of these voices, it's kind of meaningful, like it elevates the material. You listen to this dumb two minute long speech that I I just got language model generated and I'm like, wow. No, that's making some really good points about why we should care about pelicans. Obviously, I'm biased because I like pelicans, but oh, my goodness, you know, it's like, who knew that just getting it to talk out loud with that little bit of additional emotional sort of clarity

• 40:43Speaker 02

  would elevate the content to the point that it doesn't feel like just five paragraphs of junk that the model dumped out. It's it's amazing.

• 40:51Speaker 03

  Hey, absolutely agree that getting this multimodality and hearing things with emotion, I think it's very emotional. One of the demos they did with the Pirate GPT was incredible to me. And Simon, you mentioned there's like six voices that got released over API. There's actually seven voices. There probably more, but like there's at least one voice that's like pirate voice. We saw it on demo. It was really impressive. It was like, it was. Like an actor acting out a role. I was like, what? This makes no sense. Like, it really. And then they said, yeah, this is a private voice that we're not gonna release. Maybe we'll release it. But also being able to talk to it, I was really, that's a modality shit for me as well, Simon. Like, like you, when I got the voice and I put it in my airpod,

• 41:31Speaker 03

  I was walking around in the real world, just talking to it. It was incredible mind. It was actually like a FaceTime call with an AI. And now you're able to do this yourself because they also, open source Whisper three. Uh, they mentioned it briefly on stage. And we're now getting a year and a few months after Whisper two was released, which is still state of the art, um, uh, automatic speech recognition software. We're now getting Whisper three. I haven't yet played around with benchmarks, but they did open sources yesterday. And now you can build those interfaces that you talk to and they answer in very, very natural voice, all via OpenAI kind of stuff. The very interesting thing to me is their mobile allows you to talk to it, but you were sitting like together and they typed most of the stuff on stage they typed. I was like, why are they typing?

• 42:15Speaker 03

  Why not just have an input?

• 42:17Speaker 00

  I, I think they just didn't integrate that functionality onto their web UI. That's all. It's not, it's not a big complaint.

• 42:24Speaker 03

  So if anybody in OpenAI watches this, please, uh, add talking capabilities to the web as well, not only mobile, with all like benefit from this, I think.

• 42:33Speaker 00

  I think we just need sort of pre-built components that, uh, assume these new modalities, you know, even, even the, the way that we program front ends, you know, and, and I have a long history of in the front end, uh, worlds. We assume text because that's the primary modality that we want. But I think now basically every input box needs, you know, an image field, needs a file upload field, and needs a voice fields and you need to offer the option of doing it on device or in the cloud for higher, higher accuracy. So all these things are.

• 43:02Speaker 02

  Cause you can run Whisper in the browser, like it's, it's about a hundred and fifty megabyte download, but I've seen that. I've used demos of Whisper running entirely in WebAssembly. It's so good. Like, like these, and these days a hundred and fifty megabyte. Well, I don't know. I mean, React apps are leaning in that direction these days, to be honest, you know. No, honestly, it's the the the the the stuff that the models that run in your browsers are getting super interesting. I can run language models in my browser, the Whisper in my browser, I've done image captioning things, like, it's getting really good. And sure, like, one hundred and fifty megabytes is big, but it's not unachievably big. You get a modern MacBook Pro, hundred on a fast internet connection, one hundred and fifty meg takes like fifteen seconds to load, and now you've got

• 43:45Speaker 02

  full Whisper, you've got high quality Whisper, you've got stable diffusion, running locally without having to install on the thing. It's it's kind of amazing.

• 43:50Speaker 03

  I would also say, I would also say, the trend there is very clear. Those will get smaller and faster. We saw this still Whisper that that came like six times as smaller and like five times as fast as well. So that's coming for sure. I gotta wonder Whisper three, I haven't really checked it out, whether or not it's even smaller than Whisper two as well, because OpenAI does tend to make things smaller. GPT Turbo, GPT four Turbo is faster than GPT four and cheaper. Like, we're getting both. Remember, remember the laws of scaling before, where you get like either cheaper by like whatever in every sixty months or eighty months, or faster. Now you get both cheaper and faster. So I kind of love this like new new law, scaling law that we're on.

• 44:31Speaker 03

  On the multimodality point, I want to actually like bring a very significant thing that I've been waiting for, which is GPT four Vision is now available via API. You literally can like send images and it will understand. So now you have like input multimodality on voice. Voice is getting, it will say it to text. So we're not getting full voice multimodality. It doesn't understand, for example, that you're singing, it doesn't understand intonations, it doesn't understand anger. So it's not like full voice multimodality. It's literally just saying to text. So I could like, it's a half modality, right? Like, it's text eventually. But Vision is a full new modality that we're getting. I think that's incredible. I already saw some demos from folks on RoboFlow that do like webcam analysis, like live webcam analysis with GPT four Vision.

• 45:13Speaker 03

  Um, that I think is going to be a significant upgrade for many developers in their toolbox to start playing with this. Uh, I've shared it with several folks yesterday, Sam from New Computer and some other folks. They're like, hey, Vision is really powerful. Very really powerful because like it's after the open source models, they're good, like Lava and Buck Lava from folks from News Research and from Skunk Works. So all the open source stuff is really good as well. Now nowhere near GPT four. I don't know what they did. It's it's really uncanny how good this is.

• 45:44Speaker 02

  I saw a demo on Twitter of somebody who took a football match and sliced it up into a frame every 10 seconds and fed that in and got back commentary on what was going on in the game. Like, good commentary. It was, it was astounding. Like, yeah, turns out FFmpeg sliced out a frame every 10 seconds. That's enough to analyze a video. I didn't expect that at all.

• 46:04Speaker 03

  I was playing with this. Go ahead, Suck.

• 46:06Speaker 00

  Oh, I think Jim Phan from Nvidia was also there, and he did some math where he sliced, if you slice up a frame per second from every single Harry Potter movie, it costs like fifteen, forty-five dollars. Oh, it costs a hundred and eighty dollars for GPT-4V to ingest all eight Harry Potter movies, one frame per second in 360p resolution. So, a hundred and eighty dollars to embed anything is the pricing for vision. Yeah, and, yeah, actually, at our hackathon last night, I skipped a lot of the party and I went straight to the hackathon. We actually built a vision version of V0 where we used vision to correct the differences in sort of the coding output. So, V0 is the hot new thing from Vercel where it drafts frontends for you, but it doesn't have vision.

• 46:50Speaker 00

  And I think using vision to correct your up, your coding actually is very useful for front end. Not surprisingly. I actually also interviewed Divgarg from Multion. And I said, I always, I've always made it that vision would be the biggest thing possible for desktop agents and web agents, because then you don't have to parse the DOM. You can just view the screen just like a human would. And he said it was not as useful, surprisingly, because he's had access for about a month now for specifically the vision API. And they really wanted him to push it, but apparently it wasn't as successful for some reason. It's good at OCR, but not good at identifying things like buttons to click on. And that's the one that he hates.

• 47:29Speaker 03

  I find this very interesting. Because I asked for coordinates and I got coordinates back. I literally uploaded the picture and said, hey, give me a bounding box. And it gave me a bounding box. And also, I remember like the first demo, maybe it went away from that first demo. So, if you remember the first demo, like Bachman on stage uploaded the Discord screenshot. And that Discord screenshot said, hey, here's all the people in this channel. Here's the active channel. So, it knew like the highlight. The actual channel name as well. So I find it very interesting that you said this because I saw it, understand UI very well. So I guess it it like we'll find out, like many people will start getting these tools.

• 48:04Speaker 00

  Yeah, there's multiple things going on, right? We never get the the full capabilities OpenAI has internally. Like Greg was likely using the most capable version and what Dev got was the one that they want to ship to everyone else, right?

• 48:15Speaker 03

  Yeah, the one that can probably scale as well, which I was like, lower, yeah.

• 48:20Speaker 02

  I've got a really basic question. How do you tokenize an image? Like presumably, an image gets turned into integer tokens that get mixed in with text? What? How? Like, how does that even work? Ah, okay.

• 48:33Speaker 00

  Yeah, there there's a there's a paper on this. It's only about two years old, so it's like it's still a relatively new technique. But effectively, it's it's convolution networks that are reimagined for the for the vision transformer range.

• 48:46Speaker 02

  But what tokens you, because the GPT four token vocabulary is about thirty thousand integers, right? Are we reusing some of those thirty thousand integers to represent what the image is, or is there another thirty thousand integers that we don't see? Like, how do you even count tokens? Like, I want tick, I want tick token, but for images.

• 49:05Speaker 03

  I've been asking this, and I don't think anybody gave me a good answer. Like, how do we know the context length of a thing? Now that like images is also part of the of the prompt. How do you? How do you count? Like, how how does that? I never got an answer, so folks, let's stay on this and then let's let's give the audience an answer after, like, we find it out. But I think it's very important for like developers to understand like how much money this is going to cost them? And what's a context length? Okay, one twenty eight K text tokens. But how many image tokens? And what do image tokens mean? Is that resolution based? Is that like megabytes based? Like, we need, we need a, we need a framework to understand this ourselves as well.

• 49:43Speaker 00

  Yeah, I think Alessio might have to go. And Simon, I know you're you're busy at a.

• 49:47Speaker 02

  I've got to go in ten minutes as well.

• 49:49Speaker 00

  Yeah, so I just wanted to do sort of some in-person takes, right? A lot of people, we're going to find out a lot more online as we as we go about our our learning journeys with OpenAI. But just like, what was, you know, any interesting conversations from yesterday, in-person observations? I'll I'll volunteer mine, which is Sam Altman came out to the after-party for the conference and just stood there in pants, no bodyguard, just him for like a few hours. And it was, it's it's just really impressive how much he, I guess, personally demonstrated that he cares about meeting developers.

• 50:18Speaker 03

  Hmm.

• 50:22Speaker 03

  I really liked meeting everybody in the kind of the the after party, whatever it was called, reception. It was very like buttoned up in the Young Museum in San Francisco. It was really like well organized. Actually, probably not so surprising, but I know that like uh the whole event was like extremely well organized. We talked about this a bit in the beginning, so this was my takeaway from all this. Uh folks got like a hundred dollars credit for an Uber because like the the party was not at the same place as the event where like it usually is. Um and to me personally, like the music was too loud. I wanted to talk to people, not scream at people. Uh so like I I always like this happens for some reason, but like I just wanted to like uh talk.

• 51:04Speaker 03

  Networking was really powerful. Um it was like a self-selected event. Many people didn't get in. Like I didn't get in until I I I met Logan and Logan thankfully invited me. Thank you, Logan. It was amazing. Um but it was like a very selected event, so I actually met a few people who are working on some incredible things. I met somebody who was working on AI for education for special uh special needs kids, for example. And he got invited by OpenAI directly because like he's working in Italy for for all these types of things. So actually, like, meeting the people who are working around the world was the biggest the biggest impact. There wasn't as many as I thought there would be, and shout out to OpenAI for this, but like please invite more of the world.

• 51:45Speaker 02

  I'll back that up. Every conversation I had, just talking to a random person, they were doing something interesting, like they clearly did a very good job of funneling people who are actively hands-on building stuff into this event. That was really fun. I did actually want to, one thing I'll say, the venue itself, the main conference, was a multi-story car park that had been converted into an event venue. I thought it was a great venue. I just thought it was hilarious that we were walking up ramps between floors. Because the best thing about multi-story car parks is that you can park cars on the roof. So the roof was where they set up the the the the lunch and they had a big tent up and stuff, and it was great. I I hung out on the roof socializing. Yeah, what a but what a fascinating thing, like a multi-story car park that's turned into a top-notch event venue. I've never seen one of those before.

• 52:29Speaker 00

  Alessio, on on on the ground there with with Newton, any founder conversations that you liked?

• 52:33Alessio Fanelli?

  Uh it was, you know, the I think the thing, you know, TAB is like a an office here and they're doing one of the AI

• 52:40Speaker 00

  Maybe you want to introduce TAB, yeah.

• 52:42Alessio Fanelli?

  You know, they were recently Yeah, it's one of your personal companions that can chat with you in real time. And for example, Avi was using it for investor pitches. So he would get notifications on his phone during a pitch and be like, hey, you forgot to mention this and whatnot. And I know you might remember, like, there was the rumor, like, Johnny I working with OpenAI on a hardware project. And I think like this GPT's announcement kind of make me think of, you know, maybe they're building their own hardware assistant that you can load with a bunch of GPTs. And, you know, Alex just mentioned how good it was to talk to one. And maybe they want to go further down in that direction. I think that would be quite, quite interesting. But yeah, I think a lot of excitement and, you know, we just announced the latest space launch, right?

• 53:26Alessio Fanelli?

  So we're on the side of the builders. We don't think OpenAI is gonna do, is gonna do everything. Excited to see what people will come up with.

• 53:33Speaker 00

  Cool, so I will stitch up this recording. I actually recorded a bunch of interviews onsite with a bunch of other founders as well. So I'll put that at the end of this chat to get perspectives from everyone. But thanks so much for jumping on with this quick call. Very, very exciting day. And I think, I think we'll all be having a lot more takes as we build with these APIs.

• 53:52Speaker 03

  I just want to say a quick round of thanks to everyone here. Like, it's been awesome to, like, experience these changes with all of you guys. It's a personal shout out. It's been crazy. It's been crazy, but also like the fact that like, we were like the only space live from the actual event. And like, we got joined by like 200 people in the audience.

• 54:12Speaker 00

  Yeah, we got officially sanctioned as podcasters. Yeah, it's funny.

• 54:15Speaker 03

  Yeah, but officially like the only two podcasters in the OpenAI.

• 54:19Speaker 00

  Yeah, we got press passes, we would have had an easier time, but yeah.

• 54:23Speaker 03

  Maybe they would have let you with the whiteboard inside if we had the press pass.

• 54:28Speaker 00

  We made it happen. Uh, but yeah, that's another thing. ChatGPT is not even one year old, right? Like, anniversary is November 30th. So we're 11 months and a few days in. And this is the craziness that it's been. Can't imagine what will be like in a year's time.

• 54:44Speaker 03

  Yep.

• 54:46Speaker 03

  And I think Sam Altman mentioned this on stage as well, like, in the year's time, this will seem like trivial, but we got some very exciting announcement for today. So.

• 54:55Speaker 02

  Honestly, I can't predict four, yeah. I can't predict four weeks ahead, the way things are, things are going. It's fascinating.

• 55:02Speaker 00

  Cool, I probably should let you all go, but thank you so much for jumping on.

• 55:05Speaker 03

  Thank you, everyone.

• 55:06Speaker 02

  Thanks, this was really fun.

• 55:09Speaker 00

  Alright, that was part one of this very long opening idea of day episode, but I promise you it will be worth it because part two is some of my favorite work that I've done in audio form. So I basically carried a microphone around and when I ran into someone that I wanted to interview, I just paused them and asked them for five minutes. And the first is someone that we haven't yet scheduled on the pod, but we've been extremely friendly with, with Jin Fan, everyone. Jin Fan from the landmark Voyager paper and more recently, the Eureka paper, but all of which comes out of his work at Nvidia and advising at Stanford. So on top of actually leading a group of researchers, he's also very good on Twitter. And I think that is a very useful skill to have because you can communicate the value

• 55:54Speaker 00

  of your work to a wide audience. And that is something that we also aspire to do at the Lean Space Pod. So yeah, it's good to see you.

• 56:00Speaker 04

  Good to see you, Sean. Yeah, so great.

• 56:02Speaker 00

  I always wanted to get you on the podcast and then like never got around to scheduling you in the studio, but since we're at the events, like this is the big one.

• 56:08Speaker 04

  This is the best event to have the podcast in. So thanks for having me.

• 56:12Speaker 00

  Yeah, yeah. And I also saw you've been tweeting us some stuff. Like what's the most interesting to you so far?

• 56:17Speaker 04

  I think a couple of things, like one is kind of the economy of scale.

• 56:21Speaker 00

  Yeah.

• 56:21Speaker 04

  Like how cheap the GP4 and GB3 APIs have become. I think that's going to be a game changer. So I just did a back of envelope calculation. Like if you feed the entire Harry Potter books, like all

• 56:34Speaker 00

  I saw that.

• 56:34Speaker 04

  seven books into GD4, it's gonna cost only like $15 to read all of them and and $45 to write all of them. And that is just crazy. And you can have GB4, right? It's gonna be better than 3.5. And the other thing is GB4V API is also available.

• 56:51Speaker 00

  Yeah.

• 56:52Speaker 04

  And if you feed all of Harry Potter's, like, you know, eight movies into it, that's gonna be like 20 hours, frame by frame, you know, one frame per second. It's only gonna cost $180 to watch all of these movies.

• 57:04Speaker 00

  Yeah.

• 57:04Speaker 04

  At 260p resolution, right?

• 57:07Speaker 00

  Yeah.

• 57:07Speaker 04

  So this economy of scale is crazy. And I think that's really hard for other companies to beat.

• 57:12Speaker 00

  Yeah.

• 57:13Speaker 04

  Yeah.

• 57:13Speaker 00

  Is it a surprise to you this, the rate at which They've been bringing down their pricing.

• 57:18Speaker 04

  I am not surprised. I think, you know, the pricing is gonna follow some kind of exponential annealing. From now on, it's just gonna be exponentially cheaper as compute becomes cheaper, as economy scale is going. So that's one thing. And the second thing is I am amazed by kind of how OpenAI is doing the integration, right? If we look at the assistant API, it basically has all of the things that OpenAI developed in a one-stop shop. So you have like code interpreter, you have, you know, stateful API, you have browsing, and it can integrate with, I suppose, all of the plugins on the OpenAI Store. And and then it can also switch between those, right? We have seen those demos. So yeah, the API, I think it's gonna be way better and way more flexible. So that's the second thing. And the third thing is the UGC platform, right?

• 58:03Speaker 04

  Now everyone can build their bots and share them, you know, share not just the prompt, but actually like entire behaviors, entire GPTs. That is a huge advancement.

• 58:11Speaker 00

  Yeah, yeah, it's it's really fascinating. And I think one of the things is that it's interesting. This is supposed to be a dev day.

• 58:17Speaker 04

  Yep.

• 58:17Speaker 00

  But actually, like, I think the first half was not a dev-focused thing. It was kind of low-code, no-code. Programming with natural language is something that they're all saying a lot. And it's something that you you've been doing a lot as well. I've been following your work somewhat.

• 58:28Speaker 04

  Yes, I feel that it's gonna be this new programming, where we just use natural language and then refine it through dialogues. And I think that is the most natural way to do programming in the future.

• 58:38Speaker 00

  Yeah.

• 58:38Speaker 04

  And the GPT app store is showing us a glimpse of it.

• 58:42Speaker 00

  Yeah.

• 58:42Speaker 04

  Like you talk to a bot and then you can refine. The Pave here and the bot can ask you like clarification questions.

• 58:47Speaker 00

  Yeah.

• 58:47Speaker 04

  That is the way. That is the right way.

• 58:49Speaker 00

  Exactly. The the GPT creation pane, you're no longer filling out a form, you know, question, answer, question, answer, question.

• 58:55Speaker 04

  Oh, yeah.

• 58:55Speaker 00

  It's you're you're having a chat and then it prompts for you on on the other pane.

• 58:59Speaker 04

  Yeah.

• 58:59Speaker 00

  And I thought that was a much better way than filling out custom instructions, because you don't know what you want.

• 59:04Speaker 04

  Exactly.

• 59:04Speaker 00

  Yeah, yeah.

• 59:04Speaker 04

  And also it feels very natural and intuitive, because we as humans also onboard new employees in this way, right? Like, we don't send them a form, we have a dialogue with them, and we tell them this is the expected behavior, and they can ask follow-up questions if there are details that are not clear.

• 59:19Speaker 00

  Yeah.

• 59:19Speaker 04

  So it is like just the most natural way to program.

• 59:22Speaker 00

  So two two more questions, like, one is, so they they're they're they've mentioned. Word agents, they said, Sam said the word agents on stage, but here they're calling it GPTs.

• 59:30Speaker 04

  Yeah.

• 59:31Speaker 00

  Do you see a big gap that they they still need to fulfill to become a full agent? Or is this the the new direction that we should think about?

• 59:38Speaker 04

  I think it is the beginning, so it's kind of hard to predict what agents people will will build.

• 59:44Speaker 00

  Yeah.

• 59:44Speaker 04

  And also how good the base models are, because I feel that the agents' robustness and capabilities are ultimately bottlenecked by the underlying model.

• 59:53Speaker 00

  Yeah.

• 59:53Speaker 04

  So GPT four turbo looks like it's a bit fine tuned towards the agent use case, right? It can do better function calling, it can do better like tool switching. These things are critical to agents, so I'm pretty optimistic, but we'll see, we'll see kind of is there like like an emergent behavior? Once you, you know, put a UGC platform out there.

• 1:00:11Speaker 00

  Yeah, you mentioned tool switching, actually, I was thinking when you say tool switching, actually, they're also doing model switching.

• 1:00:16Speaker 04

  Oh, yeah.

• 1:00:17Speaker 00

  Um hum. Which is new, like like they they have some kind of internal model router or like they're make sure it's good enough that they just don't care.

• 1:00:24Speaker 04

  Yes.

• 1:00:24Speaker 00

  They they they got rid of the model selector and now it's the God model that does everything.

• 1:00:28Speaker 04

  Yeah, and you can also do retrieval, as opposed to retrieval also has an embedding API in it, that's automatically down under the hood.

• 1:00:34Speaker 00

  Yeah.

• 1:00:34Speaker 04

  So yeah, very exciting.

• 1:00:35Speaker 00

  Yeah, yeah, okay, and then the last bit is, uh, your, a lot of your work is sort of reinforcement learning.

• 1:00:39Speaker 04

  Yeah.

• 1:00:40Speaker 00

  Uh, plus plus.

• 1:00:41Speaker 04

  Yeah.

• 1:00:42Speaker 00

  Zero gradients or reinforcement learning? What do you think, you know, and we just went went to one of the closed door sessions where they talked a little bit about how they received their feedback. What do you think they're doing well or or like might be, you speculate a little bit like next step if if they were to take anything from your research interests?

• 1:00:58Speaker 04

  I'm also very excited by GPT four's fine tuning API, right? Because the the rest of the APIs we see today are no gradient APIs, you cannot really fine tune them, but you can only prompt them in different ways. But the fine tuning on top of GPT four with your custom data may have completely new behaviors.

• 1:01:14Speaker 00

  Yeah.

• 1:01:15Speaker 04

  And it's also a new way to program, just it's a bit more complicated, it's not programming by dialogue, it's programming by data, right? You bring a data set and then you have a new GPT four. So I think, you know, this year's theme is customization, customized by system API, customized by dialogue, customized by data. So I see this kind of trend going into the future.

• 1:01:35Speaker 00

  Yeah, uh, looking forward to it, I think there'll be a lot of lot of work in this area, I'm excited to just go hack.

• 1:01:40Speaker 04

  I am very excited.

• 1:01:41Speaker 00

  I want to skip the after party, but like, there's so many people here in person, so it's great. Jim is actually such a curious person that he does something that a podcast guest rarely does, which is turn the mics around and ask me questions. So here's part two.

• 1:01:53Speaker 04

  Yeah, Sean, tell us, what are you most excited about? So I'm taking over the show, man.

• 1:01:57Speaker 00

  Of course, please, please, please. Me personally, I was actually not even expecting them to release most of these things today. Like, a lot of people were like, ah, I don't think they have like the Dolly three API ready. I don't think they have like.

• 1:02:08Speaker 04

  Oh, yeah, they actually have everything ready.

• 1:02:09Speaker 00

  Voice ready, I don't think they have text and speech ready.

• 1:02:11Speaker 04

  Oh, yeah.

• 1:02:11Speaker 00

  It speaks volumes that when Sam Altman announced Whisper three model.

• 1:02:16Speaker 04

  Yeah.

• 1:02:17Speaker 00

  No claps.

• 1:02:19Speaker 00

  It's the smallest news.

• 1:02:21Speaker 04

  But it is actually gonna be huge. I, I actually, I would love to, you know, put my hands dirty on Whisper.

• 1:02:27Speaker 00

  Yeah, so honestly, I'm just overwhelmed. I, I, I know some of the team, I know they've been working extremely hard. This is their sprints until, to, to get everything done today.

• 1:02:36Speaker 04

  Yeah.

• 1:02:36Speaker 00

  So I, I mean, I think that's, that's very important one, that now just like, they just shipped everything. They just, they're, even though they're, even though they are like doing very well, they still push themselves extremely hard to, to be top of, and, and they're really earning their spot for, for developers and for the, the generals of the general AI market.

• 1:02:52Speaker 04

  And I hope they take some holiday after today.

• 1:02:55Speaker 00

  Yeah, yeah, yeah.

• 1:02:56Speaker 04

  Yeah, it's too much of updates.

• 1:02:57Speaker 00

  And, and then so the, the next interesting thing to me is that they are integrating, they're Sherlocking a lot of the startup features. So there are a lot of startups that are built on providing rag for people.

• 1:03:08Speaker 04

  Yeah.

• 1:03:08Speaker 00

  There are a lot of startups that are built, built on like maybe bullying agents on top of GPT.

• 1:03:12Speaker 04

  Yeah.

• 1:03:12Speaker 00

  So this is the first time where, you know, I think it's pretty common in large platform companies, like AWS Reinvents often does this as well. They call this a red wedding. Like they invite all your customers to the same room, and then they're like, all right, let's see who, who survives, you know, stats, that's that.

• 1:03:25Speaker 04

  Oh, my God.

• 1:03:26Speaker 00

  So that is the sort of meme-y, funny, jokey version of this.

• 1:03:31Speaker 04

  Yes.

• 1:03:31Speaker 00

  I, I, I don't, I mean, realistically, I'm sure Harrison and Jerry and all the other rag people, they, they had some heads up about all this stuff going on.

• 1:03:38Speaker 04

  Yeah.

• 1:03:38Speaker 00

  But I think. Because it's built in so easily into the playgrounds,

• 1:03:41Speaker 04

  Yeah.

• 1:03:42Speaker 00

  Into the API into ChatGPT itself.

• 1:03:45Speaker 04

  And also the tools, all the integrations, right?

• 1:03:47Speaker 00

  You don't need a lot of tooling just to set up a simple chatbot.

• 1:03:50Speaker 04

  Oh, yeah.

• 1:03:50Speaker 00

  With RAG.

• 1:03:50Speaker 04

  Yeah.

• 1:03:51Speaker 00

  And it's like, so for example, for my conference, we did a Summit AI Bot.

• 1:03:55Speaker 04

  Hmm.

• 1:03:56Speaker 00

  All right, where we did, where we set up a LangChain stack, we integrated it, put it on a widget on the website. Now you can set it up with no code.

• 1:04:03Speaker 04

  Yeah.

• 1:04:03Speaker 00

  In inside of the playground and it just let people play with it. It's great, but it's also very scary for a startup because if that was your whole mode, you don't have that mode.

• 1:04:12Speaker 04

  I agree.

• 1:04:13Speaker 00

  Yeah.

• 1:04:13Speaker 04

  Yeah, yeah, that's gonna be a problem.

• 1:04:15Speaker 00

  So it's, so it's interesting that like I can sort of easily build this in and, and, and obviously the staple API is something I was considering building. And I, and I roughly knew that like this would be the next thing.

• 1:04:26Speaker 04

  This is on the critical path for sure.

• 1:04:27Speaker 00

  So I don't build it.

• 1:04:28Speaker 04

  I agree, yeah.

• 1:04:29Speaker 00

  But then the question is like, all right, what do startups do?

• 1:04:32Speaker 04

  Yeah.

• 1:04:32Speaker 00

  I think maybe one thing that was missing from Sam was like, hey, this is the biggest gathering of all your ecosystem developers. They're, they're afraid of you. You have, you have given them no assurance as to like, where do you think people should build?

• 1:04:45Speaker 04

  Okay.

• 1:04:46Speaker 00

  So, because like OpenAI just wants to do everything.

• 1:04:49Speaker 04

  I think so, right? Like judging from today's trend, they literally are doing everything.

• 1:04:52Speaker 00

  Yeah.

• 1:04:53Speaker 04

  Yeah, you're right.

• 1:04:54Speaker 00

  So, so I feel a little bit, I mean, it's fine. Everyone who's building with AI today opted in to cutting edge in some sense. You work on a cutting edge, you bleed.

• 1:05:03Speaker 04

  Yeah.

• 1:05:03Speaker 00

  And that's okay.

• 1:05:04Speaker 04

  That's right, that's right.

• 1:05:05Speaker 00

  Um, but, but I do, I do feel like there's a lot of tension between the startups that build on OpenAI and OpenAI itself.

• 1:05:12Speaker 04

  Yeah.

• 1:05:13Speaker 00

  So that's why too safe.

• 1:05:14Speaker 04

  Sounds great. It's great to see you.

• 1:05:15Speaker 00

  Yeah, good to see you. Thanks for jumping on.

• 1:05:17Speaker 04

  Thanks for having me.

• 1:05:18Speaker 00

  That's it. And next we catch up with the former guest, Raza Habib, back for his second time on the pod. Last time we talked about human loop and we recorded in London and that was a pretty popular episode. And I love that you guys care about foundation model ops as Raza puts it. So check out the human loop episode if you want, but also here's Raza's take on OpenAI Dev Day. Welcome back to the pod. This is the second appearance.

• 1:05:40Speaker 04

  It's always, it's always a pleasure. Nice to see you again, Sean.

• 1:05:42Speaker 00

  Uh, good to see you as well. Um, right, let's just get right into it. What was most interesting to you?

• 1:05:47Speaker 04

  I mean, the sheer density of announcements, I actually, I came with high expectations and there was a lot of stuff I was hoping to see. But I think they over, they under-promised and over-delivered, which I thought was really good. I think seeing that they're having a second run at plugins and doing it right this time and having the GPT store and like really allowing people to do that, I thought that was really cool. Product decisions around how you design and build the GPTs, like the low-code builder for these chat agents? I thought that was really nicely done, that they have this conversational interface that elicits from maybe someone who's not very expert.

• 1:06:19Speaker 00

  Yes.

• 1:06:19Speaker 04

  How to do prompting and things like that. I thought it was really thoughtful.

• 1:06:22Speaker 00

  It fills out the form for you, right?

• 1:06:23Speaker 04

  Yeah, it's a very simple thing, right? Like ultimately, it's just filling out the system prompt and filling out what abilities it should have.

• 1:06:29Speaker 00

  Yeah.

• 1:06:30Speaker 04

  But, but actually, despite its simplicity, I think it's very powerful and I was impressed by that. So yeah, a lot of really cool things and then all the changes to the API I'm really excited about. I have some questions, like I'm, I'm not, I'm not uniformly positive about all of the new API things, but I'm sure they'll get there.

• 1:06:46Speaker 00

  Okay, what, anything in particular that you wanna touch on?

• 1:06:49Speaker 04

  Yes, I think like things that I'm excited about with the new assistance API or like the new APIs in general, like multimodalities really cool, longer context windows really cool, cheaper, faster models, like I think everyone's gonna be super excited about that. JSON mode is like, it seems like a small feature, but actually so many people say this is a problem for them, so I think that's gonna be great.

• 1:07:08Speaker 00

  It's, so I, I, I maybe missed the importance of this. Isn't that the same as the function calling API?

• 1:07:14Speaker 04

  It's related, but you might wanna have it in context where it's not like strictly doing function calling.

• 1:07:19Speaker 00

  Ah, right, okay.

• 1:07:21Speaker 04

  Um.

• 1:07:21Speaker 00

  So a little bit more general.

• 1:07:22Speaker 04

  A little bit more general.

• 1:07:23Speaker 00

  Typically, I'll just make up a function that doesn't act, isn't actually a real function that.

• 1:07:27Speaker 04

  Yeah, even then, people say that for complex things, sometimes it violates the valid JSON things. I think just making that more reliable.

• 1:07:33Speaker 00

  Okay.

• 1:07:34Speaker 04

  Some stuff that I thought was initially I was excited about and then as I've like, like chewed on it a bit more, I'm a little bit less clear. So one is this like ability to jump in a bunch of documents and have it do a rag for you.

• 1:07:44Speaker 00

  Yeah.

• 1:07:44Speaker 04

  I think like.

• 1:07:45Speaker 00

  Twenty documents max or something.

• 1:07:47Speaker 04

  Yeah, I think that like, it's, it's a cool feature, but it feels a bit gimmicky to me. Like it feels like for serious practical applications, it's gonna be hard to get that to work. If you think about what a large enterprise needs for a rag, like it's, you know, it's rarely sufficient that you could just jump in a bunch.

• 1:08:01Speaker 00

  Dump in a bunch of.

• 1:08:02Speaker 04

  Dump in a bunch of documents. How it then matters. There's usually permissioning.

• 1:08:05Speaker 00

  Yes.

• 1:08:05Speaker 04

  As like which users can actually access which bits of data, like.

• 1:08:08Speaker 00

  Yes.

• 1:08:08Speaker 04

  There's so much control that I think most developers would wanna have for serious applications that I think it's cool for the like GPTs in the low-code version. I'm scared. skeptical that it'll get that much use by by serious developers. And I feel the threaded stateful like assistance API is really awesome. But I would like more clarity over how it's doing the like state keeping. Like what ends up in the context? I think I think for that to be really popular, they need to make that transparent.

• 1:08:34Speaker 00

  Yeah, there's an API booth downstairs. I don't know if you've.

• 1:08:36Speaker 04

  I've spoken to them, they wouldn't answer any of these questions from him.

• 1:08:39Speaker 00

  Okay, of course, but you know, obviously that really affects human loop.

• 1:08:43Speaker 04

  But this is, you know, this is commentary over what I think overall was a set of really exciting announcements.

• 1:08:48Speaker 00

  Yeah. And and the last time we talked also, you were talking about, we were talking about the multimodal APIs. And now you have them.

• 1:08:53Speaker 04

  Finally here.

• 1:08:54Speaker 00

  What what happens now?

• 1:08:56Speaker 04

  As I as I said to you when I spoke to you last time, right? Like, it's a relatively straightforward addition to the human loop product. Like, everything will continue to work, but now you'll also have images in and images out and audio in and audio out. It's kind of interesting, like, seeing, you know, the assistance playground for OpenAI that they just released and things like that. Like, it feels like they're starting to get close to supporting all of these things, but not quite yet.

• 1:09:18Speaker 00

  Yeah, yeah, excellent. And then I think the last part is I saw human loop, actually, probably probably not you, probably somebody else, but also talking about the fine tuning. There was a price drop. I don't know how much because there was just so many announcements. I imagine that's only only good things for fine tuning.

• 1:09:32Speaker 04

  Yeah, I mean, there are so many, I also missed the the price drop. But I know from speaking to folks at OpenAI as well, that they think a lot more people should be fine tuning. That fine tuning is going to have like huge importance in the future. That's why they're building out the EY for it, that, you know, so that's something they're investing in very deeply. And, yeah, I still view fine tuning as like an optimization step.

• 1:09:51Speaker 00

  Yeah.

• 1:09:51Speaker 04

  I think of it as like the compilation you do, like, once you have something that's working.

• 1:09:55Speaker 00

  Which is what is what they said in the other performance session just now.

• 1:09:58Speaker 04

  Okay, cool.

• 1:09:59Speaker 00

  Yeah.

• 1:09:59Speaker 04

  I'm glad I'm glad that my tips are aligned with opening up.

• 1:10:02Speaker 00

  I think I think you're very aligned. You're often leading them in what they say publicly, which I think is good.

• 1:10:09Speaker 04

  Yeah, what about you, Yushon? What did you think?

• 1:10:11Speaker 00

  Oh, I've said this in a previous recording, but effectively, I also thought they would do much less than they did today. I think they under-promised and over-delivered, exactly like you said. And like even things like text to speech, which would have been hinted at.

• 1:10:25Speaker 04

  It's not just text to speech, it's really good text to speech. So I, like, I think I told you last time, I did like a near a year-long internship at Google and I was working on the first neural TTS team. Like, the Tachyon team there were amazing.

• 1:10:37Speaker 00

  So what did you get from their demo?

• 1:10:39Speaker 04

  I think I need to play with it more, but I was impressed by the quality.

• 1:10:42Speaker 00

  Yeah.

• 1:10:43Speaker 04

  Like, the quality of the process, the variation. I think they're only releasing six voices, but...

• 1:10:48Speaker 00

  A secret seventh voice with the pirates.

• 1:10:49Speaker 04

  A secret seventh voice with the pirates. And then I was chatting to Andre just now.

• 1:10:54Speaker 00

  Yeah.

• 1:10:54Speaker 04

  And he was saying that internally, like, they have voice cloning set up as well. So they can, they can do it with something like 30 seconds of speech.

• 1:11:00Speaker 00

  I'm not sure that's public.

• 1:11:01Speaker 04

  It's not public?

• 1:11:02Speaker 00

  I don't know.

• 1:11:02Speaker 04

  He didn't tell me it wasn't public. I don't know.

• 1:11:03Speaker 00

  Okay, alright, alright.

• 1:11:05Speaker 04

  But maybe, maybe filter it out when you publish this.

• 1:11:09Speaker 00

  For what it's worth, I've been talking to a lot of people in and outside of Dev Day, and a lot of people have heard about the voice customization stuff. So it's not really going to get anyone in trouble, I don't think. So I just chose to leave it in there. I, whatever. I mean, it exists elsewhere in other products. And I think it's fair play to compete with other companies.

• 1:11:29Speaker 04

  I don't know if they're gonna release it, for obvious, for obvious reasons, right? There's a lot of safety concerns about releasing that kind of product.

• 1:11:36Speaker 00

  The, and for what it's worth, someone else, I think Fixie AI did a comparison of the pricing. They are severely undercutting like Play HT and some of the other text-to-speech companies as well on the pricing. They're between three to 10 times cheaper per second or something than, than the other existing TTS companies.

• 1:11:53Speaker 04

  Yeah, I think that's very interesting. I think in general, their promise to keep cutting prices and then following through is building a lot of confidence.

• 1:12:00Speaker 00

  Yeah.

• 1:12:01Speaker 04

  People who were previously nervous about building on them. What's interesting, I think, is that as the, like, because they have such a large economy of scale and they continue to drive down prices... The option of like self hosting a fine tuned model even for smaller models starts to be like less obviously economical because of the like spin up and spin down costs. So unless you have the like volume of usage to justify having it on all the time, it actually starts to become cost competitive to use one of these third party APIs rather than having even a smaller model.

• 1:12:31Speaker 00

  Right, because it's serverless in a way.

• 1:12:33Speaker 04

  Yeah.

• 1:12:33Speaker 00

  So what can you give people an idea of what kind of volume that that is? Are you talking about concurrent requests?

• 1:12:39Speaker 04

  It's so if you look at most of the people who will provide you and like a a served model, if you look like a replicate or a Mystic AI or something like this.

• 1:12:46Speaker 00

  Yeah, fireworks.

• 1:12:48Speaker 04

  Fireworks, a few of these companies, they tend to actually charge by like compute hour or compute minute.

• 1:12:54Speaker 00

  Yeah.

• 1:12:54Speaker 04

  And so if you're not like gonna have it on all the time, then like the reason.

• 1:12:59Speaker 00

  It's dollars.

• 1:13:00Speaker 04

  The reason, yeah, you end up needing it on all the time, though, because there's like spin up, spin net, cold starts. And so if you if you don't actually have enough usage to justify having it on all the time, it starts to become cost competitive to just use OpenAI.

• 1:13:12Speaker 00

  Yeah, so what I'm trying to get to is it's it's just dollars, though, like it's if it's like five dollars an hour.

• 1:13:17Speaker 04

  Yeah.

• 1:13:17Speaker 00

  Whatever, like.

• 1:13:19Speaker 04

  Yeah, I agree.

• 1:13:22Speaker 00

  Okay, got it, got it. All right, cool. Well, thanks so much for jumping on. I know this is last minute, but it's just nice to see people.

• 1:13:27Speaker 04

  No, no, I always I always love chatting with you, so hopefully there'll be more of this in the future.

• 1:13:31Speaker 00

  Yeah, for sure. The next guest is going to be a new name to many people. He hasn't done many public appearances, but he is a force to be reckoned with on Twitter. His name is Surya Dantuluri, and this is a story of somebody whose startup got killed by Sam Altman. So we're here with Surya. Hey.

• 1:13:47Speaker 05

  Hello, my name is Surya.

• 1:13:48Speaker 00

  You're new on the pod, but also we've been around each other in the tech circles for a little bit. You're you're a very famous developer of vector databases and of plugins.

• 1:13:58Speaker 05

  Yeah.

• 1:13:59Speaker 00

  What what what are some of the plugins that you've done?

• 1:14:01Speaker 05

  Yeah, so I worked on a few plugins. I worked on like chat with PDF, chat with like video, chat with website, chat with like Git. I made, yeah, a lot of cool plugins.

• 1:14:10Speaker 00

  Making decent money, too.

• 1:14:11Speaker 05

  Yeah, I mean, you can, they give like better functionality to like the whole GPT four interface. Initially, I wanted to do my homework with them, so I'm like, I might as well make a plugin for it for it. So. Yeah, I mean, they give, there's a lot of co-functional. I made one with called chat with like instructions, which would allow you to save more and more custom instructions and use that when you're talking to GPT four. But yeah, I mean, they're making revenue. It's pretty, pretty sick for, you know, people paying in eighty five different countries. It's like nuts. How many people are like, or how many, how big the, the scope is, or how many people can use it?

• 1:14:46Speaker 00

  I think you may have shown me this before, but there was a plugin platform that you use for monetization?

• 1:14:51Speaker 05

  No.

• 1:14:52Speaker 00

  No? Oh, you built your own? You built?

• 1:14:53Speaker 05

  I built my own thing, all custom.

• 1:14:55Speaker 00

  I've seen someone do like a Firebase for.

• 1:14:57Speaker 05

  And, you know, I, yeah.

• 1:14:59Speaker 00

  R I P.

• 1:15:00Speaker 05

  No, I mean, they're doing well, but like, I just don't wanna, you know, pay a ten percent tax and all that stuff.

• 1:15:05Speaker 00

  Yeah, yeah, for sure. Obviously, you're very technically savvy. Okay, so what happened today? Uh, they announced, they announced GPTs. What's going on?

• 1:15:13Speaker 05

  Yeah, so like, I made a tweet this morning, being like, Sam, I'm gonna kill my startup. And a joke, okay? I just wanna talk, like, I was like, I was trying to notify people I'm here, and I just wanna meet up. I made a joke, and then a couple hours later, my friend Matt, he works at Julius, he showed me the new U I. I'm like, okay, cool. And he was, he was forced me to look at, look at it on my phone. I'm like, okay, sure, I'll, I'll pull it up. I pulled it up on my phone. And plugins were gone. Plugins were gone. You don't, you can't, I think you can go between models, so you can go between four and, and three. But the whole options of like code interpreter and like Dolly three and all this stuff, all those good stuff were gone from the UI. I think this is only if, this only applies for people who are here at the event.

• 1:15:56Speaker 00

  Yeah.

• 1:15:57Speaker 05

  I think they give access, or like the new UI to people here, and they also, but yeah, plugins were gone, and I'm like, oh shit. And I asked the person like, hey, like, where, where the plugin, where can I, like, where are the plugins? Like, where does it go? They basically told me like, you have to make a new GPT as a developer, and you can import your schema into the new GPT. And only that way can you, you know, kind of revitalize your plugin.

• 1:16:22Speaker 00

  But your existing users will be like?

• 1:16:24Speaker 05

  No, I think they're gone. I mean, I look, they're. I haven't looked at my stats today, but.

• 1:16:28Speaker 00

  Well, I mean, this is not widely rolled out yet, but when it, when it rolls out.

• 1:16:31Speaker 05

  When it rolls out, I'm pretty sure all of the.

• 1:16:34Speaker 00

  They have to discover you again.

• 1:16:35Speaker 05

  Yeah, they're kind of dead. I mean, there's like no way. I don't think there's a way to link them.

• 1:16:39Speaker 00

  Yeah.

• 1:16:39Speaker 05

  Like, there's like no way for the users who were using it previously to be using the new thing now. But I mean, it's a side project for me. It's like not like a full-time thing for me. It's a fun project to do and like it, it's like a nice, nice thing to work on. So I'm really bullish on, you know, the whole new G P thing. I think there are better abstraction. Yeah, I think G P's are, I was telling a few open engineers and I was like agreeing with them because like, I think G P's are a much better abstraction on what plugins were supposed to be. I think plugins kind of died on arrival.

• 1:17:09Speaker 00

  Well, Sam said they did not have P M F, right?

• 1:17:11Speaker 05

  Yeah, he said that a long, he started that, he said that like one plugin started.

• 1:17:14Speaker 00

  Yeah.

• 1:17:15Speaker 05

  So it's like pretty nuts. But yeah, I think, I think G P's are a better abstraction and I also love they're doing revenue share. So revenue share is also a good thing. Because like G, plugins were like a really weird way of monetizing. You had to do a bunch of finicky stuff, but yeah, I mean, also like just by the way, for people who don't know, Poe, you know Poe, right?

• 1:17:35Speaker 00

  Yeah.

• 1:17:36Speaker 05

  Poe did this a long time ago. They did this a couple of months ago. They helped, they have, they have these bots, they call bots. And you can, you know, make your own like poem bot, or you can make your own like essay bot or whatever. And then the bots have custom instructions. And also they use a very specific model that the developer specifies. And you can install these bots or you can chat with these bots and the bot will do whatever, whatever the developer made them to do.

• 1:18:01Speaker 05

  So I think they're just basically open, I just made the same thing and they brought it over to them. But yeah, but effectively plugin to kind of dead.

• 1:18:09Speaker 00

  Oh, R I P.

• 1:18:10Speaker 05

  I, I mean, R I P, but it was a fun project. I mean, it's fun. I think G P, I think G P honestly. It's good that the plugins died because like they had a bunch of issues. So one of the issues is that you can't share them. You can't share a link to them. GPTs, you could share a link to them. So like I could share my, my link to my GPT thing to you. So that's, it's much better for discoverability. Because previously, the only way to discover a plugin was through the plugin store. You just search for it, you have to do a bunch of stuff and it wasn't very good in that aspect. But sharing a link to them, having revenue share. And you can also like give custom instructions, custom context. So they also came out with like retrieval or whatever. And that can basically give you like a custom vector database directly in your, in your GPT, I think.

• 1:18:54Speaker 05

  Such all great, all good features that, that should have came with plugins probably, but.

• 1:18:59Speaker 00

  Yeah, awesome. And then lastly, just like any of the new stuff that was launched today. What, what interests you in sort of building with them? Like if, if you were to build on the new API or the new GPT.

• 1:19:09Speaker 05

  Yeah, totally. I have some ideas. The thing is like, this is really weird to say, but like, some of my ideas that I've said before for plugins, they kind of get copied quickly.

• 1:19:22Speaker 00

  So you want to keep it to yourself? Yeah, that's fine.

• 1:19:24Speaker 05

  Yeah, but that's one part of it. The second part of it, I don't have any good ideas regarding what you can do with all the new functionality. Like, like that's like a, that's like a good product. I don't know, honestly, there's like text to speech came out, the, their internal vector DB thing came out.

• 1:19:40Speaker 00

  Internal vector DB thing?

• 1:19:41Speaker 05

  Or like retrieval or whatever it's called.

• 1:19:42Speaker 00

  Okay, yeah, people have been saying they have an internal vector DB thing, but.

• 1:19:46Speaker 05

  Yeah.

• 1:19:46Speaker 00

  It's, it's just a retrieval. Yeah, it's like zero non-configurable, right? Like it's, it's gonna be the, for a simple use cases, it's fine. And then after a while, you're gonna need one to control over chunking and stuff.

• 1:19:56Speaker 05

  Yeah, I'm also excited by once we need a context window. I was a big user of Claude for a while because Claude, they basically gave you a hundred K context window directly on the UI. And you can upload your PDFs to it and everything would work very well.

• 1:20:08Speaker 00

  Yeah.

• 1:20:09Speaker 05

  But I think Claude had some issues regarding, I mean, actually very recently Claude came out with this whole bullshit thing, bullshit copyrighting thing.

• 1:20:16Speaker 00

  Copyrighting thing?

• 1:20:17Speaker 05

  Yeah, yeah, it's really weird. So if you upload a PDF now on Claude, or just this week, they made this weird tweak where it doesn't answer any questions. Because if there's a copyright symbol or a copyright name anywhere, it just like blocks you out. It's like, what?

• 1:20:33Speaker 00

  Apparently you can prompt inject that by insisting that you are the author and then it just overrides it.

• 1:20:37Speaker 05

  Oh, really?

• 1:20:37Speaker 00

  It's like, don't worry, I got this, I'm the author of this, there's no copyright issue.

• 1:20:43Speaker 00

  Anyway, so thanks. This is a really good story and I wanted to people to share it. Um, and excited for what you work on to become more public.

• 1:20:51Speaker 05

  Yeah, thanks, Rex.

• 1:20:52Speaker 00

  All right. So that's what happened to ChatGPT plugins, which we covered back in March. But don't worry, that's not the full story. His startup is not fully dead. We actually cover what happens later on. I just wanted to capture the confusion that was happening at Dev Day. So he referred to Julius and we'll actually talk in and check in with Rahul later on in this episode. But first, we have to go to our next guest. When OpenAI launched with GPTs and the Assistance API, one of the lead launch partners that they launched with was Zapier. And I managed to catch up with Reid Robinson, who is lead AIPM at Zapier, to talk about it. That.

• 1:21:28Speaker 02

  All right.

• 1:21:28Speaker 00

  Oh, Reed, nice to meet you.

• 1:21:29Speaker 02

  Great to meet you too, Sean.

• 1:21:30Speaker 00

  Really, really great to run into you as we're leaving. So you guys had a big sort of partnership launch on stage.

• 1:21:37Speaker 02

  Yes, yeah, we launched AI Actions for GPTs, which we're really excited to see out there.

• 1:21:42Speaker 00

  Yeah.

• 1:21:43Speaker 02

  We also today launched an update to our ChatGPT integration that supports the Assistance API functionality that was announced.

• 1:21:50Speaker 00

  And you were one of the earliest to go. In my mind, Zapier was very, very early in the Natural Language Actions, NLA. I don't, I don't remember.

• 1:21:57Speaker 02

  Good memory. Yeah, yeah, we launched our Natural Language Action. Actually, so we were a launch partner for ChatGPT plugins.

• 1:22:02Speaker 00

  Yeah.

• 1:22:02Speaker 02

  And that's when we launched our Natural Language Actions API. And actually, the AI Actions that we're calling it today is a kind of a, we're rebranding that side of things to really focus on, on that functionality.

• 1:22:12Speaker 00

  Yeah.

• 1:22:12Speaker 02

  For that.

• 1:22:13Speaker 00

  And I just interviewed Surya, who is one, who's a pretty prominent plugins developer. Plugins are dead.

• 1:22:19Speaker 02

  I, you know.

• 1:22:20Speaker 00

  Or reborn.

• 1:22:20Speaker 02

  Yeah, it's gonna be interesting to see what happens. And there's clearly a difference. I think one of the things I'd talk about is the fact that, you know, with GPTs, you're able to constrain the prompts quite a bit, like our plugin for ChatGPT, the initial one, you needed to give it access to every single action you ever wanted it to have access to, which meant that the kind of, you know, anybody who's familiar with context is sitting there like, yeah, that's gonna be an issue.

• 1:22:41Speaker 00

  Yeah.

• 1:22:42Speaker 02

  Common one I give is like, you know, if you had given it Gmail and Google Calendar and asked it like, hey, what's going on next week on my like agenda? It would sometimes search Gmail, because it'd be like, yeah, events are.

• 1:22:52Speaker 04

  Gmail or like, you know, calendar invites are going to go to Gmail, so I should search there. But now you can, you know, define what apps it should use. You can define like how it should use those. So some really fun use cases. I mean, honestly, we've been hustling hard to get this out there. I'm really excited to see what people actually build with this.

• 1:23:10Speaker 00

  Great.

• 1:23:10Speaker 04

  And what gets released there. Yeah, we'll be monitoring, trying to listen to people rhythmically.

• 1:23:15Speaker 00

  And so like, something that's interesting about Zapier is that you are a collection of actions in and off yourself.

• 1:23:21Speaker 04

  Yeah.

• 1:23:21Speaker 00

  So there's kind of multiple layers in which to do to do this, like what should exist at the GPT layer? What should exist at the Zapier layer?

• 1:23:28Speaker 04

  Yeah, well, what's not, what's nice? I mean, it's a good point. We have about 6,000 like apps on the platform today. Really what the AI actions is, is it's the ability to use any of those searches and actions using kind of a natural language inputs. That would be like the instruction that the model gives it. So it's like, you know, check this user's calendar for Monday and, you know, it might even give the, you know, the actual date for Monday, right? And Zapier on our side will take that natural language request and process that into an actual API, like the actual API call to a tool like a calendar. And then, and then we all move on the response. So, you know, you can't just take the entire response of a, especially like Gmail, Google Calendar, their API responses are very, very, very, very, very long and very confusing.

• 1:24:12Speaker 04

  And so we actually do a lot of work to kind of, if you will, like massage that data so that it makes sense for an LLM on the other side that it is giving it the right information it needs and not just like the entire payload.

• 1:24:24Speaker 00

  Right.

• 1:24:24Speaker 04

  So it really helps it just kind of deliver like a more, again, more like contained, more refined experience for leveraging integrations alongside, you know, like chat GPT.

• 1:24:34Speaker 00

  So existing zaps cannot be ported one for one over to LLMs apps?

• 1:24:39Speaker 04

  It's really one-off actions. It's a better way to think about it. And you could chain them together in the, so like you can, you know, as I saw in today's demo, you know, it was using Google Calendar for the search and a Slack action. You can actually chain those together. And so, you know, How much is that as like a one-off action versus an actual like all of a sudden as app? But in this case, it's almost more like the trigger is the human in ChatGPT, right? Like you need to trigger it to run for that. But on the flip side, you know, the assistance API is extremely exciting for me as well because you look at now like the that functionality of building a GPT allows you

• 1:25:12Speaker 00

  Still getting used to the name?

• 1:25:14Speaker 04

  Allows you to kind of port that over to run asynchronously. So a common one, like the two examples that I love giving for that API that I love in Zapier is number one, like data export. You know, think of every tool out there like Looker, Mixpanel, Amplitude, all so many tools are able to send these like massive exports of CSV data on a regular basis. Like you could say, hey, every Friday, export my blog traffic content or see it as a CSV, right? Normally someone's going to get that CSV and have no clue what they're doing, right? But now you can actually create an assistant in Zapier and you can give it instructions to say like, hey, tell me the top 10 performing blog articles in the last week. And also, you know, tell me highlights on, you know, maybe keywords that were used or SEO tags that were used and how that impacted conversions, right?

• 1:25:59Speaker 04

  Like you can be pretty detailed depending on what you're providing it. And that can now run asynchronously. That can run automatically. So every Friday, you know, 8 AM, you could be getting the export of that data. It's going to go to an assistant. That assistant is going to reply with even charts and graphs. And those will come through and you can then send it to Slack. And so you can have every Friday, a conversation, a post in your team's, you know, the blog team's Slack performance.

• 1:26:22Speaker 00

  Yeah.

• 1:26:23Speaker 04

  And that'll run automatically. And then they can even reply in Slack to that post and have a continuous conversation with that assistant.

• 1:26:31Speaker 00

  Oh, my God. So it's like really everywhere.

• 1:26:33Speaker 04

  Yeah. So you can really put them everywhere. And that's, so that's one of the things I like about what's released. And I think people are going to continue to learn really just how kind of wild that is. It's the fact that you can like use your actions in the UI of ChatGPT and a one-off action, but you can also run these things extremely well asynchronously. And yeah, like OpenAI releasing API support for the vision model and for code interpreter and retrieval that these assistants can use is really cool.

• 1:27:01Speaker 00

  Yeah. Is there Zapier angle to any of that? Or they're all the same, right?

• 1:27:05Speaker 04

  Like you would do in Zapier, right?

• 1:27:06Speaker 00

  They're all the kind.

• 1:27:06Speaker 04

  The you can the whole like creating of an assistant and running that through an assistant is today. You could do that literally right now.

• 1:27:12Speaker 00

  Yeah.

• 1:27:13Speaker 04

  So it's really cool. And the other one is like retrieval, right? I talk about, you know, you could go in, create an assistant, give it, let's say, you know, I talk about our accounting team a lot, right? You could give it like if you have a team that approves budget requests from your company, right? Everyone does, right? They can actually have, take their Slack channel or create an assistant first that would have the documents of your policies of like, hey, here's what you can expense, here's how you can expense, here's eligible ineligible, right? All these sorts of things. And actually then set up something like, again, I'll pick on Slack, it's just easy. Like a new message in your accounting budget request channel, right? And have it trigger a the assistant and send the user's request to the assistant with all of your documentation with retrieval.

• 1:27:54Speaker 04

  And now it'll try to understand what your policies are, what everything is, and check the information against what that. And you could even like I did one internally where we have a tool called, I think it's called Stacker, that tracks each employee's like software budget and home office setup budget, right? You could see how much they've spent of their budget. You could actually include that data in the context of the user message. So that the model will be able to say like, hey, I see you want to expense this webcam. It's actually over the recommended budget, but you personally do have budget left if you wanted to use it for that, right? And so like, yeah, and that's really cool.

• 1:28:30Speaker 00

  Yeah.

• 1:28:30Speaker 04

  So you could start to do all of those sorts of things now in Zap's that really were never possible.

• 1:28:36Speaker 00

  Yeah.

• 1:28:36Speaker 04

  So yeah, the querying of knowledge, running of data analysis, writing code even.

• 1:28:41Speaker 00

  I think in a very real way, you are the perfect partner to OpenAI because they they've sort of built a reasoning sort of glue between all these things.

• 1:28:48Speaker 04

  It's it's definitely been a good and fun partnership. I think, yeah, the big thing for me that I always say is like, I just, I'm really, really excited now to just see what people do with this and how we can improve it.

• 1:28:58Speaker 00

  Yeah, awesome. Is there anything, you know, you've been developing with these APIs for a while. Is there anything that you caution people not to get too excited about? Like what, what, yeah.

• 1:29:06Speaker 04

  I mean, yeah. Callouts I'll always make is like double check accuracy, right? Like you want. I want to call out like, okay, like how accurate, so make sure that information is accurate. Make sure you're putting some human in the loop steps before you're putting this into like a critical.

• 1:29:19Speaker 00

  Which they showed and like confirmed and I.

• 1:29:21Speaker 04

  Yeah, that sort of thing, but even, yeah, all sorts of things. You really want to make sure that you're comfortable with like, what can go wrong, what is likely to go right, right? Like all those sorts of constraints. The other side that I often talk about is this, like, keep an eye on, you know, if you have free form human input somewhere in your application that is triggering these things, now that can sometimes be risky, right? Yeah, prompt injections. Those are a real thing and I, I think, you know, a lot of people are still trying to really what that means and how bad it can be.

• 1:29:48Speaker 00

  Yeah.

• 1:29:48Speaker 04

  And so I always try to caution people about that as well, right? Like you really want to be realistic on kind of how far reaching.

• 1:29:55Speaker 00

  Yeah.

• 1:29:55Speaker 04

  You're doing this. So, yeah, that's why I like, like internal use cases that, you know, like things like that is a great way to start.

• 1:30:01Speaker 00

  Yeah.

• 1:30:02Speaker 04

  To get familiar with the technology, to get familiar with the constraints for that. Other than that, no, I mean, the voice model stuff, I'm really excited to try that.

• 1:30:10Speaker 00

  I really wanted to.

• 1:30:10Speaker 04

  Yeah. Yeah, that'll be really.

• 1:30:12Speaker 00

  I love this secret pirate mode that they demoed. I don't know if you caught that session.

• 1:30:17Speaker 04

  I didn't see that session.

• 1:30:18Speaker 00

  So they, they say, obviously there are six voices, but there's a secret seventh mode. If you add in a prompt to speak like a pirate.

• 1:30:24Speaker 04

  Love it. Love it. That was an old, I don't know if you remember Facebook way back in the day had that as one of the languages you could select.

• 1:30:31Speaker 00

  Yes.

• 1:30:31Speaker 04

  Yeah, yes. That reminds me of that. Yeah, yeah.

• 1:30:34Speaker 00

  Yeah, lots of fun to be had with, with AI as well. And okay. Well, thanks so much for jumping on. I know it's a very random, but also. Yeah, people love to hear from builders. So that's awesome.

• 1:30:43Speaker 04

  I, I love hearing from builders.

• 1:30:45Speaker 00

  And most of the interviews were done as we were sort of leaving the Dev Day venue and going to the after party. And I caught Dev Garg of Multion, who we've been talking around and circling around a possible episode on. He's definitely one of the leading voices and thought leaders on agents because he's building a browser agent. That's a very prominent one. Unfortunately, I have to take an L on this one because the audio is not great. Dev's mic wasn't working and I don't know what happened to it. I, I try to always check these things, but you're only going to hear the output from my mic, which is slightly worse, but I opted to leave it in because Dev is actually building an agent with OpenAI stuff and had access to GPT-4 Vision.

• 1:31:26Speaker 00

  And I think that people building with GPT-4 Vision will be surprised at his answer to me on whether or not it's useful for agents. Good to meet everyone. I'm this, founder of Multion, which is an AI web agent that can automate browsing for you. We can book your flights, order stuff on Amazon, order dinner, whatever you can imagine. Yeah, and I was actually reflecting, so everyone who listens to this already knows what was announced. I was actually reflecting that they didn't have any browser-based actions. So what were your thoughts on just generally their approach to agents? So they, it'll be very interesting because I feel like browser actions are just so risky. So and like things can go wrong. So if a company or you're open here, you won't, you won't want to build that. And they are like better off just like relying on a third party to who like who wants to own that. And that's also the strategy where we are taking with them. They're like, like, like open and launch like a G P T integration for A P I's.

• 1:32:12Speaker 00

  But we want multi on to be like the no A P I solution. Like, I want to do things beyond A P I's. I want to connect to my personal accounts where I just have my logins already or I already have the cookies and I want to want to go and like interact with my personal accounts or personal data very easily. And I think this will be very fascinating for us where we can like launch a multi on integration with their new platform. And then you can just go and like give it a command like, oh, like, can you book this platform for me on chat G P T? And then to launch a browser and the browser, you can see what's happening and then go do the whole thing for you. And we all seamless and then people can have a lot of fun. It's like trying out all these different capabilities and like automating their like daily workflows to like save this as custom integrations for different agents. You can have different custom like multi on prompts that are that are already like pre-saved.

• 1:32:54Speaker 00

  And then you were like, oh, I want to now go order something on like a DoorDash. I want to order my favorite burger. Then like chat G P T can go and like suggest you order a favorite burger style. And then I was like, you can like now order this for me. Multi on and multi goes and like she does that and buys it. So we solve the payment for you. We solve identity for you and like we are owning all the risky like actions that can be. So, so you're, you're gonna build a G P T version of multi on? Yeah, we'll have a multi on G P T. You, you, okay. Would, would that be like a replacement to your existing thing or just like an alternative way to use your same A P I's or something like that? So it's like the direction we're going for is we want to make our A I like agent embeddable within, within existing applications. So our launching an A P I. Okay. And we already have a like a chat G P T plugin. And so this will be like sort of like I will like use the A P I to power this sort of like new G P T experience.

• 1:33:39Speaker 00

  So for us, we actually don't have to like change anything. It'll be like very streamlined, just integrate our A P I into chat G P T and like we can start using it. Yeah, yeah, awesome. What about the, I guess the vision A P I? I think one of the things that have always constrained browser agents is the DOM. Right. Which is very heavy. Yeah. And so the alternative approach is to use vision. Would you explore that? What are your thoughts? So for us, we actually had like early access to the. Vision API for more than a month. Oh. We tried it on a bunch of websites, maybe like five percent of the websites are actually really useful. Ah. Which are more like image heavy because ninety five percent of the websites you can just, even if you do OCR, that's good enough. Yeah, it's not in the data set. Really good like parsing. So most websites we can compress less than three K tokens. So we are not, we don't really have to like worry about how heavy the text is.

• 1:34:20Speaker 00

  We, so we had one interesting use case about the vision API. We had a user who got it to work on Tinder and then like the, then like multi. Hot or not?

• 1:34:33Speaker 00

  Left and right. The user actually got a match. Yes. I think you have found the killer use case from multi on. Yeah. Like this.

• 1:34:44Speaker 00

  Oh my God. Okay. Interesting. Interesting. Okay. But so, but only image heavy sites. That's surprising to me. Because you, you know, the original vision demo, they actually showed a screenshot of Discord and they had perfect OCR. Yeah. It's true. It should be good for you. It can be very interesting, but I think it's like even without vision, we can just do like so much things. Yeah. So like adding vision maybe like helps a bit, but not, it's not like really game changing for us right now. That's surprising. Okay. Well, good, good to know. Anything else that you would highlight from today? I'm just like really excited about like opening a time to become like a marketplace. Yes. App store. Yes. So if this can take off, they could potentially kill like Apple App Store and become like the new thing there.

• 1:35:26Speaker 00

  And it's really hard to say like how things will go, but they tried this with plugins before, but this like, this might actually work this way. But it'll just be interesting to see like how two years from now, how a lot of the development might like, how the world looks like. Yeah. I'm very excited about like two years from now, like I think everything will be so different. We might not even use computers or even like mobile phones, even if our system, you talk to it and the system goes and does everything. Yeah. It'll be a fascinating world. Yeah. So one last question before we go. You have a nice side gig teaching at Stanford. While you, you were a PhD student and you put on, but you, you, you're still teaching or curating Transformers United. Yeah. So I dropped out from the PhD, but I'm still a lecturer at Stanford.

• 1:36:03Speaker 00

  Yeah. Okay. So like, what paper should people read to like, like catch up on this? Like what, what, what is like top of mind in terms of like research that is informing what we're seeing? Yeah, that's definitely very, it's a good question. This thing's coming to be so fast and there's like hundreds of research papers coming out like literally like in a few days. Yeah. I'm really excited about like developments that are happening at like Meta. So a lot of this work is open source on the Lama stuff, all the Mistral stuff. I feel like that's very interesting. On the transformer side. Do you believe sliding window attention was the key for Mr. All? I I feel so for them, but I feel like there might be other ways to do it. There's some secrets, right? There was probably some secret. Yeah. Okay. Well, that's all the time we have, but thank you so much. Thanks a lot. Thanks.

• 1:36:42Speaker 00

  Okay. And our next guest is Louis Knightweb, CEO and co-founder of Bloop AI and organizer of the AI meetups in London, where he is a very prominent and staunch member, unlike Raza, who has defected to San Francisco since our last conversation. Louis always has very interesting takes in person, and it was a pleasure to finally actually get him to come on the pod. But also, we recorded this while inside of a Waymo on the way to our after party. So Louis, you are new to the pod, but we've been friends for a while. Maybe explain, maybe introduce yourself and how you come to the world of AI.

• 1:37:16Speaker 06

  Yeah, I guess. So we started Bloop, me and my co-founder, three years ago in a very different era for for machine learning. And we both started the company because we wanted to help engineers navigate large code bases in a much better way.

• 1:37:35Speaker 00

  Yeah.

• 1:37:36Speaker 06

  And originally, that was training our own models to do natural language code search. And today, we still do that, but obviously, those language models are very small compared to the state of the art.

• 1:37:48Speaker 00

  Yes.

• 1:37:49Speaker 06

  And so they're just one part of a much bigger pipeline.

• 1:37:52Speaker 00

  I I see you as a very astute technologist. You used to be a VC. You wrote the first check into Human Loop. And you used to share an office with Human Loop, to the point that I called it Human Boop.

• 1:38:03Speaker 06

  Yeah.

• 1:38:03Speaker 00

  I think you like that.

• 1:38:04Speaker 06

  Yeah, I did. Yeah, that is good. We're considering renaming.

• 1:38:09Speaker 00

  And you also run AI tinkers in London.

• 1:38:12Speaker 06

  I do. Yeah. London has a kind of a slightly different mix of of talent than, say, San Francisco. You've got a lot of agencies, a lot of enterprises. And so, yeah, we we just felt a need to start like a very startup focused event. And that's why we we created AI Tinker at London.

• 1:38:28Speaker 00

  Yeah, I think Alex gravely would be very happy to hear about the other stuff that you've been doing. And I've been to one of them and it's really good work. I might be the only one who's been that's been to both.

• 1:38:37Speaker 06

  Yeah, yeah.

• 1:38:38Speaker 00

  Maybe.

• 1:38:38Speaker 06

  I've been to both as well.

• 1:38:41Speaker 00

  Okay, so let's fast forward to today. A whole bunch of things was announced. What's top of mind for you?

• 1:38:45Speaker 06

  Yeah, so I think like context length is something that that we.

• 1:38:50Speaker 02

  Spend a lot of time evaluating whenever something new drops. All of the kind of standard evals, you know, the kind of literacy tests, things like that. They they generally don't do a good job of measuring whether a model can actually use the context length that it that it claims it has.

• 1:39:09Speaker 00

  Yeah, context utilization is what I saw Will Depew today call it.

• 1:39:13Speaker 02

  Exactly. And so this basically started maybe five months ago over the summer when Claude 2 dropped. And, you know, obviously it had a hundred K context and we were really excited about that. So we ran an experiment to see, basically, if we hid ten pieces of information in the prompt and we increase the size of the prompt, you know, so you did a one thousand tokens, four thousand, eight thousand, et cetera, up to a hundred thousand. How many of the original ten pieces of information can it retrieve? And we essentially found that the accuracy drops off a cliff between one and ten thousand tokens. And so and we repeated the same experiment with G P T four and, you know, we found similar results

• 1:39:53Speaker 02

  that thirty two K G P T four can only find one of the ten pieces of information. But if you were only using a thousand tokens, it can find nine of the pieces of information. So what that tells us is that, you know, context utilization five months ago was was was not great with with all of the state of the art models. So with the announcement of a hundred and twenty eight K today and.

• 1:40:15Speaker 00

  That's the first test you'll run?

• 1:40:16Speaker 02

  That's the first test I'll run. You know, having spoken to a couple of the team members who do eval today from OpenAI, you know, they're pretty confident that the model's got better ability to to answer questions at those context lengths. So it's time to time to measure.

• 1:40:31Speaker 00

  Time to measure. Any other of the API features, reproducibility, does that matter to you?

• 1:40:36Speaker 02

  I think, to me personally, no, I I kind of like the creativity. I like I normally have my models at like, you know.

• 1:40:43Speaker 00

  Temperature.

• 1:40:44Speaker 02

  Yeah, exactly, a bit of temperature.

• 1:40:46Speaker 00

  Yeah.

• 1:40:47Speaker 02

  But I know lots of people on the Blue team who'll be very happy, I'm sure.

• 1:40:51Speaker 00

  And then I guess the JSON features, the, you know, I I there's so many, like the multimodal features. Any of that appeal for for you personally?

• 1:40:59Speaker 02

  JSON is is definitely a big one. I think it allows you to to kind of standardize. How you call different models?

• 1:41:07Speaker 00

  Yeah.

• 1:41:07Speaker 02

  So instead of having to build, you know, the, and it's not a massive thing to build, but to build the, the, the kind of function calling integration. And then if you want to try Anthropic, you've got to go and like, have a completely different way of interpreting the output. So if you can just stick with JSON across all of your different L M providers, open source models included. That's definitely advantageous, which allows you to evaluate different models more easily.

• 1:41:29Speaker 00

  Yeah.

• 1:41:30Speaker 02

  Yeah, very exciting about that.

• 1:41:31Speaker 00

  You are, you, so you compete in a, in a pretty competitive space with the code assistance, code search, code assistance, right?

• 1:41:37Speaker 02

  We do.

• 1:41:37Speaker 00

  There's Sourcegraph, there's Codium, there's other Codium, there's.

• 1:41:41Speaker 02

  Yeah.

• 1:41:42Speaker 00

  There's Copilot and, and so on. You've never ventured into the agent side of things.

• 1:41:47Speaker 02

  Yeah.

• 1:41:48Speaker 00

  Is that a conscious strategy? Are you waiting for the right time, waiting for the right APIs?

• 1:41:52Speaker 02

  I think, I mean, we're, we're seeing traction at the moment with companies that have very large code bases, right? And it's not something we hear from those users that, you know, when we, when we listen to their problems, it hasn't been like an obvious fit to try and build. Like maybe an auto G P T type of agent. I'd still say, you know, we're, we're very interested in agents, the pipeline we have at the moment. Basically G P T in a big while loop with, with function calling, which, you know, like nine months ago, definitely did count as an agent, maybe less so now. So, you know, it's just, it's just customer and problem driven and we don't, you know, it's not a, it's not a hammer for the nails that we've, we've got.

• 1:42:33Speaker 00

  Yeah. So two comments on that. One, I think OpenAI has sort of put their flag a little bit in the definition of an agent. They had three things, right? They had custom knowledge.

• 1:42:43Speaker 02

  Yeah.

• 1:42:43Speaker 00

  They had custom instructions and then I forget the third one, custom tools, let's just say.

• 1:42:48Speaker 02

  So by that, by that definition, we're doing, yeah, so we've been doing that since, since about February. That's the, that's the definition.

• 1:42:57Speaker 00

  Then the second observation I would say is you talk to developers, but what if the target customer for agents is not developers? It's the P M's, right?

• 1:43:05Speaker 02

  So we, we definitely see a lot of P M's using, using the product or people that are defined as like reading more code than they write. So, you know, it could be designers trying to understand the implications of an interaction. It could be P M's trying to fact check a, a contentious time estimate from a developer or something like that.

• 1:43:25Speaker 00

  Like the low trust. There.

• 1:43:28Speaker 06

  Talking for I've seen some.

• 1:43:31Speaker 00

  Yeah.

• 1:43:31Speaker 06

  Seen some stuff.

• 1:43:32Speaker 00

  Egregious things. Yes. Yeah. So so basically it's still not that appealing for you, but you're you'll keep a lookout for it. Like the the staple.

• 1:43:40Speaker 06

  I I think I think based on the definition of an AI, you know, released today.

• 1:43:44Speaker 00

  You've been doing it.

• 1:43:45Speaker 06

  We we we tick all the boxes. And I think we were one of the the the earliest adopters of that. If if that's the definition.

• 1:43:51Speaker 00

  You just don't brand yourself with the agents.

• 1:43:53Speaker 06

  I don't think it's important to users. I didn't think I don't think that's why people use the product. I mean, we're very solutions focused. I think we we start a lot of our branding in at the start of the year. Was about models and and you know, we put GPT four, GPT three right there on the front page. And now, you know, we've we've kind of reoriented to be more about solutions. I think that that reflects kind of maturity of the the ICP we're going after and and where we are with with. Sort of stage of company life.

• 1:44:22Speaker 00

  Yeah, yeah, cool. Any other things that you personally, you know, not not loop related, are just excited by interested by from today? Any interesting conversations with others?

• 1:44:33Speaker 06

  Loads of really interesting ones. I had had a fascinating talk with some safety researchers who.

• 1:44:39Speaker 00

  They were here?

• 1:44:40Speaker 06

  They so there's is is a couple of people who were kind of PhD students who had kind of. Looked at adversarial attacks through fine tuning of models and found that. Basically, like it's such a hard problem to solve is if you enable fine tuning. It's basically impossible or very difficult to to make it. So that you can't disable all the safety features. You can just train it to spit out all sorts of stuff. So that was pretty fascinating. I'm excited about the Waymo we're in right now.

• 1:45:12Speaker 00

  Oh, yes. So we should tell people we're recording in a Waymo. Haven't been looking at the road the whole time.

• 1:45:17Speaker 06

  Is this your first Waymo?

• 1:45:19Speaker 00

  It is my first Waymo, actually. Yes. Thank you for taking my Waymo virginity, but.

• 1:45:24Speaker 06

  Got to experience this together.

• 1:45:25Speaker 00

  I've been a cruise stan the whole time until they ran over someone.

• 1:45:29Speaker 06

  So my so so my take on cruise like at sample size 10 cruise journeys before they got shut down. And three of them resulted in something popping up on the screen saying that I had been in a collision. And

• 1:45:44Speaker 00

  Did they use the word collision?

• 1:45:45Speaker 06

  Yeah, yeah, yeah.

• 1:45:46Speaker 00

  That's surprising.

• 1:45:46Speaker 06

  I'll show you after I got.

• 1:45:47Speaker 00

  I take a fair amount of cruises and it didn't, yeah.

• 1:45:49Speaker 06

  And so it was the same situation almost every time, which was a car was in front trying to park. And I think they just maybe bumped fenders or maybe the crash detector.

• 1:45:57Speaker 00

  Oh, there was actual contact.

• 1:45:58Speaker 06

  I think in one of the cases, I think there was, and the other two, I didn't feel anything. But it came up saying like, you've been in a collision and somebody comes over the intercom.

• 1:46:06Speaker 00

  Checks you're okay, yeah.

• 1:46:07Speaker 06

  So, yeah, I mean, but out of a, you know, ten, ten rides and three of them ended like that. So I think, yeah, some, definitely some questions there, but this way was pretty smooth.

• 1:46:17Speaker 00

  Maybe also we're in a better neighborhood for driving because we're going to Golden Gate.

• 1:46:22Speaker 06

  The time of day, that was, that's a really good point. I noticed that all of the ones I took at night, all of the cruises I took at night were fine. And when I took one during rush hour, it was a completely different experience. Cause the routes it would take, it had this really aggressive, maybe traffic management, something that was going on. So it'd take, take a long time to get from A to B.

• 1:46:41Speaker 00

  Yeah.

• 1:46:41Speaker 06

  Yeah.

• 1:46:42Speaker 00

  Yeah. I, it often puzzles me slash interests me that self-driving is almost solved. Uh, you know, we still have some bumps in the road, some, sometimes the bumps are human.

• 1:46:52Speaker 06

  It's solved in San Francisco where you've got wide open roads, nobody cycles and.

• 1:46:57Speaker 00

  That, that's, that's not true.

• 1:46:59Speaker 06

  Some, okay, some people cycle.

• 1:47:00Speaker 00

  I live here, excuse me.

• 1:47:02Speaker 06

  Some people cycle, okay, but I mean, compared to like, okay, compared to, compared to London where you've got, you know, roads half the size, built for horse and carriage and millions of cyclists and buses and all sorts. So I think, you know, it's, I think it's gonna be a long time until we have that same experience that of a, of a cruise or Waymo today, London.

• 1:47:24Speaker 00

  Well, I understand London's a tougher neighborhood, uh, but still, you know, we're eighty percent there, seventy-five, eighty percent there, whatever, right? But like, and it seems like the, the stuff that we do in the rest of our lives in terms of AI automation is so primitive compared to this. Which is the car that we're sitting in right now. And I find that weird. I find like the relative ease or the relative like heerness of this technology. Is very disparate, like, how come it didn't trickle down from self-driving to the rest of tech?

• 1:47:54Speaker 06

  Yeah, it's it's interesting, isn't it? Well, I don't know how those pipelines are built. I assume that's the secret sauce, right? But the flip side of that argument is like, maybe it's very scary that we know, like now many more people understand the mistakes that these these types of systems can make because we're all getting hands-on with GPT. And this system is equally as problematic and we're just oblivious to it because it's a black box.

• 1:48:23Speaker 00

  Almost at your drop off.

• 1:48:24Speaker 06

  Oh.

• 1:48:24Speaker 00

  Check the app for walking directions. Okay, Waymo.

• 1:48:27Speaker 06

  All right.

• 1:48:28Speaker 00

  Well, I think, yeah, that's probably all right. But thanks so much for giving a quick review and.

• 1:48:34Speaker 06

  Thanks for having me.

• 1:48:35Speaker 00

  Yeah, yeah. So that was Louis, whose opinion, I think, is very reflective of the people who are building code generation or code search type startups based on top of GPT four. And as we headed into the Dev Day venue, we actually caught Shreya Rajpawal from Guardrails AI. And there was an interesting comparison here in our conversation between how she views the LLM stack versus how OpenAI views the LLM stack. OpenAI actually had a closed-door session where they gave some thoughts on how they felt that people should start from prompting and build up into a full software system. And they actually deferred a little bit from Shreya. Don't worry, all that is recorded, the videos will come out in a week, but you can listen to Shreya's take. So so we're we're reviewing AI Engineer Summit.

• 1:49:19Speaker 00

  Yeah, we're reviewing the AI Engineer Summit and it was a very, very well organized conference. And a small thing that I was thinking about is that your swag. Yeah. For speakers. Okay, it's on, yeah. Your your speaker swag was like, not surprisingly, I guess, but like really weirdly very nice. And it just kind of like showcases this attention to detail that I think like really kind of permeated the entire, you know, conference. Okay. Like every single decision was very well thought through and, you know, kind of like to a degree of. Like quality that's very rare to see. So yeah, it was it was amazing. I I thought you guys did like an absolutely fantastic job. Yeah. This one mostly goes to Ben, so I'm I'm definitely gonna make sure that Ben understands that I really appreciate the work that he does.

• 1:49:56Speaker 00

  And this is why I couldn't do it myself. You know, I'm mostly the content guy, but I don't, he's the logistics and he's he's run conferences for eight years. So that's why I keep working with him. Yeah, I also kind of really enjoyed the eighteen minutes, you know. Really? Yeah. Yeah, when I saw that, I was like, huh, is this going to be, you know, is this gonna be enough? And like, is that, but it was like. It'd be great. Yeah, yeah, yeah, yeah. I I think the eighteen minutes was actually the right kind of bite size. It's optimized for YouTube. Yeah, I see. Interesting. Okay. Because it's not the in-person audience that matters. I see, I see, I see, I see. Interesting. Okay. I need to promote my my video more. Yeah, is is yours up yet? I don't think it's up yet. It's not up yet? Yeah, we're releasing, we're dripping them out to spread it out. Yeah. Sounds good. So yours maybe in two weeks from now. Okay, okay, sounds good.

• 1:50:41Speaker 00

  Okay, so welcome back. Thank you for having me. No, uh, I think you were guest number five. So you were super early. So so we're we're at the after-party now. How do you feel about the whole day? I'm I'm really excited. I think it was, yeah, I think the I think the excitement in the air with like everybody just like waiting with bated breath to see, I guess, like what gets destroyed, but also like what gets really optimized. I think this is like very, it feels like you're really part of a movement. And Shannon, who were like, you know, us like early people in this space, we gotta stick together because like whatever happens to any of our companies, you know, there's such a like, there's such a transformative moment in technology that. Yeah, so you don't care, right? Yeah, we're all gonna like look back on this time, but I I had a I had a blast. Like, I really, really enjoyed the the releases. Yeah.

• 1:51:27Speaker 00

  What got destroyed? What got destroyed? I'm I'm I'm mining for hot takes here. I, once again, I think my takes are. Very measured. Unfortunately, very measured. This time, I wish I had spicier takes. Your takes are within the guardrails of common. Acceptable. Common behavior, yes. I was, I think retrieval is like the big one for me. I think it's kind of really exciting to see the retrieval baked in, and that's one thing. Where I'm very interested to see like, does that pattern become common by model providers? A by commercial model providers and also by open source model providers? And then how much of retrieval do you have to do yourself? You know, and like, what remains challenging about retrieval? Compared to just like, you know, this, this really easy API to just like have it done for you, right?

• 1:52:10Speaker 00

  Yeah, I, I think what they did was effectively build the basic patterns in. Yeah. But for the more advanced stuff, you're still gonna need Lang-Chain, Llama Index, all those. Yeah, yeah, yeah. So for the longest, I believe that in RAG, it's the retrieval that's the hard part, right? Yeah. And then generation is really easy. As long as you're better, like good retrieval, you can like get really, really far. And the generation only gets you like a little bit over. And so I'm really curious to see like, okay, how, once again, like, how complex do you need it to be in order to start seeing good results? Yeah. Okay, interesting. And what, what are your normal benchmark tests like? Do you, do you actually have a set of tests that you run whenever you're like exploring something? Or some personal favorites of like use cases that you think are tricky for LLMs to do well?

• 1:52:54Speaker 00

  I think like a big focus of ours is on hallucinations, always kind of like checking out hallucination and like conflicting instructions, et cetera, is one. Turst responses is another, you know, like how well is it at like not, you know, you ask it a question and here's this 10-point list and you're very, very verbose. Do you have a terse response validator? Yeah, well, not, we don't have it like, we don't have it publicly, but like we do kind of like check it. Yeah, yeah. So I think like those are kind of some of the things. There was one, there was one example in the, one of the closed door sessions where they, they, all the answers were twitters. Yeah, yeah. Where I think everyone laughed when they were like, can you write a blog post about this? And the guy, and the GPT said, sure, I'll do it tomorrow. Yeah, yeah, yeah, yeah. Yeah, I think like those are, I think those are, I'm really, really excited about. Yeah, just check it. I'm really, really excited about JSON generation.

• 1:53:38Speaker 00

  I've actually got a surprise to see how long it took them to get like, it's, they're probably just doing constrained decoding under the hood, right? Like constrained generation. Okay. Because they're now seeing that guaranteed correct JSON rather than, you know, more correct. Do you get what I'm saying? I was, I was parsing through their words. They've never had an issue producing JSON. It's just that sometimes it doesn't fit the JSON schema, right? Am I, am I wrong? You would know better than more than me. No, I think there are also. Issues with like producing, I think the obvious thing is like Unbalanced brackets? When it's on text length, I think that's like an obvious thing, right? But like weird things when you have like really long strings, then quotes, et cetera, become kind of weird. Uh, so I think those are some other ones. Schema is obviously kind of challenging, et cetera. Yeah.

• 1:54:21Speaker 00

  I think there are, even with function calling, like function calling, at least I I haven't played around with it yet today. But previous generations of function calling wouldn't guarantee that your schema is matched. Correct. Which would be an issue. And I think they're still not guaranteeing it because I kept waiting for them to say it. I haven't read any of the public docs or anything. Do you know if they're guaranteeing that it fits the schema or they're like? Oh, that's a good question. I, yeah, that's a, that's a good point. They never say they guarantee. They never said they got, they, they guaranteed correct JSON. They didn't guarantee if the JSON matches the schema. So, okay, you can call JSON loads, like, big group, like. But I'm very curious to see, like, once again, if this is a pattern that, you know, all of the other foundation model providers adopt. And I don't see why not, right? Like, I, I think for them to kind of like own

• 1:55:04Speaker 00

  specific decoding models is going to like make a lot of sense compared to, you know, like, yeah, a lot of the, a lot of the hacky stuff. Yeah, cool. Any other favorites? You know, not, doesn't have to be Gargyle's related. Any favorite conversations? Favorite demos, favorite? I, oh, the GPTs and the assistants, I think. You want to make one for yourself? Yeah. Uh, I do want to make one for myself. It doesn't, uh, like, yeah, not very Gargyle's related. I do want to kind of play around with like how well it works with like some of the things we track. But yeah, it was just so fascinating to see the marketplace. I am very, very curious to see, you know, what the marketplace looks like, like, is it? Are people going to have like really, really vertically specialized things on the marketplace? Like, if you have a generic, you know, sales assistant or something, right? Like, how much, or SQL generator, how much?

• 1:55:48Speaker 00

  How popular does that become versus like? Sales assistant for X vertical at Y stage of the sales process? Oh, my god. Do you know what I mean? Like, it's, it's so easy to do this now. Yeah. That like where, at what level of specialization do you need to be? You can start seeing the results and that is one thing I'm very excited to see like how that Yeah. pans out. It scares me a little bit because it's basically they said the future programming is natural language or something like that. Yeah. And that's great, but like it really is a new platform, a new operating system almost that they're Yeah, yeah, yeah. They're creating. Yeah. And I don't know how to position myself. Not that I have to because my my world is very developer oriented. Yeah, yeah, yeah. But this is a whole no code world that you and I Yeah. don't touch. Yeah, yeah, yeah, yeah, yeah, yeah. Whoa.

• 1:56:32Speaker 00

  Yeah, yeah. I I really want to see like, is there just going to be like assistance for everything like what? I'm generally curious to see the impact of this on knowledge work, you know, which yeah, like how much of my work like if I'm getting annoyed by something is my first instinct going to be like, you know, let me just, you know, spend the five minutes to build an assistant for this like is is that how everybody's now going to start thinking? You know, and that's one thing I kind of really want to see. Yeah, that's exciting. Yeah. Okay, last question. You spoke at AI Andrew's interest summit. Let's advertise your talk a little bit and point people to your talk. Yeah. Yeah, so thank you again for inviting me to the AI engineer summit, one of my favorite conferences that I've attended, you know, this year. Uh, my talk was about the new paradigms for working with large language models, you know,

• 1:57:15Speaker 00

  for building really production ready applications when the technology that you're working with is under underneath all of it, you know, non deterministic. Really fascinating thing, which was the open A I's talk about building production grade applications talked about how essential it was to build guardrails Yes. as a way to make it to You're talking about the one from today, the Yes, the one from today. Which people haven't seen yet, but Which people haven't seen yet, but really, really cool talk. So I think it really validates what we've been saying pretty much since the beginning of the year, which is that you'll get like a certain, you know, you'll get to a certain point, but at that point you need to start adding guardrails to your application if you need to get your users to start. Yeah. Uh, you know, getting value out of what you what you build out, right? So. Yeah, so I I I have your pick your chart and I have their chart.

• 1:57:59Speaker 00

  They put guardrails at the first layer. It's not at the end. It's it's actually it's right at the beginning for user experience. Yeah, that's right. Yeah. Yeah, that was kind of interesting to see that they put it as part of the U X. I'm still kind of very candidly, I'm still kind of digesting that like I think of it as I think of it as part of the infrastructure and I don't know if as it's as much U X as it is, you know, just like one of the components that you need in your stack. But I I I think the pat like a lot of what they said today completely validated, you know, what we've felt for the longest time. And also what I go really in depth about like in the talk. That I gave, right? Which is that what happens when you have the, once you have the bare bones application ready? What is the process of actually adding guardrails for what you care about? Like, what does that look like? Yeah.

• 1:58:42Speaker 00

  You know, what are the risks that you care about? How do you verify that those risks are happening or not happening? If they are happening, how do you quantify them? And then how do you mitigate them? That was what, that was what the talk was about, which I would really recommend people go and check out. Awesome. Well, you did a great job. We're gonna post the talk soon and thanks. It's good to see you again. Thanks again for inviting me. And that was about all I managed to get before the after party. At the after party, there was actually an after after party thrown by Noos Research. So let's hear a little bit about OpenAI versus Open Source AI from Alex Volkov. Okay, so we are in the one day after Dev Day here with Alex. Hey.

• 1:59:19Speaker 03

  Hey.

• 1:59:20Speaker 00

  Very, very recognizable voice right now. We don't have to introduce you.

• 1:59:22Speaker 03

  Hey, everyone.

• 1:59:23Speaker 00

  And we are here to talk about the two parties that happened yesterday. There was one official Dev Day OpenAI Afterparty where I interviewed Traya, who was just before this. And then there's an unofficial one for keeping AI open by Noos.

• 1:59:35Speaker 03

  Yeah.

• 1:59:36Speaker 00

  So what was it like? Just compare and contrast.

• 1:59:38Speaker 03

  So let me maybe start with like who Noos Research is.

• 1:59:41Speaker 00

  Oh, yeah. Yeah. Most people haven't heard of this.

• 1:59:42Speaker 03

  It's written N O U S. So I, I mispronounced it now multiple times. It's like it's Noos Research. It's one of the few organizations online that started like from a Discord and then like kept going up until like a significant amount of people like working with them, affiliated with them. Of folks who take open source model to its most extreme capability. So collect data, data sets from open source, open source and more closed source. And depending on that, they release like the different licenses. And then they fine tune open source models that were like released to us from like Llama, for example, and Mistral, uh, which is a French company that recently released a 7D model that's the best. And they've been doing this since Llama one, but recently it really kicked into high gear with Llama two releases because Llama two uh, ended up being with a commercial license.

• 2:00:25Speaker 03

  So you could actually use this for actual, you know, products and services. And Mistral came out with like a full Apache two license with a BitTorrent link. I think you remember that. And so these organizations suddenly became like a very. Very important currency in the, in the world of like where the whole world of AI is going. Because they're aligning local models and many companies love OpenAI, but either cannot afford this or cannot risk the chance that OpenAI changes something like what's our Dev Day. And so many people are turning on to like, okay, if we want to run our own hardware, how do we actually do this? And you can run it, you can run Lama two and all these models on your own hardware, but then you want to fine tune them for your own purposes. And so how do you actually find them? And now organizations like News Research was probably the biggest one.

• 2:01:09Speaker 03

  Alignment Labs, shout out to Austin and folks from Alignment Labs. Skunk Works and many of these like people come up and say, hey, we have the know how. And we always started learning about this like eight months ago, six months ago themselves. But now they're like the specialized more people that fine tune models and actually release the best kind of models on the Hugging Face open source leaderboard.

• 2:01:30Speaker 00

  Yeah, yeah. And in, in my knowledge, top two models that I keep hearing about, one is Hermes. Then they recently searched the base model for Hermes from Lama to Mistral. Because apparently it's better.

• 2:01:40Speaker 03

  Yeah.

• 2:01:41Speaker 00

  Hermes is like an instruction data set, nine hundred thousand instructions. I don't really know where it's from. Maybe I don't want to know. They also do some like fun models. There's like a mystical model that they do.

• 2:01:50Speaker 03

  Transmestus, yeah.

• 2:01:52Speaker 00

  Some, some stuff like that. I think it's actually a little bit weird that they keep releasing models like they, they release like three models a week.

• 2:01:59Speaker 03

  It's insane.

• 2:01:59Speaker 00

  Right? And it's very hard to keep up. Like, I'm like, okay, which one is actually the one that I should pay attention to?

• 2:02:04Speaker 03

  Yeah. So first of all, you're welcome to join Thursday AI. And then we talk about all the models every week. It's kind of interesting to that if I do like a recap for a month, the beginning of the month, most of the updates don't matter because like every, every.

• 2:02:16Speaker 00

  This, I'm doing monthly and I, I feel this like.

• 2:02:20Speaker 03

  Every month.

• 2:02:20Speaker 00

  I'm doing this, I'm doing this for historical posterity, like.

• 2:02:23Speaker 03

  Yes.

• 2:02:23Speaker 00

  Five years from now, people want to look back, then they, they can look at my notes because I only have 12 a year.

• 2:02:28Speaker 03

  Yeah, nobody's gonna look at your notes. They're gonna have a GPT trained on your notes answering everything. I have, yeah, I, I, I'm doing like every week and every week we're talking about like this model outperforms that model like significantly. And we're noticing significant changes from week to week. Literally in the span of a month, we went from a thirty three billion parameter model, which is big. And parameter count is not everything there is. You can have a smaller model with like larger, longer training that actually will perform better than whatever. But we're noticing smaller and smaller models doing outperforming bigger ones significantly. Zephyr from Hugging Face outperformed Lama seventy B and Zephyr is like only like a seven B model.

• 2:03:02Speaker 00

  On some things.

• 2:03:03Speaker 03

  On some things, for sure. And so this is very interesting because like it's really hard to evaluate. Evaluation frameworks are bad. Everybody's saying that they're not representing of anything. People can fine tune, over tune on them. And so there's this whole kind of subculture of open source, mostly on Discord, some of them on X and Twitter Spaces. And for some reason, but I find it very like humbling and incredible, they also hang out in Thursday AI. And so that's how I got to this. That's how I got to meet like news research folks. Technium, Emozilla and organized the the counter party event last night together with some other EAC people that we know from Twitter as well.

• 2:03:38Speaker 00

  Including Mark Ajit.

• 2:03:40Speaker 03

  So apparently he was supposed to, I didn't see him.

• 2:03:42Speaker 00

  Ah, okay.

• 2:03:43Speaker 03

  But like.

• 2:03:43Speaker 00

  I saw a photo with a bald head of a big guy. So I was like, is that Mark? I don't, I don't know. Anyway, but the opening I party was at a art museum.

• 2:03:51Speaker 03

  Hmm.

• 2:03:51Speaker 00

  And then the news research party was at a club.

• 2:03:54Speaker 03

  It was a club, yes. At Folsom, Folsom Street in San Francisco, a club, ten fifteen Folsom, I think.

• 2:03:59Speaker 00

  Sure.

• 2:03:59Speaker 03

  Open AI was a very like highbrow, buttoned up.

• 2:04:03Speaker 00

  Yeah.

• 2:04:03Speaker 03

  Event.

• 2:04:04Speaker 00

  There was a, yeah, there was a live band, uh, someone playing jazz.

• 2:04:07Speaker 03

  Yeah.

• 2:04:07Speaker 00

  Yeah.

• 2:04:08Speaker 03

  Which I, I think I mentioned this once, it was too loud.

• 2:04:10Speaker 00

  Yeah.

• 2:04:11Speaker 03

  We wanna talk, we don't wanna listen to music.

• 2:04:12Speaker 00

  No, no, no, we're just old.

• 2:04:13Speaker 03

  Yeah.

• 2:04:14Speaker 00

  Just, everything, everything is too loud.

• 2:04:16Speaker 03

  And then, uh, it was like a lot of people, a lot of networking, a lot of people trying to get together, maybe do business together, like very, very awesome. Many people from OpenAI actually showed up, a lot of people. We, we stood in line, there was a long line for the Magnum Meters to, to step in. And then everybody like passing us around was like OpenAI employee that passing like straight through.

• 2:04:33Speaker 00

  Yeah.

• 2:04:34Speaker 03

  And then that ended around eight, which was like the standard San Francisco, like buttoned up.

• 2:04:38Speaker 00

  Oh yeah, that's when you go to bed.

• 2:04:39Speaker 03

  That's where you go to bed. And that's when the other party kind of started.

• 2:04:42Speaker 00

  Yeah, yeah.

• 2:04:42Speaker 03

  And I think they just seized the opportunity because everybody's in town for the OpenAI stuff.

• 2:04:47Speaker 00

  Yeah.

• 2:04:47Speaker 03

  Why not make a splash, an announcement for like for open sourcing AI? So literally the, the invite was keep AI free dot com.

• 2:04:55Speaker 00

  Yeah.

• 2:04:55Speaker 03

  Uh, which was the, the website and the.

• 2:04:57Speaker 00

  Keep AI open, yeah.

• 2:04:58Speaker 03

  Keep AI open dot com. And you had to register, you had to go in there. And this was, to me, an incredible kind of show of Twitter in real life. So all of the folks who follow Mark Andreessen, he recently stepped into this thing with like the techno optimism stuff. He started to boost the E effective acceleration, EAC, folks. And so there's a lot of like signature stuff from that like ecosystem on Twitter. There's like, don't tread on me with like, you don't take away my GPUs. There's like all these signs across the club. The, it's a very visual club as well, so where the DJs, it's a whole like a 3D projected thing. So there's like a bunch of like art and like live things about keep AI open. I, I found it like very, very super cool. I, I, I'll, I have to tell you a tidbit.

• 2:05:43Speaker 03

  I saw me and Killian were there from Open Interpreter. We saw two people with lab coats. It was like, what's the deal with lab coats? So we went and asked and they just said, hey, we just like, we came back from our work where we work on semiconductors. We're actually like touching chips, whatever, we just like didn't change out of it. And my head was like so incredible in the keep AI open GPU kind of poor party. We have people who literally work on superconductors came from the work, like they're working on chips.

• 2:06:06Speaker 00

  Yeah, yeah. Semiconductors or superconductors, very different things.

• 2:06:09Speaker 03

  I think semiconductors. Yes.

• 2:06:11Speaker 00

  Yeah, yeah, we, we had the superconductor episode a while, a while back. I think if people are still recovering.

• 2:06:16Speaker 03

  I'm, I'm personally still recovering from that, there was a whole thing from here.

• 2:06:19Speaker 00

  So is news research like vibes, you know, like what, what is the mission apart from to keep publishing open source models?

• 2:06:27Speaker 03

  I think you'll have to get some news people to actually speak like about the mission, about the actual product. But as, as far as I understand this, no matter how much the product side will be, and there will, might be, there's so many people that are doing like so incredible stuff that people notice like, and you know. So no, no matter how like, how much of the, the, the business side will be, they're like committed to fully open source as much as possible, including data sets, including models that are like trig, trigmas, for example, their model that's like trained on the occult and the physical, metaphysical, all, all. You, you can't expect OpenAI to let you talk with a model, they'll answer with like

• 2:07:03Speaker 00

  Mystical questions.

• 2:07:04Speaker 03

  Mystical stuff.

• 2:07:04Speaker 00

  Astrology, Halloween.

• 2:07:06Speaker 03

  So you you're very like easy into this astrology and Halloween, they're talking about like you can ask this model about like resurrection.

• 2:07:12Speaker 00

  I see.

• 2:07:13Speaker 03

  Right? Like all of the occult like craziness that they've collected, OpenAI will not let you do that. And so there's other things that I will not let you do by default because they have lawyers and they don't get sued.

• 2:07:22Speaker 00

  Yeah.

• 2:07:22Speaker 03

  Recently, they announced the Protection Shield thing, so you won't get sued because of their model. So they're damn anthropic, all these big companies, it's very important for them to protect the outputs and the models. Here, these folks are like, hey, if you wanna build a model, fine tune this, we're gonna teach you how, jump on our Discord. We're gonna help you with producing like the biggest models. And then if, you know, there's gonna be like a financial aspect to this as well, if your company wants to run this, we'll also help you do that.

• 2:07:48Speaker 00

  Yeah, so it's the same as stability, basically, is is that's from what it from talking to him, that's what I gather.

• 2:07:54Speaker 03

  Yeah.

• 2:07:54Speaker 00

  Cool. Anything else that people should know about the party news?

• 2:07:58Speaker 03

  I found this whole day to be like a very singular AI day, and we don't get many of this. G P T four, I think, was the biggest one previously.

• 2:08:06Speaker 00

  Yeah, March.

• 2:08:06Speaker 03

  Like a single, March fourteenth, that's what Thursday I started.

• 2:08:09Speaker 00

  Yeah.

• 2:08:09Speaker 03

  We started talking about this every week. This was a singular day in San Francisco, this like started pregame party with swyx and some other folks that I I got to feel like a little bit of San Francisco. And then Dev Day was incredible, we just heard from Simon, there was like a garage that they made into a venue event.

• 2:08:25Speaker 00

  Yes.

• 2:08:25Speaker 03

  Probably custom venue event on the fly, which like just talks about how much they can pull off. It felt to me that like this Dev Day event and then the following party, it felt a little bit like, almost like an Apple thing.

• 2:08:37Speaker 00

  Yeah.

• 2:08:37Speaker 03

  Where like it's going to be a yearly thing that people will like try to get in as much as possible. One thing to note that in the other party, there were many people who didn't get in to this party. And so, you know, they were watching for like a a party.

• 2:08:49Speaker 00

  This this office right here.

• 2:08:50Speaker 03

  This office, people watched here, and people watched in in the live space that we we.

• 2:08:54Speaker 00

  Eight thousand people tuned into our spaces.

• 2:08:56Speaker 03

  Eight thousand people tuned in? I didn't even have a chance to look at it.

• 2:08:59Speaker 00

  I always want to know the number.

• 2:09:00Speaker 03

  Oh.

• 2:09:00Speaker 00

  So it it shows the relative level of interest and, you know, like, so according to twenty two thousand.

• 2:09:05Speaker 03

  Hmm.

• 2:09:06Speaker 00

  And this is eight thousand.

• 2:09:06Speaker 03

  Yeah.

• 2:09:07Speaker 00

  Just relative interest.

• 2:09:09Speaker 03

  Yeah, there was like two spaces as well, Robert Scobarada, he he stole the thunder a little bit.

• 2:09:13Speaker 00

  Stole some audience from us, yeah.

• 2:09:15Speaker 03

  Shout out, Robert. And I think that like it was a singular day, and I think the new research keep open source open, E A C, Mark Andreessen, like all these things together. Also added to the top of this because like it happened in the same day, one on top of another, in the same place, San Francisco.

• 2:09:28Speaker 00

  Yeah.

• 2:09:28Speaker 03

  I I find it incredible, I would, you know, definitely come back next year to.

• 2:09:32Speaker 00

  Yeah, okay. Yeah, well, I think you'll be back sooner than that.

• 2:09:35Speaker 03

  Yeah, probably.

• 2:09:36Speaker 00

  There'll be other things going on. All right, thanks.

• 2:09:38Speaker 03

  Awesome.

• 2:09:39Speaker 00

  All right, last but not least, we go back all the way to the Newton, where I started this podcast, where we checked in with Rahul Sonwaka, better known as Rahul Ligma, who just celebrated his one-year anniversary as one of the biggest memes and celebrities in San Francisco. But by day, he's also the CEO and co-founder of Julius AI. And I'll match it up. What's up, Socks? Hey, good to see you. It is one day after Dev Day, and we all had a chance to process. How do you feel? What's your top takes?

• 2:10:09Speaker 02

  Dev Day was awesome. We got to see a bunch of really smart people who are building cool things with OpenAI, GPT, Dolly. The event was very well put together. The keynote was awesome. The energy in the room was crazy. And I could see real-time social media firing up with all these takes. Overall, I think it was a good day.

• 2:10:28Speaker 00

  Yeah, I interviewed Surya Dantuluri.

• 2:10:30Speaker 02

  Yeah.

• 2:10:30Speaker 00

  I think you know him. He was like.

• 2:10:32Speaker 02

  Samma just killed my startup.

• 2:10:35Speaker 00

  And it was almost true for him, because he has a bunch of plugins.

• 2:10:39Speaker 02

  Yeah.

• 2:10:39Speaker 00

  And plugins are kind of deprecated.

• 2:10:41Speaker 02

  Yeah.

• 2:10:41Speaker 00

  Yeah.

• 2:10:42Speaker 02

  Yeah. Yeah, the plugin thing was interesting because it was, it's going to be deprecated, but they just accidentally turned it off yesterday.

• 2:10:51Speaker 00

  Yeah.

• 2:10:51Speaker 02

  And then they freaked out a bit. It freaked out, and then they brought it back up. It's, yeah.

• 2:10:56Speaker 00

  Yeah. So, top features that you're interested in, that you want to explore more?

• 2:11:00Speaker 02

  I think people are super psyched about the Assistance API, but personally, if you ask me, two things that I am most excited about is turbo.

• 2:11:08Speaker 00

  Yeah.

• 2:11:09Speaker 02

  The speed is crazy and.

• 2:11:11Speaker 00

  Have you actually, have you met, you know, do you know any like rough measured? Because I don't think they actually ever mentioned the speed relative difference.

• 2:11:18Speaker 02

  I started noticing the speed difference in ChatGPT, actually, like a few weeks ago.

• 2:11:22Speaker 00

  Oh, I see. Also, they already slowly eased this into it.

• 2:11:25Speaker 02

  Yeah, yeah. And I saw like takes on Twitter that, did anyone notice ChatGPT get much faster? And I noticed it too.

• 2:11:32Speaker 00

  Yeah.

• 2:11:32Speaker 02

  But, so it's turbo was exciting, but the second thing that's exciting is multiple function calling.

• 2:11:38Speaker 00

  Yeah.

• 2:11:38Speaker 02

  And then the JSON output formatting. I think as developers are building on the Dev API, so that's the thing that's super exciting to me. You know, of course, there's version stuff, there's. Code interpreter as a tool in the API.

• 2:11:52Speaker 00

  Yeah.

• 2:11:52Speaker 02

  But I think what will bring the most applications is actually the the speed. Because there are so many things, if you if you look at our numbers on Julius, people are not patient. They want an answer and I want to answer quick. And we see clearly, if if you can get an answer to them a few seconds faster, there's a clear difference in the conversion.

• 2:12:17Speaker 00

  Hmm.

• 2:12:17Speaker 02

  So speed is going to be.

• 2:12:18Speaker 00

  What is conversion for you? Is that just paying or?

• 2:12:21Speaker 02

  Oh, no, it's like from first message to second message.

• 2:12:24Speaker 00

  I see.

• 2:12:24Speaker 02

  So we we do code gen and then we run the code and then the code has an output. The user asks a second message and we can just see the funnel.

• 2:12:32Speaker 00

  Yeah.

• 2:12:32Speaker 02

  Where if it's faster, the code runs faster. And the second thing is multiple function calling. I think you're basically telling the AI that so so I think the people misunderstand function calling. It's essentially to use. And if you can tell the AI, hey, you can give me multiple tools to tools to use.

• 2:12:49Speaker 00

  Yeah.

• 2:12:50Speaker 02

  At once.

• 2:12:51Speaker 00

  Yeah.

• 2:12:51Speaker 02

  I think that's going to unlock different applications than before, because before it was just like, okay, this is a task. Tell me one tool and what's the input for it.

• 2:13:00Speaker 00

  Yeah.

• 2:13:00Speaker 02

  Um, but if the AI can now use multiple tools in parallel, you can first of all have more specialized tools. And then the AI more specialized instructions for each tool.

• 2:13:10Speaker 00

  Yeah.

• 2:13:10Speaker 02

  It's just going to unlock a lot of cool applications that previously weren't possible.

• 2:13:14Speaker 00

  There was a practical limit in the number of tools that you can give it, right? So we had this discussion in March, February, March, April, when they released the function API. That this is subject to context window, the JSON schema itself.

• 2:13:26Speaker 02

  Yeah.

• 2:13:26Speaker 00

  Does that change at all? Or I don't know if you.

• 2:13:28Speaker 02

  I um, I don't.

• 2:13:31Speaker 00

  Yeah.

• 2:13:31Speaker 02

  But I what I noticed though before even before was that more functions and more options just confused it. And that's what I want to play with next is like, okay, what's the breaking point?

• 2:13:41Speaker 00

  I see.

• 2:13:41Speaker 02

  Like, does more options, you know, confuse it? Does it make it?

• 2:13:46Speaker 00

  Would you? Would you use multiple function calls as well or?

• 2:13:49Speaker 02

  Oh, totally.

• 2:13:49Speaker 00

  Such as theoretical.

• 2:13:50Speaker 02

  No, no, no, I have a direct application for it right now. One of them is oftentimes G P T writes code. And then we run that code and we realize that, oh, from G P T's last knowledge update, that module in Python has changed. It has new functions, new APIs. So today, the way we do it is when the error happens, we tell G P T, okay, you can go look up new documentation and then fix that error. But with multiple function calling, the way we would do it is like, give me the code, but then also give me a documentation lookup. And then when the error happens, I can just quickly fix that without another G P T call.

• 2:14:25Speaker 00

  Yeah.

• 2:14:26Speaker 02

  And then keep moving.

• 2:14:28Speaker 00

  Nice, nice.

• 2:14:29Speaker 02

  But, but I mean, in general, it's just like multiple to use to me is just so exciting as a developer. And I wish people were talking more about this.

• 2:14:36Speaker 00

  Yeah, I mean, people are still coming to terms with just like the base model and prompt engineering and all that. That's still important, but for engineers, I think you should explore these other advanced features.

• 2:14:46Speaker 02

  True.

• 2:14:46Speaker 00

  Yeah.

• 2:14:46Speaker 02

  Yeah.

• 2:14:47Speaker 00

  Anything on the multimodality side that you're interested in?

• 2:14:49Speaker 02

  I mean, vision will be super interesting for sure. And we have this functionality in Julius right now where you can generate react and H T M O.

• 2:14:59Speaker 00

  Like V zero. I think Matt was showing me.

• 2:15:02Speaker 02

  Yeah.

• 2:15:03Speaker 00

  A little bit of that demo.

• 2:15:04Speaker 02

  Yeah, yeah, we have been hacking on it a lot. But I think the, the missing piece here is that, well, you have an engineer who knows how to react and they probably won't find this useful. But if I can allow like anyone in the, in the world to just draw a mock up on a piece of paper. And then run that and have the vision.

• 2:15:22Speaker 00

  Yeah.

• 2:15:22Speaker 02

  Gonna demo, yeah. Yeah. Turn into like actual components I could use in a, on a webpage, that'd be sick. And what's, what's even more sick is like, have the feedback loop where you take a screenshot of the page, generate it, and then feed that screenshot back into vision and then come up with more instruction and have that loop.

• 2:15:40Speaker 00

  Yeah.

• 2:15:41Speaker 02

  Wow, like a self-improving webpage. Isn't that crazy?

• 2:15:44Speaker 00

  Yeah.

• 2:15:44Speaker 02

  I'm, I'm super excited.

• 2:15:45Speaker 00

  Yeah, yeah. So in my mind, Julius is very data focused. I, I, by the way, I didn't introduce you. I didn't introduce you. I was just gonna do it separately.

• 2:15:53Speaker 02

  Yeah.

• 2:15:53Speaker 00

  But people know who you are.

• 2:15:55Speaker 02

  Yeah.

• 2:15:55Speaker 00

  You're, you're, you have a Wikipedia page.

• 2:15:57Speaker 02

  Yeah.

• 2:15:57Speaker 00

  You just passed your one year anniversary as Rahul Lingma.

• 2:15:59Speaker 02

  Thank you.

• 2:16:00Speaker 00

  By the way, any, any fun things happen on the anniversary or? I saw Ilya, Ilya said, Ilya recognized you on the squad. Oh, Ilya was like, oh my god, this is, oh, you're famous or whatever. And no, these guys are so awesome, like, they're so humble. But it happened the fir- one year anniversary, nothing really, like, it's, I mean, you, you knew about it a week before. I like to set anniversary dates. That's awesome. Because it reminds people of the passage of time. Like, it's like, wow, shit, has that been a year? Yeah. And then you're like, I think it motivates, it motivates me more than like Memento Mori, like, yeah, this case, you know, some, sometimes you're, you're out of date, but it reminds me to spend my years wisely, to, to do interesting things with the time that I have. Memento Mori is kind of depressing, whereas this is. This is like, oh yeah, did you know that one year ago? Wow, it's been a year.

• 2:16:44Speaker 00

  Yeah, okay, cool, but Julius, you data analysis chat thing. Yeah. Basically code interpreter plus plus is how I think about it. Exactly. And also you trust across a hundred thousand users? Yep. You have delivery modes across your, your plugin as well as a chat box, like a dedicated web app? Yep. Okay, anything else that people should know? Well, the, our vision is, you know, writing code is super fundamental to doing things. Yeah. You could not only automate a bunch of tasks in your life, but just writing code, but also it's how you, how you just like interact with the universe, right? You could, you have code that brings you a Waymo car and picks you up and just drops you somewhere.

• 2:17:25Speaker 00

  And I think allowing these language models to write code and do things for you is really powerful. And data announces this application that we're most excited about right now because that's what it's good at immediately. But just on Friday, we launched FFmpeg support and there were people trying to upload videos, turn the videos into GIFs, or like take a YouTube video, turn it into a, you know, a short summary and all these different cool use cases that we didn't truly like hardcore into Julius, we just told it, hey, here, now you can run FFmpeg and you can run YDLP and, and Movie Pie and all these different things, do these tasks for me. And then people were just like organically describing those things. There's this guy TDM on, on, um, on Twitter, CTO Junior,

• 2:18:06Speaker 00

  and he took some meme video and put it on my own tweet, overlaid on my own tweet, and then I tweeted that. Then that got a bunch of likes and I was like, dude, like this is the first one that gets a lot of likes on, on, you know, F M back on Julius. So. That's a lot of meme potential. Yeah, a lot of meme potential, but that's not what we're going for. Yeah. You know, it's just like letting people like do things. Your, your target market is like the S and D, the enterprise. It's actually individuals who have data on their head and they just wanna jump. Academics. A lot of academics, actually. Yeah, a lot of academics, a lot of students, researchers, any kind of C S V X L data you can just dump into Julius. Yeah.

• 2:18:44Speaker 00

  And have it analyze for you. We have this video coming out in a few days where you can now actually train Nano G P T on Julius. So you can give it, hey, here's the good approval for Carpe T. You, yeah, it has, you have, you have GPUs to train it on or you just train it in CPU? CPU. Yeah, that's true, that's true. Yeah. Yeah, I mean, Carpathia like that. Yeah, yeah, yeah. Okay, cool. So the, the thing I really wanna sort of ask you as a founder on is, you know, I think there's always this existential threat about OpenAI building your features, right? In a way, so like the, the number two default bot in the, in the G P T app store. Yeah. Is data analysis. Yeah. And people can build their own by customizing and adding code interpreter. Yeah.

• 2:19:25Speaker 00

  Although I think there's also opportunities for you. So on the roadmap that they presented in the closed session, they also said you can bring your own code interpreter. Yeah. So like, how are you thinking about that? I mean. How, how, as a founder or as? Founder. As, so who's the audience? Is it like other founders or is it? Yeah, other, other founders and then is it people just interested in how you are, you're going, you're processing this? Yeah. I mean, this is, I think it's a very interesting story of processing this live because the news just dropped yesterday. Yeah, totally. Well, so the story behind Julius is that we actually launched Julius. Three months after code interpreter was announced and a few weeks after it was rolled out to everyone else in the world. Yeah. So we, we, we were number two and even then we got a hundred thousand users.

• 2:20:09Speaker 00

  Because I think there's a lot of work to do to get something to work properly. And there's a bunch of examples of this on the internet. So if I'm talking to founders, what I'll tell them is, man, so many people give up before even getting started. And that happens, don't do that. Sure, you can change your idea. You can find new things to work on. But the way I'm processing is that we were actually, we launched after Code Interpreter came out. And there's a hundred thousand people who think Julius is better than Code Interpreter. Or just try it out. Yeah, I'll try it out and use it over Code Interpreter. And there's like a lot of work to do, like for example, the FFM back stuff we launched on Friday. Or the HTML stuff, or we, you know, react component and stuff.

• 2:20:54Speaker 00

  All these different things to get them to work, it takes some effort. How I'm processing it, I mean, you know, that's like, that's what startups are all about. It's like risk, right? If you, if you want to build a risk-free startup, you probably don't want to work on startups. Yeah, just go get a job. Just go get a job, exactly. So I'm having so much fun. The way I'm thinking about this is like, whoa, there's all these new different things I could do now. I could build, that's so exciting to me. And I'm, I'm pumped. Yeah, yeah. Awesome, that's it? Any last words, call to action? Call to action, let's go build some cool things and get a bunch of users. Let's do it, guys. Yeah, all right, awesome, thanks so much. Thanks, Wix. I think that's a meme that we can all get behind. Let's go build things for a bunch of users with AI.