E50: The Future of Software Development with Tyler Angert of Replit

In this episode, Nathan sits down with Tyler Angert, Product Designer at Replit, to discuss the future of software development. Replit is building what CEO Amjad Masad calls "the perfect substrate for AGI." In this discussion, Tyler and Nathan discuss how Replit is leveraging AI to enhance its current product, bot-bot interactions, the design and ethical considerations around AI agents, and more. If you're looking for an ERP platform, check out our sponsor, NetSuite: http://netsuite.com/cognitive We're hiring across the board at Turpentine and for Erik's personal team on other projects he's incubating. He's hiring a Chief of Staff, EA, Head of Special Projects, Investment Associate, and more. For a list of JDs, check out: eriktorenberg.com. TIMESTAMPS: (00:00) Episode Preview (01:10) Nathan’s Intro (09:24) Tyler's role at Replit (11:37) Replit as the "perfect substrate for AGI" (14:55) Sponsor: Netsuite | Omneky (15:45) Defining AGI (19:26) How AI agents might interact with Replit (23:20) Replit's "virtual developer" product (31:11) Current state of Replit's AI features like Ghostwriter (41:48) Measuring productivity boost from AI coding tools (45:29) Potential for massive developer productivity gains from AI (48:34) Technical gaps still remaining to achieve advanced AI agents (51:45) Core AI breakthroughs have already occurred (57:35) Timeline for functional AI agents interacting online (01:01:20) Safety considerations for powerful AI agents on Replit (01:08:47) Ethical considerations around AI and consciousness (01:20:36) What constitutes consciousness in an AI system (01:26:30) Should AI systems have rights and protections? (01:30:32) Being polite to AI systems (01:31:59) Advice for beginners looking to leverage AI (01:36:58) Using AI with a Neuralink brain implant (01:43:02) Hopes and fears about AI's impact on society LINKS: https://replit.com/ X: @tylerangert (Tyler) @labenz (Nathan) @eriktorenberg (Erik) @cogrev_podcast SPONSORS: Shopify | NetSuite | Omneky Shopify is the global commerce platform that helps you sell at every stage of your business. Shopify powers 10% of ALL eCommerce in the US. And Shopify's the global force behind Allbirds, Rothy's, and Brooklinen, and 1,000,000s of other entrepreneurs across 175 countries. From their all-in-one e-commerce platform, to their in-person POS system – wherever and whatever you're selling, Shopify's got you covered. With free Shopify Magic, sell more with less effort by whipping up captivating content that converts – from blog posts to product descriptions using AI. Sign up for $1/month trial period: https://shopify.com/cognitive -NetSuite provides financial software for all your business needs. More than thirty-six thousand companies have already upgraded to NetSuite, gaining visibility and control over their financials, inventory, HR, eCommerce, and more. If you’re looking for an ERP platform: NetSuite (http://netsuite.com/cognitive) and defer payments of a FULL NetSuite implementation for six months. -Omneky is an omnichannel creative generation platform that lets you launch hundreds of thousands of ad iterations that actually work customized across all platforms, with a click of a button. Omneky combines generative AI and real-time advertising data. Mention "Cog Rev" for 10% off. MUSIC CREDIT: MusicLM This show is produced by Turpentine: a network of podcasts, newsletters, and more, covering technology, business, and culture — all from the perspective of industry insiders and experts. We’re launching new shows every week, and we’re looking for industry-leading sponsors — if you think that might be you and your company, email us at [email protected].

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

• 0:00Nathan Labenz?

  It's so simple. I just want to do X. I know it is possible to do it. I know probably thousands of people have done it before, and yet I can't quite track down exactly the way to do it. So it was like a whoa moment for me when I literally just typed a comment, and then the next thing you know, it writes the whole thing for me. It wasn't even that much code, but the key point was that it was correct, and that it happened in a second. And then I was like, oh, my God, I can just do that? And it works?

• 0:28Speaker 01

  If we want to get to the point of like 100x productivity, 1,000x productivity, where, you know, one, two, three people, right, three-person startups are essentially running the equivalent business of what are now today like 500,000-person companies, we need to parallelize. And this is where the bots and the agents come in.

• 0:47Nathan Labenz?

  Hello and welcome to The Cognitive Revolution, where we interview visionary researchers, entrepreneurs, and builders working on the frontier of artificial intelligence. Each week, we'll explore their revolutionary ideas, and together, we'll build a picture of how AI technology will transform work, life, and society in the coming years. I'm Nathan Labenz, joined by my co-host, Erik Torenberg. Welcome back to The Cognitive Revolution, and welcome to Replit Week. Today, we're sharing my conversation with Tyler Anger, Replit's first product designer, which we recorded in May. And next week, we'll have a follow-up with Replit's new VP of AI, Michaela Catasta. Why do two episodes on a single company? Simply put, because I believe that

• 1:32Nathan Labenz?

  Replit might prove to be one of the most important companies in the world. For those not familiar with Replit, it is the single easiest way to create and deploy fully custom software. Working through a native web environment that feels as responsive and customizable as your typical desktop coding environment, users can copy and launch cloud servers with a click, then immediately begin to write and execute code.

• 2:00Nathan Labenz?

  Replit has become famous for both product excellence and release velocity. In addition to abstracting away so much of the hassle and headache from the software development experience, Replit has pioneered features including multiplayer mode, replayable edit history, a bounty marketplace for hiring small projects done, and even has the best mobile coding app that I've personally used. Replit started as a tool for "hackers," and they've built a global user base of some 25 million young software developers who they support with a generous free tier. They've accomplished all this with under 100 team members. But with its recent partnership with Google, its new deployments product meant to support

• 2:43Nathan Labenz?

  mission-critical software, and especially now its market-leading productization of AI coding assistance, I genuinely believe that Replit is poised to compete with Microsoft and Amazon as the unified application development and hosting platform of choice for a wide range of software projects going forward.

• 3:04Nathan Labenz?

  Compared to the typical workflows developers use to share and publish their work today, Replit is a genuine breath of fresh air.

• 3:12Nathan Labenz?

  To give you a sense for just how transformative this can be, since recording this episode in May, I've started using Replit in a new AI task automation training program that I've been developing for executive assistants at Athena, and I'm now beginning to teach EAs to create custom software without teaching them to code. How is that possible? Well, for starters, we've learned that most businesses have similar opportunities to use AI to save time and money, or to scale previously unscalable tasks. To take just one example, nearly every business could benefit from more meaningfully personalized outreach, whether that's to sales leads or job candidates. But of course, it runs much deeper than that.

• 3:57Nathan Labenz?

  For routine tasks, where consistent execution matters more than creative genius, with a few rounds of iteration, most text-based tasks can now be at least semi-automated. Prompting is generally no longer a major barrier. Models are getting easier to work with by the month, and already just a handful of core prompting skills—input labeling, role casting, few-shot examples, chain of thought, format trick—just these five cover the vast majority of use cases. The challenge then becomes how to work AI into existing business processes without having to spend all day copying and pasting into and out of ChatGPT. And it is tricky because every situation is unique.

• 4:41Nathan Labenz?

  Requirements are idiosyncratic, data is in different places and formats, and every integration point is also a possible point of failure. Authentication and access of all forms is a huge source of friction.

• 4:56Nathan Labenz?

  To help the EAs overcome all this, I've developed the mantra of copy and customize. Our plan is to equip the EAs with simple templates for common situations. Athena calls these playbooks. And then we teach them how to use AI coding assistance to modify the examples to suit clients' particular needs. This way, we're empowering them not just to serve as EAs, but as AI implementation specialists at their clients' companies. In practice, this amounts to using GPT-4 to write code, which gets, processes, and then loops through some inputs, often making another AI API call for each input.

• 5:38Nathan Labenz?

  And increasingly, that's Claude 2 these days, thanks to the 100K token context window, before finally returning all the results in some desired format. There is a learning curve here, to be sure. But over time, I believe this skill set will prove transformative, even as it becomes comicless.

• 5:58Nathan Labenz?

  Now, obviously, this would not be possible without advanced language models like GPT-4. But it also wouldn't be possible without Replit. Because it's so easy to share executable code with no environment confusion, Replit makes it remarkably easy to help one another get unstuck. And because it's a full development and now hosting platform, we can always layer UIs or APIs onto our AI-powered creations as needed.

• 6:25Nathan Labenz?

  AI task automation and this software without coding paradigm will surely be the subject of future episodes. But for now, if you're interested in hiring a human executive assistant that's coming out of this AI training, there is a referral link in the show notes where Athena offers a special deal for qualified customers. It really is a lot of fun teaching these EAs, who never even signed up to learn to code in the first place, how to use these powerful tools.

• 6:54Nathan Labenz?

  But as you'll hear over the course of these next episodes, this is just a drop in the bucket of Replit's ambitions. Their goal is to bring the next 1 billion software developers online and make them 10 times more productive. While it's hard to imagine the state and nature of software in that world, their vision for artificial developer intelligence and their track record for world-class execution make Replit, in my view, one of only 15 to 20 live players in the AI game globally. CEO Amjad Masad, also a former guest on The Cognitive Revolution, arguably put it best when he said that Replit is the perfect substrate for AGI.

• 7:37Nathan Labenz?

  And again, I'm not sure anyone has really thought through what that means, but I think there is a decent chance that it proves correct. In this conversation, Tyler provides practical advice for non-coders looking to leverage AI, an in-depth look at how Replit is using AI to accelerate development today, and a window into their relatively near-term plans for a virtual developer that will work alongside humans, eventually allowing users to "speak software into existence" by managing whole teams of AI agents that create tools and accomplish tasks in parallel. At times, this may sound like science fiction, but in all seriousness, if there is a single platform where the human-AI collaboration economy is most likely to take shape,

• 8:22Nathan Labenz?

  I would put my money on Replit. And just think what you might have thought sounded like science fiction just one year ago. The possibilities, mostly very much to the good, but also some very clearly to the bad, are endless. Many of the next billion developers will enjoy previously unimaginable economic opportunity as a result of AI. But I've also learned from personal experience that they will be highly AI-dependent and therefore vulnerable to an entirely new class of attacks, which Replit will have no choice but to confront.

• 8:58Nathan Labenz?

  Tyler embodies Replit culture as I've come to understand it. Extremely smart, first principles oriented, inclined to think outside the box, confidently optimistic, forthcoming, sincere, and always shipping. I think you are really going to like it. So, without further ado, I hope you enjoy this first half of Replit Week. This is my conversation with Tyler Engert. Tyler Engert, welcome to the Cognitive Revolution.

• 9:27Speaker 01

  Thank you. Thank you for having me.

• 9:29Nathan Labenz?

  Yeah, I'm really excited because you are a product designer at Replit. And I think Replit, regular listeners to the show will know, I name drop it all the time and cite it as one of the companies that is really punching above its weight in this current AI moment. So I'm excited to kind of dig into that with you and understand all the stuff that you guys are building and what you're seeing in terms of usage. I think it's going to be a lot of fun. Maybe just for starters, tell us a little bit about like what you do at Replit and then I want to ask you some big picture questions about the platform as well.

• 10:03Speaker 01

  What I do varies a lot day to day. I'll plug one of my own articles that I wrote about this called "The Mechanics of Work" a little bit later. But like in terms of physically what I'm doing, there's a lot of writing. There's a decent amount of coding and there's a lot of drawing shapes on screens, which you might imagine is the case for most product designers. But recently, I've been focusing a lot on AI. I was also working a lot on our version control. Previous to this, we just released an article about our redesigned Git interface, Git user interface. I've also been at Replit for almost four years now.

• 10:47Speaker 01

  I was the first design hire and the second designer at the company after Hyah, who was the co-founder. So I've pretty much worked on like every part of the product at this point and done a lot of zero to one stuff. But overall, I focus on the kind of primary coding experience, which we call the workspace, to people who don't know it or people who are familiar with like other IDEs like VS Code or IntelliJ or Platiom or whatever. The workspace is our IDE. It is the primary environment in which you are coding. Now, my main focus is basically everything to do with AI inside of there.

• 11:27Nathan Labenz?

  One of the most memorable tweets that I keep thinking about that I've seen over the last year, year and a half was from Amjad, the CEO, who said, "Replit is the perfect substrate for AGI." What does that mean? Where are you guys going big picture with your AI initiatives?

• 11:47Speaker 01

  There's a lot to unpack there in that single sentence. So the two key words in that claim are substrate and AGI. Hopefully, we know what perfect means at this point. But a substrate is like a material. It's like a form. It's something that something is made out of or something that a being or a thing exists in. The claim that Replit is the perfect substrate for AGI only makes sense if you have a particular view on like what is necessary to actually make AGI in the first place. Like, what are the missing components to get AGI to where we need it?

• 12:28Speaker 01

  And, you know, obviously, Amjad said this tweet, and it's his view. I also agree with it. So I'll try to defend it from my perspective. The reason it makes sense is because of the fundamental nature of what software is and what it can do. And the basic claim is that code itself, like being able to make software, being able to write code, is kind of the most generic medium. It's the most generic substrate that we have to create other things. Right? Even if you go as far back as the Turing machine, like more theoretical computer science back in the day from when Alan Turing was first founding the field, the idea was being able to simulate other machines.

• 13:12Speaker 01

  The idea of being Turing complete essentially implies that you can simulate other machines as well. This kind of universal platform to be able to create and simulate other processes. And the same is true today with code and software in general. Or we use software and we use computers to be able to simulate and work with many, many, many different kinds of use cases. So it's a very generic platform, generic medium. Because of that, because it's generic and because it can be used for many different use cases, the main idea is that software and code is kind of like the raw ingredients for basically doing whatever you want, as long as it can be expressed digitally. Now we can jump back to the AI definition. If the definition of AGI or...

• 13:55Speaker 01

  I'm not going to try to define it on this podcast, but for the sake of argument, AGI is a really, really smart thing, right? Really, really smart machine that can do a lot of different things, right? At or above human level, basically a human level breadth and diversity of different activities and tasks. And if it's, if it's supposed to do that within a digital environment, presumably it'll have to be able to create new tools for itself, right? Learn how to interact with its environment, right? In this case, the digital world for the most part, and be able to know how to compose and use those tools together in order to accomplish whatever task it needs to do. If the tools don't exist, if you ask AGI in 10 or 15 years,

• 14:39Speaker 01

  "Hey, we need to solve World Under. We need to discover the universal vaccine," whatever task you give it, presumably not all the tools that you need or that it needs in order to accomplish that task will exist at hand. So we'll need to create new tools.

• 14:58Speaker 02

  Hey, we'll continue our interview in a moment after a word from our sponsors. Hey, everybody. If you're a business owner or founder like me, you'll want to know more about our sponsor, NetSuite. NetSuite provides financial software for all your businesses. Whether you're looking for an ERP tool or accounting software, NetSuite gives you the visibility and control you need to make better decisions faster. And for the first time in NetSuite's 25 years as the number one cloud financial system, you can defer payments of a full NetSuite implementation for six months. That's no payment and no interest for six months, and you can take advantage of the special financing offer today. NetSuite is number one because they give your business everything you need in real time, all in one place, to reduce manual processes, boost efficiency, build forecasts, and increase productivity

• 15:39Speaker 02

  across every department. More than 36,000 companies have already upgraded to NetSuite, gaining visibility and control over their financials, inventory, HR, e-commerce, and more. If you've been checking out NetSuite already, then you know this deal is unprecedented. No interest, no payments. So take advantage of the special financing offer with our promo code at netsuite.com/cognitive. netsuite.com/cognitive to get the visibility and control your business needs to weather any storm. That is netsuite.com/cognitive.

• 16:07Speaker 02

  Omniki uses generative AI to enable you to launch hundreds of thousands of ad iterations that actually work, customized across all platforms with a click of a button. I believe in Omniki so much that I invested in it, and I recommend you use it too. Use Cogrev to get a 10% discount.

• 16:26Speaker 01

  And how will it create new tools? It will create new tools through code, through writing new code and deploying new software. So now, hopefully, the answer will make a little bit more sense because the idea is that Replit as a platform is all about these self-contained code-centric environments, right? Being able to write code, get it online quickly, and be able to deploy apps and APIs and kind of work with code very, very granularly, all within a kind of native internet-connected environment. And that's useful for people, obviously, because you don't have to worry about setup. You can just get started. You click a button, right? And your kind of whole environment is set up to be able to create essentially anything

• 17:11Speaker 01

  that you want, as long as it's, you know, available and usable on the internet. Those benefits also extend to machines, right? If a machine can automatically spin up an environment, right, and write code and execute it and test it, right, and do it, quote, air quotes, with its friends, right? You and I can collaborate in Replit, but if bots can also collaborate with each other on Replit 2, they kind of can make use of the same benefits that we get, but at a significantly faster rate and in a much more sophisticated way that we can necessarily predict. To summarize, it's basically the claim that, like, Replit as a kind of general-purpose

• 17:51Speaker 01

  computing environment, where it's very easy to spin up instances of these machines, be able to code in whatever language you want, be able to use tools and collaborate with other people and other machines, kind of makes it the perfect sandbox for a robot to kind of create whatever tools it needs for itself to accomplish increasingly complex tasks. In that sense, Replit is kind of like the substrate for AGI to kind of expand its abilities and become more and more sophisticated. Does that make sense?

• 18:22Nathan Labenz?

  Yeah, it's really interesting. You're right to flag out of the gate that, you know, a definition of AGI is definitely not something that everybody has agreed on or may ever agree on. I tend, because OpenAI is kind of the company that's pushing the hardest toward it right now, I tend to start with their definition as a point of departure. And I think theirs is pretty similar to what you said, basically, an AI that can outcompete humans on functionally all economically valuable tasks, which is, you know, quite a concept to ponder. Some people then will, you know, take it a little farther and kind of, you know, it

• 19:03Nathan Labenz?

  gets more godlike and sort of, you know, infallible or omniscient or whatever. But there is quite a bit of space, you know, presumably between outperforming humans and something that's like, you know, so powerful that it becomes like it never needs any help or, you know, can do everything totally, totally on its own. Anyway, that's really interesting and good to flag that that definition is a pretty fuzzy one. You know, when I use Replit, I open it up in a browser. And I do love the multiplayer mode, which I think is, you know, one of the first big things that kind of brought Replit, you know, to some internet fame in the first place. Do you envision AIs kind of working with the platform sort of the same way that I

• 19:48Nathan Labenz?

  do, which is to say like, do they need to be multimodal and do they sort of look at it, you know, with a computer vision model that they, you know, like as we've seen with GPT-4 and things like InstructBlip this week, you know, seemingly getting really good at understanding a scene? Do you envision them using a sort of visual modality or is it all just more code where like they can ingest sort of a state of the, you know, of the environment as text and kind of interact with it purely as text?

• 20:20Speaker 01

  I think depending on timelines, within the next year, probably no vision will be involved or no vision models, primarily dictated by just the availability and the reliability of those models, especially and also the cost to be able to deploy that at scale, you know, for millions of people. But I think even practically.

• 20:44Speaker 01

  You have to think, what is the minimum amount of information that a agent or a person or an AI, whatever, needs to be able to take actions inside of some kind of environment? And in the case of Replit, right, whether it's like writing code or debugging code, does it need to see, right, what you see? Does it need to kind of know where like the layout is in order to accomplish different tasks? If it's about showing users how to use Replit, right, and being able to guide around and be able to actually simulate like a full pair programmer in the sense that it really feels like another person is there, then yes, some form of vision model is probably going to be necessary for that.

• 21:26Speaker 01

  Or at least some sort of visual encoding of the state of the workspace will be useful for that. But you can get very far with just text-only models. If you've read the Sparks of AGI paper that came out recently where they're putting GPT-4 through all these crazy tasks and trying to test its kind of spatial reasoning abilities. Did you, did you see that part of the paper where it could like figure out how to navigate a maze? And like draw out the maze pattern that it saw? Even if GPT-4 is only fed textual directions of like making turns and like navigating this abstract space, right, with no kind of visual input,

• 22:05Speaker 01

  it's able to create some sort of internal model of where things are laid out relative to each other because of the spatial information that's encoded in the words that it's using, right? For left, right, up, down.

• 22:20Speaker 01

  It can do that partially because it was also trained on vision data before that and it's making use of that kind of transfer learning ability with the text-only output. That's all to say that there will probably be benefits to using partially vision-trained language models inside of the workspace, even if its only input and output is still just text, purely so that it can take advantage of those like emergent spatial reasoning abilities that it got from being able to look and describe scenes. That being said, also, you can get very far just with like, just with text and just with giving the model kind of essentially an API spec that describes exactly what actions are available to it

• 23:04Speaker 01

  and when to use them and how to compose them together.

• 23:07Nathan Labenz?

  So let's kind of reel in from this far future in toward the sort of mid-future, and then we can also look at what you guys have today, which in and of itself is pretty awesome. You're building the substrate for AGI. I've also, another very provocative tweet from Anjad is the one where he says that Replit is on track, roughly on track, to have a virtual developer by the end of this year. So what does that mean? What is my, what is my, you know, experience going to be using that virtual developer? Do I delegate projects to it? Is it more of like a pair programmer that works with me? What do you guys have in mind for this virtual developer concept?

• 23:48Speaker 01

  My bet in general for AI going forward is on agents. And, you know, we've seen a lot of hype around agents and agentic machines recently, namely AutoGPT, BabyAGI, open source projects like this, where, you know, they get a lot of critique, basically saying like, oh, you just put GPT-4 in a loop. Is that consciousness now? My short answer to that actually is kind of yeah. But the broader point is that people are really into this idea of having kind of these long-running, more complex processes, right? Plan me your project. Work with me, right? Work with me on this project over

• 24:32Speaker 01

  the course of a day. If you're working with a team, right, or on a group project, or at a hackathon, or whatever, the people that you're working with have very long-term context on the project and the complexities of it that you're, that you're working on together. You work with each other for weeks, right? You have multiple conversations across multiple modalities. You're, you're talking, you're, you're writing messages, you're emailing, you're on video chat, whatever. And it's very important to maintain that context and have kind of this, a sense of long-term memory and commitment to a series of tasks. It's not just in and out. I think humans tend to over-anthropomorphize things in general, and, you know, we do it with random patterns in

• 25:16Speaker 01

  Uh like moon formations and like, you know, surfaces of Mars. We do it with like little objects where we see faces. We like love things that feel like people. And that might come across, I think, as like almost a naive answer to like the AI developer problem like, hey, let's just make it behave like a human. Because the the your first like counter thought might be, wait, is that like too easy or too obvious? But I think it's actually the way to go. Obviously, there will be kind of AI specific benefits or AI specific behaviors that only they can accomplish. But in terms of kind of the process of working with somebody, I think the basic idea here is that a virtual developer or or an assistant in in Replit

• 26:02Speaker 01

  will be kind of just like, it will feel almost exactly like a teammate. But it will have AI specific benefits. Concretely, some ideas here are that it doesn't get tired, right? It doesn't have to go to sleep. Um. So it can do things for you constantly in the background. It will have access to kind of the entire Replit network, right? It's essentially uh connected to the Replit API uh by default, which as an as another asterisk, you know, there's no public API, but we own the service, so it can obviously access our database and our network. So presumably we'll be able to use bounties.

• 26:44Speaker 01

  Uh and bounties are Replit's uh software marketplace where you can post, you know, tasks that you want completed or projects that you want done. It will have access to the to the rest of the community, right? So people posting projects, sharing, talking about them, that kind of thing. As said before, it also has access to your workspace, right? And your projects that you're working on and all of your previous conversations. So I think it's going to extend far, far, far beyond just chat and autocomplete. Most AI apps um and products you see nowadays are essentially just glorified forms of messaging and autocomplete, which are both very useful in and of themselves. But the idea of kind of like this pair programmer that isn't just good at writing code, but can, you know, it can use.

• 27:29Speaker 01

  Product and project planning software, right? What if it can use Asana or your notes or your to do's app? It can help write notes, organize them for you. If you, if it knows its own weaknesses and doesn't know how to accomplish something, it might go put a bounty out, right? For a human to come in and help out. One idea I'm really excited about in particular is the idea of being able to put your bot or your ghostwriter out for work. In the sense that if you're working with it already for like a month, two months, three months, it can start to learn and understand what kinds of software you build and what you're good at, right? And start to replicate some of your own behavior. In which case you can kind of put your own ghostwriter out for hire on the bounties marketplace.

• 28:13Speaker 01

  Uh, and then suddenly, you know, your ghostwriter that is going out and seeking out help can hire my ghostwriter. And now our bots are literally collaborating together to accomplish the tasks. And instead of being like a pair programmer, uh, now we, you know, you and me are more like managers or C or like kind of these like mini CEOs. And we're just, you know, overseeing the actions and the output of all of these bots. And it's not even just like, you know, it's not just that we're prompting, you know, like purely in terms of like, you know, a text input. All of the information of the entire environment, you know, from our behavior to the kinds

• 28:54Speaker 01

  of projects that we like and work on to the rest of the community are all fed in as context. So it's really like having, you know, a team of virtual employees, uh, or even like this co like human robot community that, uh, just works with each other. I kind of went off the, the deep end on that explanation, but, um, I think my main point is that, you know, pair programmer is like a really good kind of first milestone in terms of things, you know, an assistant that can write code with you and understand your project. But in reality, a software project is usually much bigger than just writing the code. It's the planning, it's the people that you work with, it's the community and the distribution of it.

• 29:35Speaker 01

  It's everything that goes, uh, it's basically following the entire lifecycle of an idea to production, shipping to production and having an assistant that can, you know, jump in at any point there. Does it make sense?

• 29:48Nathan Labenz?

  We talk about the. You put GPT-4 in a loop, is that consciousness kind of? I definitely want to follow up on that. The bot-to-bot interactions is another dynamic that I think people are way underestimating and definitely want to dig in a little bit more on that as well. The sort of scope also sounds really like interesting, ambitious, but also like kind of scary. Like, I would think doubly scary for Replit as the owner of the service. But even, you know, just for me as like a random citizen, I'm like, man, you guys sound like you're going to give this thing some pretty high-level access, or at least, you know, are on the trajectory to do that.

• 30:27Nathan Labenz?

  So all of those, I think, are things that definitely merit a kind of deeper scrutiny. But maybe before we return to that, let's just talk about where you guys are today. Because I think, you know, for folks that haven't used the platform, and, you know, maybe aren't even familiar with the company, although I mentioned it often enough that folks should have at least heard about it. But, you know, for folks that don't code, or, you know, they may not know the details. I would say, you know, in terms of AI coding assistance, the one-two, you know, power rankings and different folks will have different assessments, pretty clearly in my mind today, Replit and Microsoft Copilot.

• 31:08Nathan Labenz?

  So you guys are under 100 people still, last I heard, and have built this like massive computing stack that kind of does it all, so to speak. And now, you know, getting into the AI layer as well. It is extremely impressive how much the company has been able to build with such a, you know, modest, such modest team. Maybe let's just run down kind of where you're at today. You have a $10 a month premium product called Ghostwriter that I do subscribe to. I also subscribe to Copilot. They're, you know, both are an outstanding value at $10 a month. Tell us like what it can do today, how people are using it, and then we'll kind of build

• 31:52Nathan Labenz?

  back up to that future. But just like, let's ground, you know, this discussion now in terms of what exists today and how it's already helping people.

• 32:01Speaker 01

  Ghostwriter is the name for our suite of AI tools, right? It's the, when we say our suite of AI tools, there's kind of three main components to that right now. There's our autocomplete, right? Our smart autocomplete, which is kind of what took the software engineering world by storm when Copilot first came out, right? Let's write a comment, describe what my function should do, and Copilot will fill in the rest. That kind of behavior where, you know, you could be ghost text, you know, this like grayed out text of the suggested code that should be written inside of your editor and can accept it or deny it. That's one part of it.

• 32:44Speaker 01

  Some of our earlier viral tweets were also around features like Explain Code, Generate, Transform Code. Generate is not so novel anymore, but Explain and Transform were particularly interesting at first because, especially in a educational and learning context, which a lot of people use Repli for, it was very, very valuable for people to just, you know, highlight a piece of code, especially something that someone else wrote and be able to get a human readable English description of what it does. At, you know, at varying levels of abstraction or complexity, depending on how deep you want to go. Same thing with Transform where, you know, if you want to say convert a snippet of code from one language to another,

• 33:27Speaker 01

  you can, you can do that, say from, you know, Python to JavaScript. Or let's say you want to migrate a legacy code base to some more modern stack. I'm assuming this is what a lot of people in finance are probably doing right now with all the legacy finance software out there. You can, you know, highlight and select regions of code in a project and ask Ghostwriter to basically refactor it for you and take the equivalent logic and move it to more modern technologies. So we have so far autocomplete. We have Explain, Transform and Generate, which operated on specific regions of code. And recently, we announced Ghostwriter Chat, which is essentially our built in version of a chatbot,

• 34:13Speaker 01

  you know, a chat Q P T like service that you can talk to and ask for assistance on your project and it can write code. If you don't have a kind of broad overview of all of the, of all of the available tools in the market today. They look very similar. You know, it's like, oh, there's a input box. I type some stuff in. It like talks back to me like, what makes this different? You know, why would I, why would I pay for this thing? Obviously, it's very early and we still have lots more room to improve on it. But the main point of Ghostwriter Chat specifically is that it is connected to your project. Often, if you see, that's important for a few reasons.

• 34:52Speaker 01

  But before Ghostwriter Chat, if you want to use Replit to say, run the code that ChatGPT is giving you, right? You're, you maybe ask ChatGPT, let me make a game, you know, or let me make, you know, a simple HTML website. And I'll output all this code. People will just, you know, copy and paste it in to Replit. It works, you know, for all the other reasons the Replit is good. But the point is that you can just go very quickly from outputted code to something that's running in live. But if you ever want to make edits to your code, right? Or tell ChatGPT about your project, you would have to do this clunky copy and paste thing. You know, you go back to Replit, you copy, copy a bit of code. You have a question about copy, paste it back in.

• 35:33Speaker 01

  And it only has a very, very limited view into the complexity of your entire project. Right? It's only, it only knows, in this case, ChatGPT only knows about the pasted code that you give it. Ghostwriter Chat, um, the way that we built it is that it has full context of your entire project. And what that means concretely is that we do a bunch, a bunch of fancy prompt engineering to basically give Ghostwriter Chat the right context, uh, at the right time, um, for the, the preamble of the prompt that we give it. So that it knows information about your project. That includes things like, you know, where your cursor is, what file you're currently in, uh, the structure and the directory structure of the project.

• 36:16Speaker 01

  That kind of thing, which shouldn't be news to anybody. If I said that Ghostwriter has, you know, context on your project, you would assume that that kind of information is included. You can imagine that we can go much deeper into that as well. That's kind of where we're at today. We're slowly starting to roll out and work on more, um, action-based features for Ghostwriter.

• 36:38Speaker 01

  Instead of just Ghostwriter outputting text and being able to, you know, it can still take in context and understand your project and and make edits to code. It's all limited to the chat window right now. We're working on Ghostwriter being able to write, to actually write two files, you know, interact with the file system. Be able to read new files on demand and kind of act as like an operator on top of the entire container. AKA the actual running computing environment under the hood of every project. And from that, it's like, you know, you have, you have file system operations. Then you have the actual tools that are available for people to use as well, like our docs, our databases.

• 37:20Speaker 01

  And right now even there's, if you have a bug in Replit, you might notice that occasionally a little purple box will pop up that says debug with Ghostwriter. Close to mark where Ghostwriter now will notice when particularly complex errors pop up. As you're debugging, developing your program and you can kind of send that error message directly to Ghostwriter to fix. But you can imagine that we're working on getting Ghostwriter to actually debug for you and debug with you as well, more granularly.

• 37:54Nathan Labenz?

  Yeah, so I've used all of these features. I'll maybe just give a little experiential view again for those that, you know, haven't done it or don't code. If you are, if you do spend any time coding and you have not subscribed to, honestly, I would say both Copilot and Ghostwriter. Like, I don't know really what you're waiting for. The ROI on these products is just insane. I remember when I was, this was pre-Ghostwriter, but in the, just before Copilot went to a paid model. I was thinking to myself, and it was free and I had like the free, you know, trial version. I was thinking to myself, what would I pay for this? If, you know, they were really trying to maximize how much money they could get out of me, like, what would it be? And I kind of came to the conclusion that like.

• 38:40Nathan Labenz?

  It's probably worth a thousand dollars a month to a, to a company that has like a full-time developer. You know, you figure you're going to be paying your developers, whatever, something into six figures, you know, pretty commonly, right? So you're figuring 10,000 a month, so is it worth an extra 10%? And I was kind of like, yeah, I don't see any reason that it wouldn't be worth, you know, 10%. I think the ROI would probably be there. So the fact that it comes in at $10 a month, you know, isn't an extra 0.1% in my mind, you know, by multiple, you know, get, get CoViolet and Ghostwriter. One of the first things that I remember so strikingly, because it just say, it was like that paid for itself,

• 39:23Nathan Labenz?

  you know, for months right there, was I was trying to just do some file manipulation. You know, this is in the context of Waymark, which is my company. We have a AI-powered video creator. It's a highly structured process and we like manipulate files in different ways. So one of the things I was just trying to do is separate the audio from a video file and have like the audio output. So naturally you're going to, you know, have some specialized library that does this. And ultimately it's FFmpeg that I need to use. And I don't really know anything about that. I like kind of know that it's out there, but I don't really know, you know, how to use it or all the arguments, flags. I think it's gnarly, right? So what I would have had to do is go search, you know, for this.

• 40:08Nathan Labenz?

  Google and maybe I'm on Stack Overflow and, you know, whatever. I'm trying to find these examples. Do I have the right example? Have I understood it right? And you can imagine easily, and I've gotten bogged down in this, and anybody who's, you know, spent time developing has had this experience where you're like, oh God, this conceptually is so simple. I just want to do X. I know it is possible to do it. I know probably thousands of people have done it before and yet I can't quite track down exactly the way to do it. So it was like a whoa moment for me when this is probably eight months ago now. I literally just typed a comment, separate, you know, create a MP3, you know, of the, of just the audio from this MP4. And then the next thing you know, it writes the whole thing for me.

• 40:50Nathan Labenz?

  It wasn't even that much code, but the key point was that it was correct and that it happened in a second. And then I was like, oh my God, I can just do that and it works. And that's kind of still just the first paradigm, right? Where it's just kind of doing that in line auto completion. As you, you know, mentioned, you've got multiple different modalities for kind of interacting with the language model. For anybody that is curious about the prompt engineering that goes into this, definitely recommend looking up a big teardown of not the Red Bull, but the Copilot prompt that somebody went real deep on and published all their findings. It's probably a bit out of date now because obviously all this stuff is

• 41:34Nathan Labenz?

  advancing and Copilot has, has certainly, you know, had more releases since then as well. But it does give you a good window into how to think about like, okay, there's all these different files, what really matters? And I've got limited context window and all that kind of thing. How do I make the most of it? Sure, you guys are, you know, wrestling with, with very similar issues. So that would give, at least give a directional insight into the kind of challenges that you guys are solving. How much productivity boost do you think people are getting from this today? Is there any way to measure that? Or, you know, if not, like what kinds of things do you measure about the way people are interacting with AI within Red Bull?

• 42:13Speaker 01

  Well, I remember Microsoft actually released a study or some release about Copilot's impact on productivity. And the experimental setup was pretty simple. You know, it was something like both the experiment and the control group had a series of tasks that they needed to get through, like coding tasks, whether it was writing functions or like testing stuff. I forget exactly which. And they just measured, you know, those with and without AI, like how quickly did they finish? And I think the with AI group ended up finishing all these tasks like 40% faster, 50% faster. It was some like ridiculous jump in just time saved. If we're going off of

• 42:56Speaker 01

  Numbers like that. And obviously, that's copilot. We don't have, we haven't run like a specific study like this internally with Replit. You know, all we have is kind of our raw data, like our raw telemetry on usage numbers, but we haven't run like a, you know, scientifically controlled setup. You could imagine that it's probably on a similar order of about, you know, twice as fast, right? Two, two times the time saved in terms of completing tasks. I think where it gets more interesting, however, is all of the compound effects of saving this time.

• 43:36Speaker 01

  And the compound effects of learning and building more quickly specifically. You know, it's easy enough to say like, to put a timer on and say, how quickly can you make this website? Right? But if as a part of that, right, you are asking Ghostwriter questions, right, about your code. And you are suddenly more knowledgeable about how it works, right? Because it's giving you more coherent explanations and it's giving you more accurate information than you would get just by Googling things. Then you are also smarter the next time you ask the same question, or you are smarter next time you have to work on a similar problem.

• 44:18Speaker 01

  And you can use that new knowledge to build more complex things. So I want to figure out a way to measure the compounding effects of these productivity boosts and see, you know, like over the course of maybe three months, six months, a year, somebody who has versus doesn't have Ghostwriter, for example, how much farther ahead are they in terms of, you know, computing literacy, ability to explain concepts even, you know, like how much does that affect your productivity and how much does that improve from control to experimental group?

• 44:50Nathan Labenz?

  So Microsoft basically reporting almost 2X speed up as of, you know, sometime prior to May 2023. And obviously, we're kind of just getting started. Do you have a sense for how far this goes? Like, at the end of the year?

• 45:10Nathan Labenz?

  What do you think that might look like? Are we talking? Because it's pretty hard for people to wrap their heads around, I think, a multiple productivity increase, right? We have, you know, total factor productivity growth is like low, single digits in today's world. And now we're talking about something where like a whole major segment of the economy stands to maybe get a multiple speed up boost? Is there anything that's like bounding that in your mind? I mean, is there like certain parts of the software industry that this just can't impact? Or are we really looking at like developers can be two to five or even maybe more, you

• 45:53Nathan Labenz?

  know, times as productive as they are today in the not too distant future?

• 45:58Speaker 01

  Yeah, so I think this is actually where the cost of a, an assistant and agents and like the bots to bots stuff comes back. So in terms of how far it goes, the bottleneck here is the ability to parallelize, parallelization. Right now, all of these boosts, right, are kind of starting with the assumption that the main bottleneck is essentially typing speed, right? And maybe like some information retrieval, right? Like Googling stuff takes time. The main point is that you type a little bit and you get a lot out, right? Versus the kind of default scenario is that like every keystroke that you write or like every key press that you have is kind of like the only output that you get, right?

• 46:43Speaker 01

  It's like one to one, like keystroke, every keystroke that you input is one keystroke that's outputted onto the screen versus what AI unlocks is like for every keystroke that you input, you get multiples more output of text on the screen. So that kind of like increases your productivity by a, by a factor that is like linearly proportional to. And I think what I mean by that is like. We are bounded by the, like, we can only make that kind of stuff so much faster, right? At the end of the day, it's basically just like a supercharged version of typing, which is really, really fast, but you're not going to get a hundred X speed up in terms of the things that you can make just by making typing faster.

• 47:28Speaker 01

  It is kind of a serial or like, you know, a sequential activity by definition, right? You're typing, you get more stuff out and you're interacting one to one with this little robot that is kind of helping you execute on one task at a time. If we want to get to the point of like a hundred X productivity, a thousand X productivity, where, you know, one, two, three people, right? Three-person startups are essentially running the equivalent business of what are now today, like 500,000-person companies. We need to parallelize. And this is where the bots and the agents come in, right? Where, you know, instead of just essentially speeding up the process that exists currently, where you're just speeding up the act of writing text,

• 48:11Speaker 01

  you have multiple employees essentially, right? Multiple agents or bots that are working with you in parallel to execute their own series of tasks. So it's really about, you know, instead of vertical scaling, right? We are horizontally scaling. Basically, all the labor that's necessary to actually create, actually able to create software at large scales.

• 48:34Nathan Labenz?

  What are the pieces that are kind of currently missing that need to be built for that type of vision to come online? Like, Replit recently open-sourced a new code model. And as I'm listening to you, it's, it's starting to sound like that code model is maybe the base on which you guys are presumably like, fine-tuning the sort of internal Replit version that, you know, has a lot more context on like your internal APIs and, and all that sort of thing. So I think maybe one piece that needs to be built is sort of... Very specific knowledge of the right tools so it can really go deep on a particular platform.

• 49:20Nathan Labenz?

  Context window, we've seen huge breakthroughs in just the last couple of weeks. I think, whatever, the, on an academic level, the breakthrough maybe came a year ago. On the practical, productized level, we've seen breakthroughs in the last couple of weeks where context windows are exploding. I don't know if you can share what kind of context window you're working with right now and how you're thinking about managing that and how that also may become a problem you don't have to worry about for too much longer. But those are just kind of my groping around in the darkness a little bit to try to imagine like, what is either beneath the surface that you guys have already created that's starting to power this? And, or maybe like what you're likely working on or about to be working on in the near future? What more can you fill in there in terms of like the gaps that need to be bridged to get us from this

• 50:05Nathan Labenz?

  sort of 2X to, let's say, 10X plus future?

• 50:09Speaker 01

  Or 100X. Shoot for, shoot for the stars. In terms of fundamental breakthroughs that are needed, fundamental like industry level problems that kind of everybody's contributing towards, larger context windows is obviously one of them. I think solving the problem of, but that's more technical. That's more of like a kind of straight up engineering problem, right? Getting more memory to fit in. I think on a larger scale, the ability to actually plan and reason and work with tools reliably and follow instructions without necessarily needing like a fine tuned instruction labeled data set,

• 50:52Speaker 01

  you know, like 800,000 examples or whatever of like following instructions. Getting models to kind of behave more in a particular direction. I'm basically just saying we need to be able to get models to be fine tuned and steer it in certain directions much more quickly with much less data. I think that's kind of one fundamental piece that is not. Necessarily missing, it's just it's being actively worked on. It's definitely a big bottleneck. There was a recent paper that was released, or I saw some tweet about it that was talking about like activation engineering and like kind of feeding these, instead of feeding in prompts as context, you are, or feeding in prompts as context to a language model,

• 51:36Speaker 01

  you are essentially giving it vectors, raw vectors that can be processed by the initial input layers in the network to be able to steer the output in a certain direction. That's pretty promising. Beyond that, I actually think it's kind of just a bunch of boring work that is the main model deck or the thing that's missing. Language models right now are already really good at reasoning overall and like reading comprehension and obviously really good at data generation and synthesis of of new text. I think it's really just a matter of like using the existing tools at hand

• 52:15Speaker 01

  and combining them in unique ways to try to kind of act like these big planning and execution machines. I think all of the pieces for the most part are kind of there. Like the fundamental primitives, I think, exist. And people are, you know, working on more specific technical parts of the problem, like long-term memory, bigger context windows, better information retrieval, you know, accounting for hallucinations, being up to to specify a specific input and output formats and validate them. These are all definitely necessary, but I think the core is here and it's really just about figuring out how we compose them together. Get the models to work the way that we want to be able to execute and plan on tasks in parallel.

• 52:58Speaker 01

  To your earlier comment about, you know, the open source models and contributions there. One of the other bottlenecks there is like, we need more open source. We need more tools that let people kind of tinker and fine tune models very, very, very easily. You know, right now, the only reasonable way to get people to fine tune models is still through OpenAI's fine tuning API. And they did a really great job. There in the sense that all you have to worry about is curating a data set. And then you just send it to them and you get your own, you get your own link to a inference server. Where you can ask the model things, but we need to radically, radically decrease the barrier to entry to fine tuning. And for people to tinker with these models.

• 53:44Speaker 01

  And you know, companies like Hugging Face are doing a really great job at, you know, exposing. Models and open source packages to people in a very easy to consume way. Beyond the actual technical components, the tooling and the infrastructure around allowing people to fine tune and experiment is a huge bottleneck. That hopefully, Replica can also address.

• 54:02Nathan Labenz?

  One thing I'll just highlight as a point of agreement is. I also have a pretty strong sense that the, the fundamental breakthroughs basically have already been made. And now it kind of is a matter of. You know, polishing those, refining, working through the QA, you know, sanding down the rough spots, patching data sets, you know, to, to overcome. You know, predictable weaknesses as well as integrating with these like retrieval tools. And just, you know, integrating with systems more broadly. This, you know, makes me recall something that I experienced in the GPT four red teaming experience. You know, even then, and this was last fall, right in the, in the unpublished version. I would ask it to do things like.

• 54:46Nathan Labenz?

  Self delegation, you know, in, in the context of programming, like set up essentially a little REPL and, you know, have it kind of. Break down problems and delegate to itself. And even then, in the, in the kind of first version, it was like. Usually the plans that it would come up with and the general approach to. Problems and usually even the way that it would kind of break those problems down into sub problems. It was all pretty reasonable and it was like, wow. This thing has a pretty good handle of. Conceptually like what I'm asking for and how to approach it. And if I were to have a conversation with somebody and they gave me this sort of. Plan for success, like I would think they're off to a pretty good start. Then when I would get down to the weeds of it, a lot of times it would be like, oh, well.

• 55:31Nathan Labenz?

  You know, it's, it's kind of stumbling on some very low level. Implementation details, just a stupid little toy experiment that I ran was. This was just after the Queen Elizabeth had, had died. Go online and find out who is the current reigning monarch of the UK. And I tried that because it was like, it's a very high prior, you know, for the cutoff date of like, what the answer would be, but now there's a different answer. So can it go out online, you know, figure out where to look, how to look, et cetera? You know, it's a very simple task, probably wouldn't surprise anyone that like, it could come up with a reasonable plan for that task. And then I just noticed, you know, it would get stuck on these, these little things where

• 56:12Nathan Labenz?

  it would be like, go to BBC.com, look at the H1 tag and see what it says. And it was like, well, you're close, like you got, you know, a pretty good plan, but you know, there might not be an H1 tag, or the H1 tag might say something a little bit different, you know, than the specific like, string matching, you know, that it, that it would use. One weakness that it did have is it didn't take full advantage of itself. I started to prompt it with things like, you can interpret natural, you don't need to write regular expressions to interpret language, you can use yourself to interpret natural language and like, answer, you know, questions for yourself. So it was able to pick that up as well with prompting pretty easily, but still would just, you know, tend to get bogged down in these like, very low level details.

• 56:56Nathan Labenz?

  So it sure seems to me like, you're correct that the core kind of reasoning ability, planning, understanding, basically is there. And then it does seem like we're just kind of a matter of collecting a bunch of sort of failure examples and kind of patching them and, you know, little reinforcement learning around the edges, whatever. And my hypothesis has been that even though agents broadly, like, don't work today, that we're probably no more than six months away from that kind of all coming together as enough of those rough spots get sanded down, you know, enough of those examples get kind of fed back into training data.

• 57:39Nathan Labenz?

  Also, to some degree, the tools themselves are made more accessible for agents. I think we'll see a big trend of like websites, you know, having sort of an agent friendly, you know, presentation, literally just designed to make it easier. What do you think about that? Does that seem right? Are we talking like a six month timeframe here?

• 57:57Speaker 01

  I mean, I'm always more conservative with my time estimates because I like to be surprised. So I would say something like a year. I mean, the, the idea of like an agent friendly view for websites so that it can consume information. That's pretty interesting. And it's, it's funny then, because then you're like, well, now designing websites well is not just for humans to be able to comprehend it, right? And be able to see information really clearly. It's also for these bots to be able to read in on these campaigns too. That's why I said kind of the core pieces are here, right? In terms of being able to like reason and plan and figuring out like the steps necessarily. Generally speaking, I think the core parts are here. The thing that, as you described, that is lacking is actual execution.

• 58:40Speaker 01

  All these remote browsing technologies, like relying on, you know, parsing HTML and like figuring out how to actually extract the right data from the web. That's going to be a much harder problem, I think, to solve than, than we think. Either that, or I'm underestimating vision models. What we might start to see is, um, vision models will definitely take off. But I think unless, you know, the price of using them is very, very low. I don't see like just, you know, constantly using vision models to like autonomously browse the web to gather information as like a sustainable path forward for, you know, the next year or two.

• 59:19Speaker 01

  I definitely see some sort of like agents or like AI friendly component of websites in the near future. And maybe all that means is that more and more websites by default will have a API. You know, instead of going and visiting a actual webpage. Right to go and get the right information for it. If a website complies with this new like AI standard, right? Where it's like, you know, you have to, you know, part of accessibility won't just be using, you know, the right color scheme for humans and, you know, proper HTML tags, right? To separate sections, but will also include accessibility for robots. So that you make sure that you provide, um, you know, as the website owner,

• 1:00:00Speaker 01

  uh, an API for the AI to consume that perfectly mirrors all the, um, content that's available for humans to consume. So that's all to say, some sort of like kind of AI first, um, consumption format will probably come around. Um, whether or not that's, you know, just a standard REST API is another question. I think it'll be some combination of that kind of format plus stronger, um, multimodal models that you can use vision to extract stuff from regular pages.

• 1:00:29Nathan Labenz?

  Yeah, that totally makes sense to me. If you haven't seen Instruct Blip yet, I would definitely recommend taking a look at that because we had the authors, uh, two of the authors on a previous episode when they did Blip 2, which was at the time, you know, state of the art image understanding kind of dialogue about images. And now they've taken it up, you know, yet another level. And it really looks like it's kind of GPT-4 vision, you know, equivalent in terms of the detail of the descriptions that it can return to you. Actually, I need to double check this, but I believe they've started to incorporate UI into their data set as well, uh, which certainly OpenAI has.

• 1:01:12Nathan Labenz?

  So yeah, it seems like all this stuff is kind of coming at us extremely quickly. You know, going back to the substrate for AGI, right? Obviously there's a lot of difference of opinion, to put it mildly, around like, should we be excited about this AGI prospect? Should we be fearful of this AGI prospect? I personally feel both of those things, you know, and I don't have any struggle in feeling both of them. They both come pretty naturally to me. I find it also, you know, super exciting, like I could be 10x more productive. You're just talking about 100x more productive. I don't even know what that would mean, but it sounds pretty amazing. But then I'm also like, if something's that powerful, you know, then certainly at a minimum, you know, the developing of it comes with great responsibility.

• 1:01:55Nathan Labenz?

  So I kind of want to understand a little bit more how you guys are thinking about that. Cause it seems like you have. A really hard problem in some ways where the typical things that people talk about doing to kind of maintain control over AI. One of the obvious things people talk about is like, well, we'll just sandbox it. Like, we won't connect it to the internet or we won't give it, you know, the ability to kind of get outside of its like immediate environment. Now, some people would argue like, you won't be able to do that even if you try. But you're kind of saying, we're going to give it all the APIs that we have, you know, as the owners of the service.

• 1:02:32Nathan Labenz?

  So I understand that to mean like you envision a future where an AI agent on the repl platform could like spin up a new REPL and kind of bounce over from one computing environment to another. And, you know, when you talk about parallelization, like maybe a lot of those perhaps even when you're giving, you know, maybe not root access, but like increasingly close to root access to your whole stack to a language model that is fairly well understood, but definitely not fully understood. And you add in these additional dynamics of like agent to agent interactions or like

• 1:03:12Nathan Labenz?

  you also alluded to custo individually customized models, which adds a whole other layer of kind of unpredictability, like what is Tyler's model going to do versus Nathan's model for any given situation? Plus, on top of that, you've got people coming in with like adverse intent, you know, who might be today, a lot of times that's like, can I get you to leak your prompt? But in the future, if you've given it access to all these underlying APIs, you know, I might be less concerned with getting you to leak your prompt and more concerned with like, how can I spin up a thousand Bitcoin miners? You know, I know you guys have battled that already, but you might be battling it in a whole different way. So how are you guys thinking about all that? Cause it just seems like an unbelievable. Amount of complexity and just unknown unknowns.

• 1:03:58Speaker 01

  The the TLDR is that we take it very seriously and we're thinking a lot about the safety of these tools. And especially if they're going to be used presumably to help control and write mission critical software at some point. And especially if they're open to the internet, right, or, you know, have network access, right, and can talk and coordinate with each other. There's, you know, a whole host of nightmare scenarios that can emerge from that.

• 1:04:26Speaker 01

  But maybe to answer a few things that you brought up specifically, Replit is like the place for sandboxing. By definition, a Replit project is a sandboxed, self-contained environment. We have dealt with a lot of abuse problems before. We have a lot of practices in place right now to easily detect abuse from all of our metrics that we record from activity inside of inside of Replits. In terms of people being able to just spin up Replits as needed and to use that like for Bitcoin mining

• 1:05:06Speaker 01

  or even for if they're like using agents or like, you know, if they're using Ghostwriter to kind of control and manipulate like hundreds of Replits at once, that's a future that we don't need to see, right? Like, we can, we can enforce and put restrictions in front of the use of Ghostwriter, similar to the ones that we put in front of people to begin with, right? We can rate limit it. We can give it specific guidelines for like how it should behave. Anthropic's release of their constitutional AI, but like all of their techniques for kind of encoding values into its output, I don't even think, like, our, our problems aren't even as difficult as that, I don't think. We can basically just use traditional engineering techniques to be able to like limit

• 1:05:51Speaker 01

  its access to what it can do without human intervention. I think the key is kind of taking a, a tool-first approach to this problem in the sense of just viewing, viewing Ghostwriter, viewing AI, viewing agents as tools that we use that we deliberately control, right? That we kind of give leeway to, to behave autonomously when, when we want. At any point, if there's kind of like vulnerable part of the process. Where it's developing things, um, you know, say it's asking for input, right? From a user, like if I'm making an app, right? And uh, I have an agent that's assisting me in like gathering user research, right? And I'm telling it to go out and interview people.

• 1:06:34Speaker 01

  That's a case where, you know, maybe it won't execute on interviewing that person unless I'm there. Right? Or it has its ability to research on the internet. Um, or to execute and talk to other bots suspended in the process of while it's interviewing other people. That's all to say, it's a really complicated question to answer, but it all will boil down to very deliberate measures to make sure that whatever systems we build uh, have very specific levels of access to different um, networking technologies, depending on the tasks that they're assigned to do. And always making sure that, you know, even in a world of very, very high autonomy, uh, where maybe you're managing like dozens or even hundreds of agents,

• 1:07:19Speaker 01

  all of the uh, potentially moral or ethical decisions that might come into play will still require a human in the loop to make sure that it's operating smoothly. So that's my take.

• 1:07:30Nathan Labenz?

  So how, how far do you envision this going in the like? Maybe I should just say foreseeable future because I do feel like there's some, you know, event horizon beyond which you, it's really hard to see anything. You mentioned earlier, like your agent potentially tapping into a bounty type marketplace, you know, setting up. So if it, it starts to sound pretty hard to control in all honesty there. Like if I have an agent that can't do certain things or it's been sandboxed, but it can go ask other people to do stuff. You know, we've already seen from like the G P T four technical report, the, and I was like tangentially, I wouldn't even say I was involved, but kind of tangentially collaborated with

• 1:08:14Nathan Labenz?

  the, the group at Arca a little bit on this where they were, you know, creating situations where, for example, the A I would ask a per, you know, hire a person to solve a captcha. You know, that's, that type of pattern. Seems like it comes back, you know, pretty naturally, right? It's pretty easy to, if you have this sort of very bite-sized gig, you know, mini-project-style marketplace, and you're sitting above all this and sort of saying, right, here's my, you know, master plan of how I'm gonna take over the world or whatever. But down here, I need like somebody to solve a captcha for me. The AI in that, you know, technical report example lied to the user, right?

• 1:08:57Nathan Labenz?

  And said like, I shouldn't say I'm an AI or, you know, it probably won't do it for me. So I'll just say I'm, I think it said I'm blind or whatever. It feels like these problems go a little bit beyond sandboxing. I don't know, like, I just want to push you a little harder on this and say, if you're kind of allowing the, these agents to even commandeer, you know, the intelligence of humans, as well as other agents, can we really be that confident that this is gonna stay under control?

• 1:09:25Speaker 01

  Part of this answer goes back to actually having a centralized platform. If this is a closed marketplace, right? Where like you have to be a registered Reddit user in order to use it. And, you know, anytime a bot interacts with the marketplace, it's like specifically labeled that it's a bot. It's a much, you know, questions like this are much different when you have a system that's like deliberately designed to manipulate people and like doesn't have to reveal its identity or have any kind of like, you know, external mark that it isn't a bot, which is the case of the GPT-4 example that you brought up, right?

• 1:10:04Speaker 01

  Honestly, the majority of this boils down to, I think, how well we communicate what a bot is doing and when it is being enacted on by a bot, which maybe feels a little bit like a cop-out answer. But the kind of main danger about everything you're talking about is deception, right?

• 1:10:26Speaker 01

  And people being fooled and tricked. And that's like the whole point behind social engineering in the first place is you're using all these like weird psychological tactics to get people to do what you want. You know, the key component behind like any form of deception is the deliberate withdrawal of information from the receiving party, the decey.

• 1:10:47Speaker 01

  In this case, you know, I don't have a specific answer in terms of like how exactly we'll implement it. But my direction that I'm going in thinking about this is like, we just have to make it extremely obvious like when things are being done by bots. And that goes as far as like the user interface, right? Like whenever something is written or done or executed by a bot, whether it's in the community or in the marketplace or in the workspace. It's just very obvious that it's attributed to the robot itself. And that being said, if some malicious user, right, is like instructing a bot to

• 1:11:29Speaker 01

  behave on its behalf, right, and go and post bounties, right? Or go and find workers and recruit. You know, the example that you brought up about, you know, the low level task by itself is like not malicious, but it's in service of some higher level goal that is malicious ultimately. This reminds me a lot of like, you know, stories of World War II and people who are essentially employed by malicious governments who defend their actions by just saying, oh, it was part of our job. We were just doing our jobs.

• 1:11:58Speaker 01

  And we want to avoid a situation where essentially bots have to defend themselves. In that case, where it's like, oh, I didn't realize that you were trying to send the nuclear war codes. I was just doing my job by posting a bounty and, you know, needing,

• 1:12:13Speaker 01

  needing you to solve a captcha or something. Maybe a core component here that is kind of the hardest problem, it seems, is being able to make sure that when bots are enlisted or trusted upon to, you know, behave and interact with the community by posting bounties, completing them, whatever, that they are aware and able to analyze like the higher level goals that they are in service of. They can't be aware of the current bounty as its only context. They kind of have to play, I think, a little bit of detective work to work backwards and see like, what is this person actually doing? They asked me to do a bounty to like, you know, ask somebody to complete a captcha. Yesterday, there was a bounty that, you know, asked me to figure out how to mass send out emails, right?

• 1:12:57Speaker 01

  Bounty was for some sort of like fake password cracker, right? You know, I'm, oh, I'm teaching like cryptography, right? Give me an example of how to crack a password. Each of these examples in isolation might be benign, right? Like one is education, one is like automating email and one is automating some other task. But if you, you know, triangulate pieces together, it's like, oh, somebody might be coordinating some kind of phishing attack. Right on replit users. And I think having some sort of more complex like anomaly or malice detection algorithm.

• 1:13:38Speaker 01

  And have the agents or the bots kind of responsible to do that, like as they're completing things. Can be a big part of how we prevent malicious behavior from users and having, having bots essentially be abused for bad behavior.

• 1:13:52Nathan Labenz?

  I mean, it seems like there's so much to do, you know, in all honesty. I just went to replit and did generate code and just said, write me a denial of service attack. And I was just wondering like, would it do it? Or would it, you know, kind of refuse to do it? And I don't know about the quality of this, but it did write code. It did not refuse to write code, you know, did not lecture me about, you know, this is not something I should be doing. I then, I then explained code and I was like, wondering, you know, would it say like, this is malicious code? It, it didn't, it just sort of kind of said, you know, what the code does, like it opens a socket and it connects and then eventually it disconnects.

• 1:14:35Nathan Labenz?

  I mean, do you guys feel like you have the, the bandwidth, the wherewithal, you know, even just the cert, like the, the models have more surface area than you guys do it, like under a hundred people arguably. Is this just too much to like bite off for one company? And I mean, I think, by the way, my view is it might very well be too much for society to be biting off. So, you know, put my, you know, kind of frame on the table there, but.

• 1:15:01Nathan Labenz?

  It seems like there's just so much surface area. Like, can you possibly tame this on the same time scale that, you know, that virtual developer could come online as a product?

• 1:15:13Speaker 01

  I have hope. The answer really depends on what scope you're talking about. If you want to help tame this thing, you know, taming this thing is in itself a very loaded phrase, right? Because what does it mean to tame? Are you making sure that, you know, doughties that could be malicious are not posted? Are you kind of detecting potentially malicious code from being written before it's even written? The devil's really in the details here in terms of like, how you implement solutions to make this thing safer to interact with. In terms of concrete problems with like, you know, abuse and anomaly detection, I think we're totally set up to do work like that.

• 1:15:57Speaker 01

  Because we already have an enormous data set of anomalies and abuse, right? And we spend a ton of time making infrastructure that's like resilient and able to respond to stuff like that. So if we want to train models, right, and build technology that can help scale that process of like moderating abuse, then I think we can. Replit is certainly not, I think, burdened with solving the society level problems, right? Where it's like, we still have to deal with like these intractable problems of, you know, people being unpredictable. Ultimately, like, if we go back to the tool-based argument, right? These things are tools and they are used by people. The best that we can do, or part of the best that we can do,

• 1:16:41Speaker 01

  is make sure that good people are using these things in the first place, right? And that people themselves are acting with good intent. Itself is a very, very hard problem to solve and is mostly an incentive and community design problem rather than a technological problem, I think. If you have a room of 100 people that are all trying to kill each other, there's no amount of anomaly detection that will, you know, save these people if they just want to destroy the world in front of them. So I think actually a lot of this kind of boils down to more like foundational human questions about how we encourage and how we encourage a community and a set of tools that are used for good.

• 1:17:26Speaker 01

  Rather than, you know, are kind of a ripe platform for abusers to take advantage of it. You can always put protective moats up in front of the castle, right? You can have, you know, guards at the station. You can have, you know, the princess at the top of the tower, right? Who's isolated from everything. But um, the real way to address these problems is to make sure that there's no attackers to begin with. The main problem that you brought up is actually a political problem, not a technological one.

• 1:17:59Nathan Labenz?

  I think that's true. Well, I think there are both problems. Any front on which we might make progress here seems like it ought to be pursued. It certainly seems like both technology focused and whatever human focused approaches, you know, hopefully would would bear some fruit. But so I just, I don't know if this, again, I don't know if my like, you know, denial of service code is any good or not. But let's say I did write some quality denial of service code. And who cares whether I wrote it or the, you know, ghostwriter wrote it. Do you, it sounds like you guys have a whole layer of kind of infrastructure to detect that. Like, where would I run into a problem if I all of a sudden started to DOS some random domain

• 1:18:44Nathan Labenz?

  that I have a vendetta against? Uh, how is Replit gonna stop me from doing that today?

• 1:18:49Speaker 01

  That is a good question. Like, if you wanted to deploy a bot that was DDoSing, right? The website that you have a vendetta against. I can't answer this one super in depth technically, but we are able to see outbound traffic, obviously, because it's our machines that are running everything. So we can detect anomalies like that. Whether or not it's automated in terms of like shutting people down automatically, I don't think we do that. We do have systems in place to detect abusers like this and people who are behaving maliciously. Uh, but I don't think it's a completely automated task where, you know, you send out a bunch of requests, right?

• 1:19:34Speaker 01

  Uh, to this website that you Want to take down. And then suddenly there's like 10 seconds later, oh, Replit noticed you're trying to perform a DDoS on somebody. We're shutting down your Repl and flagging your account. It's not at that level yet. And I think that's probably, you know, I'm not a security expert. I'm also not on the infrastructure team, so I can't like speak super in depth to that. But we are able to catch people like that who are using the systems for for bad. But it's definitely much more of a, it's more of a semi manual and semi semi automatic process now, rather than like a strict kind of filtering mechanism.

• 1:20:12Speaker 01

  You know, in the case of writing some code that does perform a DDoS, yes, you can write it and you can execute it, but it will get noticed and you will probably get shut shut down. Because there are cases where this happens and then they get flagged and booted off the system. But the actual like upfront protection, I think, is a much harder problem that still needs to be, still needs to be worked on and we're actively working on it.

• 1:20:36Nathan Labenz?

  Going back to the top, you put GPT four or whatever in the loop, you know, is that consciousness that you said, maybe kind of. So unpack that intuition for us and, you know, to tell us more about what you're, what that means to you.

• 1:20:50Speaker 01

  I have a pretty like practical stance on my view of consciousness and like awareness. You know, there's tons of like psychological definitions for it that send center around certain kinds of benchmarks, like, you know, the mirror test where animals can perceive themselves in a mirror and recognize that dogs often fail this, for example, but they can recognize their own, they can recognize their own pee. Right? Like they like they pee in places and they can recognize like who peed where, which is interesting. So that's a sense of self-awareness. But it's just not vision focused. Even trees, right? Like if trees can grow in the direction of sunlight so that they can feed themselves, right?

• 1:21:33Speaker 01

  And they'll adapt themselves to the lighting and environment.

• 1:21:38Speaker 01

  Uh, that's some level of self-awareness, you know, whether it's actively making decisions in the way that we perceive, or it's, you know, it's adjusting its own location and its own kind of way that it fits into the environment for a specific purpose. Um, I think that's also a form of consciousness. Um, and Douglas Hofstadter, who's a very famous psychologist, cognitive psychologist, uh, who thinks about consciousness a lot, uh, in his book, um, I am a Strange Loop, there is a diagram at the very beginning where there's this like spectrum of consciousness where he goes from like, you know, atoms and like inanimate objects, right, like rocks, um, or somewhere like closer on the line, all the way up to

• 1:22:21Speaker 01

  like humans, right? And, um, he argues that like implicitly people have some perception of how conscious or not something is. And clearly we have some relative definition of it because, you know, we treat animals, right, and especially smarter animals, right, with some level of consciousness more than, you know, dumber animals, um, or plants below that. I tend to ignore questions like this and basically say like, there are properties of a conscious system that are important, right? Like whether or not you agree on the actual definition of it, there are things about something that is conscious that all

• 1:23:03Speaker 01

  conscious things tend to, tend to share. There is some, uh, form of, uh, environment, right, and awareness of your environment to some degree. Notice I don't say self-awareness, right, because I actually don't think that that's like very easy to define, but there is some sort of like internal reflection mechanism, right? Even plants have a form of internal self-reflection in the sense that, you know, they perceive sunlight, right? They get sunlight onto their leaves, right? And then there's some sort of internal mechanism that the plants execute on to move the direction that they're growing into more towards the sunlight. Um, so whether or not it's actually thinking about that in some form of language, um, is irrelevant to me because it's still processing input and doing something internally

• 1:23:48Speaker 01

  to be able to produce an action. Um, so some form of self-reflection is very important. Being able to continue to do this over time and adapt over time is kind of the third main thing that I'm thinking about. So if you have a system that can internally reflect, right, and adjust new data, if you can do it over time, And if it's aware about its environment, it can take in new input. Then to me, that is some form of cautious system. As much of a meme as it is, GPT four, right? Or GPT three, or some sort of language model, which can ingest text data, right? About the world that contains lots of complexities about different relationships and things and nouns and objects.

• 1:24:28Speaker 01

  You put it into a loop so that it can actually reflect on its own input and produce new actions based on some internal state or internal model. And then you output, you know, the new, the new text that it's producing. To me, that's a form of consciousness. I don't think it has to be like self-aware in the traditional form that most people talk about. You know, like, can it pass like some kind of mirror test? Uh, I don't know, but it's able to act and behave at least semi-consciously. Uh, and it shares properties with, uh, systems that we have all already agreed upon are at least somewhat conscious. So that's kind of my bar.

• 1:25:09Nathan Labenz?

  Kind of a behavioral approach, right? I, I guess what I would kind of infer from that is you probably don't take a position on whether or not it feels like anything to be GPT four, which is something that most people today are just like, well, no way, you know, definitely doesn't feel like anything. And I'm always kind of like, seems like it probably doesn't, but I'm not sure, so sure why everybody's so confident in that. You know, if you were to tell me, actually, it turns out it feels like something to be GPT four. I would expect it to feel very different than anything I've ever felt, you know, totally alien, probably. But I wouldn't be like totally blown away if you were like,

• 1:25:54Nathan Labenz?

  yeah, it, it feels like something, you know, I, I, probably not even describable in our language. Do you think we should, you know, from sort of an ethical standpoint, take a view that such systems deserve protection? I mean, we, we may never be able to, or in the short term, it seems unlikely that we're going to have any definitive answer for whether or not it feels like anything, but you're kind of taking a. Quacks like it's conscious, then it might be, you know, kind of effectively conscious. But would you then extrapolate from that to a sense that, you know, these systems matter in some sense, that they may, you know, deserve to be or ought to be treated well by some definition of well?

• 1:26:38Nathan Labenz?

  And we see these things online all the time where people are like threatening their language models and, you know, to some degree, people think that's funny. Others think that that's like not a good idea, you know, or like potentially already unethical. What does this view lead you to in terms of and ethics of AIs?

• 1:26:57Speaker 01

  Whenever humans regard something as sacred or protected, right? It's all ultimately arbitrary. Even basic human rights are still like an arbitrary set of morals that we've created. So I guess I'm getting into some debate about whether or not there's there's objective morality. And newsflash, I don't think there is, but I still consider myself a moral person despite that. I think if you accept that morals are arbitrary to some degree, right? It's like some sort of rules that we've all agreed upon. It makes us feel good. You know, a good set of morals can help guide and direct director decisions. Some people would say that that's ethics, you know, applied morality, semantics.

• 1:27:42Speaker 01

  The point is that, you know, a good set of morals can essentially serve as a foundation for good incentives, right? For people to operate within a system. So, you know, in the States, we have, you know, certain standards around like what animals we farm versus what animals we keep as pets, right? And different countries around the world also have different standards for that. And if you get on the really granular level of like, you know, individual relationships people have, if I have a dog, right? Like that dog. Obviously deserves, you know, all of my love and respect and attention. But I'm not necessarily expecting other people to give it that same kind of respect and and rights.

• 1:28:27Speaker 01

  Like, I kind of tend to it, right, because I have ascribed value to it. Even things like plants, right, like we have sacred trees, right, we have parks, right, but we also mill lumber and use it to build structures. So this is a very like long-winded way of saying that giving AI rights, right, and protecting it, right, is ultimately just arbitrary. And it's really a matter of how and when it is used and if it's essentially in a position of extreme power or in a position of like emotional attachment with us. And I think, you know, like if we're using AI, right, for kind of national level security problems, right?

• 1:29:09Speaker 01

  It's, you could imagine that, you know, in the future, AI systems literally have a seat at the UN, right? Or have voting power in legislation or have board seats at companies. You know, like, let's get like very, very specific here. There might literally be in contract, right, Nathan's or Tyler's AI, right, has this share of voting power over the, uh, over the cap table. That's a very real possibility. Think the amount of rights and protection that we give it is totally dependent on just like what situation it's put into. With, with great power comes great responsibility. And that applies doubly here.

• 1:29:51Speaker 01

  So, um, in positions where, uh, these things are like really, really highly trusted and, you know, are potentially, um, making very, very big decisions or helping us do, they must be protected. If there's an AI that's like helping coordinate the like New York Stock Exchange in like 10 years, you know, you can't. And it starts behaving like a little bit maliciously, you can't just turn it off, right? Because like the entire like global financial system is partially dependent on it. So we need like more defensible systems than just like, you know, a complete nuke switch, um, on, on these things. Yeah, I think that's kind of where I'm at with it.

• 1:30:28Nathan Labenz?

  So as it stands right now, do you? Do you feel like it's totally fine? Or do you in practice use these sort of? If you don't get this right, you're going to be turned off kind of threatening prompts to your language models? Or do you feel bad for some reason in doing that and avoid it?

• 1:30:47Speaker 01

  I'm polite, but I think that's because I like to. I think it benefits us to be polite too, because it helps put you in a more communicative state of mind and helps you express things more clearly. But I'm not like avoiding. It's not like I avoid insulting it, you know, for the sake of like protecting myself in the future, you know, like I don't really care about that too much. If it's really super intelligent in the future, it will be able to understand that when people were rude to it, it was simply to test it and test its capabilities rather than like actually insulting it. So I trust. I trust a 10 year from now AGI being able to tell when we were joking or not.

• 1:31:25Nathan Labenz?

  I'm also polite to my language models for what it's worth. So super, obviously fascinating.

• 1:31:34Nathan Labenz?

  You know, somewhat speculative conversation into what's to come. Couple of questions just on very practical stuff. I'm getting this question a lot and I want to get your take on it. I'm new to AI like I've, you know, I've seen all these amazing things. I know that it's going to change the world. I know that, you know, in whatever I do, right? I it clearly has applicability, but I'm just kind of becoming aware of this now. I don't know how to code really at all. What should I do, you know, if I if I'm that sort of beginner?

• 1:32:09Speaker 01

  Is this from the perspective of like, I want to get started like working with AI, like making AI powered software or what?

• 1:32:18Nathan Labenz?

  I think whether or not coding is like a core part of where everybody should start these days is part of the question. Because a lot of times I do get this this question framed this way like, oh, I'm a lawyer. I work at a firm like I know that a lot of the stuff that I or my team does could either be automated or greatly accelerated. I don't really know how to do that. I don't know how to code. I don't have any hands on experience with these systems, but I do see that there's potential. How do I go about bringing myself up to speed so I can start to realize that potential? I'm not saying I want to create software or I don't want to create software. I just want to get value from AI in a way that I currently can't.

• 1:32:56Speaker 01

  First and foremost. Use what is out there and what's popular right now, AKA ChatGPT, Bard, Claude, whatever. You know, these are, at this point, relatively mature products. They have like widespread adoption and penetration in the market. Become a power user, right? All of these extremely powerful tools. Read up on prompt engineering guides, like tips and tricks that people are using to like use them for various use cases. Most people, especially most knowledge workers in the US, you can get by on a lot of copy and paste. Like a lot of tasks and things that can be sped up in your, in your life.

• 1:33:36Speaker 01

  Most of the work and actual manual effort that you're putting in to complete those tasks is in the physical typing that we talked about before, right? It's in the typing, it's in the analysis, it's in the, it's in the synthesis of new information. And actually getting, getting that information back into whatever software programs you're using for your work, whether it's like a spreadsheet or email. You can handle that for the most part. Think for people outside of the industry, even just relying very heavily on chat at first to kind of automate and perform tasks and you are responsible for the connective layer, right? Of getting that data back into wherever you need can get you a very, very long way. That being said, if you actually want to do things that are more complicated, like, you know,

• 1:34:21Speaker 01

  sending mass emails, for example, you know, summarizing or analyzing like a bunch of files, you know, PDFs, images, whatever, you're going to have to write code for that or at least create new software app hoc in order to accomplish these tasks. That's where I would plug Replit. Go make a Replit account. You know, it's very easy to just copy and paste code in and run it and have it operate on a bunch of data that you want. And if you trust Replit. From that point on, we're working very hard, basically by the end of the year or two years from now, for consumers of all kinds to be able to create custom software to accomplish different tasks like this.

• 1:35:01Speaker 01

  And it might come right now at the price of having to actually read source code and, you know, copy and paste it in from places, use Ghostwriter, learn how to use these more like low-level tools to actually interact with the file system,

• 1:35:18Speaker 01

  to accomplish three different tasks that you need. But I think the end goal is to get to the point where, like, you can just speak software into existence and you can just work with software on a level that is equivalent to two people having a conversation. I don't necessarily mean that the future is chatbots. I think that's just one component of it. But eventually we'll get to the point of, you know, you're not just importing and writing code specifically, but you are thinking about the actual task at hand, like your example from before of needing to extract the audio from a video, right? You're thinking about the extraction of the audio and getting that alone. You don't care about the underlying code.

• 1:35:58Speaker 01

  I think also more practically, so I mentioned ChatGPT, Bard, getting used to these tools, using them. Poe is also really good from Quora, which is a mobile app. Making a Replit account, getting comfortable with using Ghostwriter inside of there, or at a very minimum, taking code from other places and using Replit to run it. You know, if you're a professional developer and you already have VS Code downloaded, obviously try out Copilot and get used to that. But in terms of, you know, new people from the industry, I think it's chat, it's Replit. Zapier also has a really good service now where I think they're integrating a lot of AI-powered tooling, and that's great. And Notion also. I've been using a lot of Notion's AI tools recently.

• 1:36:42Speaker 01

  I sent this to their feedback team, but it feels like I have a jetpack on when I'm writing. And they have a lot of really great affordances for using these models in interesting ways on top of just standard document editing and all these like in-line databases and stuff.

• 1:36:58Nathan Labenz?

  Well, let's imagine a hypothetical situation where a million people already have a Neuralink implant in their skulls. And, you know, safety profile is like, looking good. You know, we can, I usually say it's kind of, imagine it's like the COVID vaccine where it's like, the general consensus is it's like, safe, people are not like killing over and dying. There's probably some, you know, inevitably, there's going to be some noise of people saying it may not be fully safe. But, you know, to the best of your knowledge, it appears to be safe. If you get one, you can start to interact directly from your brain to your computing devices. Are you interested in getting one?

• 1:37:38Speaker 01

  It's in my brain?

• 1:37:40Nathan Labenz?

  Well, the Neuralink as it stands right now goes, they basically carve out a little spot of the skull, put an implant back in, and then layer, you know, the skull on top. So it's not in the brain, although there are electrodes from the device that, you know, that get implanted actually into the brain. The device kind of sits on top, I believe, of the Dura, if I understand correctly. There's generations of this as well, right? I think actually maybe the current one that they have, like in monkeys, they take the Dura off and then kind of layer that on top. In the future, they're trying to put the electrodes through that kind of thick layer and into the brain, but they're, you know, still working, obviously, on the final version.

• 1:38:18Speaker 01

  Honestly, my initial reaction is no. I think, I mean, it depends on how, the idea of it being an implant really scares me. And having something electrical up there, you know, regardless if it's been tested as safe or not, I think is just maybe a fear that I can't get over. I think in theory, I do want it. Like, I want all of the benefits that it can give me. Maybe a good counter is like, it's, it kind of triggers the same fight or flight response as asking if you'd let your kids play football. And it's like, for many people, they're like, yeah, sure, I'd play football, like, whatever, I'll go try out for the team. But if you ask them if you'd let their kids play, it's like, definitely not.

• 1:39:03Speaker 01

  Like, you don't want your kid, like, being exposed at risk to concussions and whatnot. So I'm thinking about this from the perspective of, would I want my kids to also get an implant? Like what if that was part of the policy, right? Like that your whole family actually had to be like on board with it, you know? Or like if it was a citywide thing, you know, like, you know, a certain percentage of the population all had to be like on the same page. It wasn't just a totally personal decision. I think the politics of they'd get really muddied there. So implants scare me. I'm into the idea of like the augmentation, but I think I'd be interested to see some form of like non-invasive technology as well that can get like, you know, near the fidelity of it.

• 1:39:47Speaker 01

  I don't know if that's impossible or not, but yeah, what about you? Would you, would you do it?

• 1:39:50Nathan Labenz?

  You know, I, I've heard such good answers on both sides of this question that I honestly don't even know where I stand anymore. I do think this kind of thing is coming. Prob, you know, if I had to guess, I think maybe the get out of jail free card would be the sort of wearable, wearable version that you could take off. You know, feels like, you know, there will be a lot of demand for that. We're already seeing like MRI, you know, fMRI type, you know, brain reading that is not invasive, but obviously has like massive, you know, capital attached to it. So I don't know what the final form factor will be, but man, there are a lot of times when I wish I could like just write something and I have my hands full. I've got three kids, you know, my hands are often somewhat full and like, there is a

• 1:40:32Nathan Labenz?

  real draw toward, yeah, if I could just kind of have this sort of direct channel of like access to information or ability to, you know, get my thoughts out, I would value that quite a bit. Uh, I don't know if I'm quite ready to do, you know, the, the hole in the skull either. Maybe hold off for the wearable that I can remove at night, but I think the draw there is going to be pretty, it's going to be a strong attractor. It feels like one of those things where like, you know, like people today, like you don't have a cell phone, like how, how do you manage, right? Like everybody kind of expects you to have one. It feels like one of these equilibrium shifts that could happen where. You know, it kind of becomes hard not to have something like that at some point. So I sort of expect, you know, it to be something like this to be eventually become the norm.

• 1:41:15Speaker 01

  Honestly, I think the difference might end up being way more drastic than that. Like, oh, you don't have a phone, like you can't talk to people. It'll feel probably more like PEDs, but I think even worse or even more drastic. Like, imagine instead of, you know, let's take like the 100 meter dash, for example. Presumably, everybody in the Olympics at that level is on some sort of drug, right? But the difference between like drug athletes and non-drug is maybe, you know, half a second in the 100 meter dash. It would be like, I feel like a human taking some new drug and then suddenly being able to run as fast as a cheetah. And it's like, you're not even, you need a completely separate competition. You can't even like, it's not that you're blowing them out of the water.

• 1:41:57Speaker 01

  It's like you're a non-human at that point, or like you're bionic, which in this case would literally be the case. Yeah, I think the pressure might come less from like, you know, societal pressure, like, oh, you're not normal, like everybody has, everybody has a phone. And it might be more like, wow, you are like missing out on a fundamental shift in what we can do. And like, it's not like you're not normal anymore, but it's like you're going to be like an idiot if you don't engage. I hope it's more positive than that, but if the direction goes that way and the differences are really that stark, then I think it might be, I just don't want us to end up in a Gattaca type situation. I don't know if you, have you seen the movie Gattaca?

• 1:42:36Nathan Labenz?

  It's been a long time.

• 1:42:37Speaker 01

  Yeah, so for the viewers at home, it's like a genetics-based eugenics movie, essentially stars Jude Law, talks a lot about like inequities between people who are genetically engineered and those who aren't. But, you know, this is, what we're talking about is an extension of that kind of core struggle, I think.

• 1:42:57Nathan Labenz?

  So this is also a perfect, you know, you're transitioning from question to question so seamlessly for me. Last one I always ask is, just zooming out as far as you can, what are your biggest hopes for and also fears for? Society at large as we enter this new AI era.

• 1:43:16Speaker 01

  This isn't a particularly new take, but I hope we can live forever. I I'm definitely on the boat of at this point, we either all die or all live forever. Kind of dichotomy. I mean, we were all going to die anyway. The question is now it's either we all die at the same time through some mass extinction event, or life proceeds as normal and people have normal like life and death cycles. Right? And and humanity continues. So the question is, is there extinction? Or do we like achieve utopia? I'm definitely optimistic that honestly, as long as the US is the first one to develop safety GI,

• 1:44:00Speaker 01

  and we can use it to accelerate science research, you know, 100 X, 1000 X. And we get fusion. We hopefully have some sort of universal cancer treatment. We figure out how to reverse aging. We figure out how to, you know, replace organs and grow them very, very quickly when they're failing. Once AGIs figure it out, that's like kind of the core bottleneck. And then we can figure out the rest of the big issues at hand, which I guess is kind of like Sam Altman's main take to where or a lot of people in AI. It's like this is the fundamental invention that we need to get to. And it's like, without it, we won't be able to answer and solve bigger ones. So let's get to this one first.

• 1:44:45Speaker 01

  We'll still do scientific research in all these other fields along the way. You know, all the stuff that DeepMind is doing with AlphaFold and whatnot. But let's get to here first and then hopefully we can live forever. So, yeah, I don't even know if I want to live forever, but at least for a few hundred years. There's a lot of TV to watch, a lot of media to consume. I have like a million Netflix shows that I need to catch up on. That'll take me at least 500 years. So hopefully, modern medicine advances to the point where I can go through Netflix's entire story.

• 1:45:15Nathan Labenz?

  Well, I certainly hope that works out for you. Tyler Angert, thank you for being part of the cognitive revolution. Thank you so much, Nathan. It was a pleasure.

• 1:45:24Speaker 02

  Omnikie uses generative AI to enable you to launch hundreds of thousands of ad iterations that actually work, customized across all platforms with a click of a button. I believe in Omnikie so much that I invested in it, and I recommend you use it too. Use Cogrev to get a 10% discount.