E38: The Virtual Biopsy Revolution with Dr. Tanishq Mathew Abraham (Part 2 of 2)

Part 2 of Nathan's conversation with Tanishq Mathew Abraham focuses on Tanishq's work with virtual biopsy technology enabled by deep learning. This unique technology has the potential to address a number of important biomedical challenges; in particular, qOBM could help during live tumor and cancer removal surgeries. Tanishq, a 19-year-old UC Davis grad and one of the youngest people in the world to receive a Ph. D, with a degree in biomedical engineering, is the founder of the Medical AI Research Center (MedARC). In case you haven’t yet listened to part 1, check it out! That episode goes deep on Tanishq’s first published paper: Reconstructions of the Mind’s Eye, which encompasses breakthrough research on reconstructing visual perceptions from fMRI scans into images. 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. RECOMMENDED PODCAST: The HR industry is at a crossroads. What will it take to construct the next generation of incredible businesses – and where can people leaders have the most business impact? Hosts Nolan Church and Kelli Dragovich have been through it all, the highs and the lows – IPOs, layoffs, executive turnover, board meetings, culture changes, and more. With a lineup of industry vets and experts, Nolan and Kelli break down the nitty-gritty details, trade offs, and dynamics of constructing high performing companies. Through unfiltered conversations that can only happen between seasoned practitioners, Kelli and Nolan dive deep into the kind of leadership-level strategy that often happens behind closed doors. Check out the first episode with the architect of Netflix’s culture deck Patty McCord. https://link.chtbl.com/hrheretics The Cognitive Revolution is a part of the Turpentine podcast network. To learn more: Turpentine.co TIMESTAMPS: (00:00) Episode Preview (03:26) Introducing Tanishq’s second paper presenting an AI-enabled biopsy (06:39) Diagnostic applications of this unique slide-free and label-free technology (qOBM) (13:06) Leveraging deep learning, specifically generative adversarial networks (GANs) (14:55) Sponsor: Omneky (15:42) Framework for model (16:22) Challenges of medical data sets (20:11) Challenges of unpaired image to image translation, addressed with a CycleGAN architecture (22:19) What didn’t work (25:44) Breaking down GAN frameworks (34:30) Simplifying data to better with the CycleGAN (36:02) Factors for errors and confusing the model (39:00) Compute and training requirements (41:35) How this technology can scale in the near term and improve patient care (45:53) Tanishq’s relationships with EleutherAI and Stability AI (47:08) The future-looking focus of MedArc and unexplored opportunities (49:00) Developments in Medical AI field and interesting applications (55:11) Early education and AI (58:00) Would Tanishq get a neuralink? (1:00:00) Tanishq’s hopes and fears for AI LINKS: Paper: Label- and slide-free tissue histology using 3D epi-mode quantitative phase imaging and virtual H&E staining https://arxiv.org/pdf/2306.00548.pdf Authors: Tanishq Mathew Abraham, Paloma Casteleiro Costa, Caroline Filan, Zhe Guang, Zhaobin Zhang, Stewart Neill, Jeffrey J. Olson, Richard Levenson, Francisco E. Robles TWITTER: @iScienceLuvr (Tanishq) @MedARC_AI (MedARC) @CogRev_Podcast @labenz (Nathan) @eriktorenberg (Erik) SPONSOR: Thank you Omneky (www.omneky.com) for sponsoring The Cognitive Revolution. 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

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

• 0:00Speaker 00

  And I'd be very excited about medical applications of AI and using AI to improve patient care. I'm excited about using AI for education, being able to learn whatever we want to learn. There may be concerns of AI being used for polarization of society, but I also think that AI could be used for bringing society together and being able to connect with each other.

• 0:21Nathan 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. Hello and welcome back to Tanishq Week on The Cognitive Revolution. Today's episode is part two of my conversation with Tanishq Matthew Abraham. If you missed part one, in which Tanishq and I discuss his FMRI to image project, which uses visual cortex

• 1:01Nathan Labenz?

  activity data to reconstruct images that patients saw during an FMRI scan, I highly recommend checking that out. But today's portion of the conversation is quite self-contained, so you can definitely feel free to choose your own adventure. Today, we're covering the other paper that Tanishq recently published, which introduces an AI-powered technique to support medical diagnosis based on clinical assessment of tissues. In practical terms, in today's world, cancer screens are often invasive, expensive, and slow. To determine whether a particular tissue contains cancerous growths, the usual workflow is to biopsy

• 1:42Nathan Labenz?

  the tissue, prepare thin slices of the tissue, stain the tissue with dyes to improve visibility of key structures, and then examine the stained tissue under a microscope, a process that often takes more than eight hours from end to end.

• 1:59Nathan Labenz?

  The new technique, in contrast, is far faster and cheaper and can even be less invasive. Using 3D image data collected with a new technology called quantitative oblique back illumination microscopy, Tanishq and co-authors were able to use a technique called CycleGAN, which, interestingly, was introduced in 2017, making it a relatively old approach by today's AI standards, to take this entire process down to just one second. A virtual slice of the 3D tissue image is virtually stained and then can be immediately examined, even in the context of ongoing surgery. While this technique is new and will, of course, take time to work its way into general

• 2:43Nathan Labenz?

  practice, the promise of more effective surgeries based on more effective diagnosis with less unnecessary damage is clear. Again, a picture is worth a thousand words, and I definitely encourage you to follow the link in the show notes to get a sense for the raw inputs and outputs before going on to listen to this episode in full. Special thanks to Tanishq for spending three full hours with me for these two episodes, and thanks also to all of you for listening. If you're finding value in the show, I'd very much appreciate a review on Apple Podcasts, Spotify, or the platform of your choice. And with that, I hope you enjoy part two of my conversation with Tanishq Matthew Abraham.

• 3:26Nathan Labenz?

  You've got a second paper, and these two came out within like three or four days of each other. Second one is called "Label and Slide-Free Tissue Histology Using 3D EpiMode Quantitative Phase Imaging and Virtual HNE Staining." That is a mouthful, but in the end, you know, I kind of, there's one diagram in the paper that I think kind of describes the before and after extremely well, and it definitely touches on, or at least alludes to some themes that are kind of outside of your research, but definitely still of interest to, you know, our podcast audience, which is like, how are things going to be done in the future? Like, what is work going to look like? You know, what's going to happen with jobs and all these things? So just the kind of before and after is like... How is it done today?

• 4:11Nathan Labenz?

  You guys have a diagram where it's, okay, it's an eight-hour process for somebody to take a piece of tissue, you know, that is like biopsied out of a patient, perhaps, and do these like real thin slices of it. Now we plate that on a slide. Now we put these chemical reagents on them to dye them colors. That is notably a destructive process where like you really can't undo that. And for a lot of them, like you can't go use that tissue for something else now that it's been kind of stained the one way. It's kind of, that's kind of used up, right? So then you, you know, plate a few of these things and stain them a few different ways. Then ultimately, somebody looks at it under a microscope and looks for different indicators.

• 4:55Nathan Labenz?

  And all of this adds up to, you know, diagnostic. The after, one second, you don't even have to, with the new method, you don't even have to slice the tissue. You instead, and I want to hear a little bit more about this, but you instead use this new 3D imaging technology that can image whole tissue. So in this sense, it's kind of similar to the voxel FMRI thing from before where it's, you know, non-invasive against a certain piece of tissue, doesn't have to destroy it. Then you can pull out these slices from this 3D imaging. You can do, you can apply the model that you've developed, which does a virtual staining of the image to convert it to what it would look like if you actually sliced it and stained it.

• 5:40Nathan Labenz?

  And then a person can look at that, or you could potentially even train, you know, a classifier model, as, you know, certainly people have done prior to this as well, to, you know, automatically process these images. And we're probably going to need one because it sounds like a lot more of this might end up happening if you are virtually doing all these stains instead of, you know, actually slicing and manually staining. Upshot of this is like, at least, I would say probably more like two orders of magnitude time reduction in what it takes to get, you know, a stained image out. And again, you know, that just means like explosion likely of like how much of this sort of imaging work is and diagnostic stuff can be done.

• 6:23Nathan Labenz?

  What did I miss in my high-level summary before we dig into some of the details about how you made it all happen?

• 6:28Speaker 00

  Yeah, I think at a high level, that all sounds pretty accurate to me. The other thing worth noting is like this sort of like eight-hour process you can imagine is very impractical for some applications. So like in a typical sort of diagnostic workflow, like that's accused, but like there are some applications where it doesn't work. So for example, during surgery, which is one of the examples that we are really interested in, where you know, a patient may undergo some sort of tumor removal surgery. So the surgeon is going there, for example, if it's a brain tumor, which is again, we're really interested in brain tumors. So if it's a brain tumor, the surgeon is surgical site within the brain and trying to remove as much of the brain tumor as possible.

• 7:09Speaker 00

  And it's often useful to provide some form of guidance to the surgeon so they know like, okay, this is the tumor or this is not, there's no more tumor here, we're done. A similar form of like, okay, we take some sort of biopsy and then we image, try to image it and tell like, okay, if there is still tumor, then maybe that means there's still tumor remaining in the patient as well. So like there needs to be some sort of similar guidance. And this is again, using a similar sort of process, but you can imagine that eight hours is way too long to be able to do this in a surgery. So there are some alternative processes that are like 30 minutes. So that's like still like kind of slow, but like it's, it's still much better than eight hours.

• 7:51Speaker 00

  But that process also provides really low-quality images and like very hard to interpret slides and images. So it's not ideal either. So like, yeah, people have been developing some of these alternatives that some of them are, you know, widely used in the clinic, but they're still very, yeah, they're still kind of difficult to interpret and hard to use. So yeah, overall, people have been developing these what are known as slide-free microscopy methods that try to, try to speed up that process altogether and don't require all that processing like you talked about and can do the sort of You know, imaging in just a couple of minutes, or in the case of this particular technology, it's not even the, you know, it's like you can even image directly and

• 8:34Speaker 00

  it can just take just like a second to image. So it's like, it, it, like, there are very slide-free microscopy technologies, but this one that has been developed as well is actually a very unique technology as well that is quite promising. And this was actually developed by a lab at Georgia Tech. And so that this paper was in collaboration with their lab at Georgia Tech.

• 8:53Nathan Labenz?

  And that is the QOBM, the Quantitative Oblique Back Illumination Microscopy?

• 8:59Speaker 00

  That's right. That's the, that's the slide-free technology. And it's also label-free. So that, so what that means is it doesn't require the staining. So there are some slide-free technologies that do utilize some form of staining. Yeah, itself, the staining isn't necessarily destructive. The more, most destructive parts are the sort of processing to get down to that single slice. So that's including the sort of formalin fixation, which basically makes it so that these structures are, you know, intact, like they just kind of stay in place. So the structures stay in place, but you know, that's still some form of like, you're basically applying formalin, you know, formaldehyde, which of course can be, you can imagine that can also be destructive. And then there's also this embedding in, in

• 9:41Speaker 00

  paraffin vax, so that allows you to section it into thin slices. So all that sort of processing, that can be damaging to the tissue, and it can take quite some time to be able to do all that processing as well. The staining itself can be a little bit damaging, but isn't necessarily the worst part of it, I would say. You can technically, if you wanted to, you could actually like, remove the stain in some cases, but you know, may not be perfect. But because of that, like, there are some slide-free microscopy methods that don't require all of these sorts of formalin fixation, paraffin vax embedding and all of those steps, but still do require

• 10:22Speaker 00

  Staining. So there are some technologies like that that use some sort of dyes that label for specific structures and things like this. But the unique advantage of technologies like QBM is that it doesn't even require that. And so you can imagine that allows, if that's what why it's called label free, it doesn't require any sort of labels, any sort of staining, any sort of dyes. So it's called a label free technology. And because it's label free, that allows for in vivo applications. Because you can imagine, if you wanted to stain something, you would actually have to remove it. You cannot like, apply your stain. It'd be kind of difficult to do that. You know, and you wouldn't want to, you know, label, you know, functioning tissue in your body. I don't think that's the most ideal thing to be doing. So

• 11:06Speaker 00

  instead, with a with label free technology, that's able to image also in the sort of slide free manner, you could also do in vivo imaging. And that's actually made more practical by the use of handheld probes. So the QBM technology has been miniaturized into a handheld probe as well by the by this lab in Georgia Tech. So then they can use this handheld probe to, for example, image in vivo. So yeah, there's a lot of interesting advantages of this QBM technology compared to even other slide free microscopy technologies that may exist out there. So that's, that's why we really think that QBM is a quite promising and unique

• 11:45Speaker 00

  technology. And then in combination with these sorts of AI tools, it can become a very powerful system for diagnostic and in the case of these sorts of surgical applications as well. So yeah, it's very powerful, I think.

• 12:02Nathan Labenz?

  Yeah, the the vision of being, I mean, everybody has had somebody in their life who's been through one of these, you know, tumor removal surgeries. And then it's like, well, how'd it go, you know? And then, well, the doctor thinks that they got the whole thing, you know? So that's good, right? And it's like, well, how exactly did they determine that? And how confident are they on things? And it sounds like, you know, without jumping to the conclusion that I understand the current state of the art, which I certainly don't, it sounds like You know, there's a lot of room for improvement in terms of being able to take a probe directly into the surgical site and be like zap, okay, cool. Now we got a 3D image of this and we can,

• 12:41Nathan Labenz?

  virtually sliced, virtually stained, virtually classify and see is there any more tumor if it's detected in this process, you could do that, you know, right in line.

• 12:51Speaker 00

  Yep, that's the, that's the hope. Yes, that's the goal.

• 12:55Nathan Labenz?

  So let's talk about then how you did it. Again, a small data, you know, regime here by comparison, certainly to like, you know, what folks are used to hearing about with trillions of tokens. So yeah, tell me, I was reading the paper and I copied down 2,358 QOBM and 1737 H and E tiles. Is that basically like you have that many thousand of the new kind of image and then 1,700 of the like actually stained tissue samples that correspond to it?

• 13:27Speaker 00

  Yeah. So these are actually 512 by 512 pixel tiles. So they're very small tiles that we're working with. And yeah, that's, that's pretty much correct. And they're only actually coming from, you know, a few actual subjects or like in this case, we had, we started out with a rat brain model. So there were maybe a few rats that we were working with. And then again, with the human, we had some human examples as well, some human specimens that we were working with again, just from a few patients. So there are, they're just from a few patients and then we were able to get several images from each of these patients or subjects or whatever. But yeah, even with all that considered, it's still a quite small data set that we are, that we are

• 14:12Speaker 00

  working with here.

• 14:13Nathan Labenz?

  So 1737 H and E tiles, that essentially corresponds to some smaller number of actual tissue samples, sliced, plated, stained, imaged, and then the tiles are just little sub bits of that larger tissue sample. So like a handful of tissue samples, a handful of actual things, and then just, you know, zoomed in 512 by 512 squares from those samples.

• 14:45Speaker 00

  That's correct. Yes.

• 14:47Nathan Labenz?

  So that's like really small data.

• 14:48Speaker 00

  Yes, exactly. Yes. Yeah, it's, it's not much data at all.

• 14:52Speaker 02

  Hey, we'll continue our interview in a moment after a word from our sponsors.

• 14:56Nathan Labenz?

  You know, if you, if you had said to me, go build something like this.

• 15:01Nathan Labenz?

  My need your correction would be like, all right, I need to go find, how am I going to get, you know, a hundred million input-output pairs? And, you know, then I'm like, well, shit, I'm going to be cutting a lot of tissue, you know, to, to get there, right? Probably not feasible. So, or maybe not. Certainly would take a lot more effort on the, you know, actual slicer than you managed to get away with. So, you use a GAN-based approach, which I think folks will probably have some familiarity with. Certainly, I've seen GANs used in the image generation days before the current crop of image generators, you know, that was, that was kind of the way. So, I think folks will maybe have passing familiarity.

• 15:45Nathan Labenz?

  But how did you decide to use that approach? Was it, you know, a direct consequence of not having that much data? Or were there other, you know, factors that went into it? Where did the inspiration for the architecture come from?

• 15:57Speaker 00

  Well, yeah, first of all, going back to the concept of data, data sets, of course, just want to point out, of course, yeah, with the case of these sorts of medical data sets, these biological data sets, it's, of course, very hard to collect a lot of specimens, as you can imagine, you know, to, to be able to, you know, in the case of the rats, that means that if you wanted to collect a lot of rats, you're going to have to sacrifice a lot of rats and you have to then, yeah, if you grow, you have to grow the tumors in the rats and you sacrifice all those rats. It's a very long and involved process. The same with the humans, it's like you have to get, you have to wait for enough humans to have brain tumor surgeries, I guess, and then hope there's enough excess tissue around that can be used for research purposes and, you know, it's, it's, it's going to be very hard to collect a lot of specimens.

• 16:41Speaker 00

  So, it's like, overall, just not feasible to do that unless you have like, like, I would, I guess if you were like going into like a clinical trial or something. So, by then, you should already have like somewhat of an established method anyway. So, I mean, overall, these are the sorts of issues that you have to worry about when applying AI to medical applications, is that you do have to sometimes worry about the lack of enough data. And especially if you're trying to use AI in A medical application that's using novel technology. Like if you had something like you wanted to use AI just for histology. I mean, we can get thousands of H and E slides, like the regular slides that is already being used. That's not hard to get because people are already doing that in the regular process.

• 17:25Speaker 00

  And they're already saving those and storing them. And so if that's something that you're working with and of trying to apply AI for diagnosis or something like that, for example. Then you could do that. But the problem here is that we're also working with a novel technology and we want to apply this novel technology for all these different applications. And trying to bring a novel technology into the clinic is going to be a little bit harder. And it's, and so it's, you know, there's a lot of different challenges around that as well. So these are the different, I guess, considerations that, you know, maybe, maybe people don't necessarily think about when it comes to applying AI to medicine and especially in this case of a novel technology. Then specifically for these slide free microscopy systems, one of the difficult things in terms of doing this

• 18:08Speaker 00

  virtual staining task is the idea is that, okay, these slide free microscopy images that we get don't look like standard H and E, which is the standard images, sort of standard visualization. The H and E represent is the sort of hemotoxin, it stands for hemotoxin and eosin, which are the exact stains that are often used for this tissues, for the tissue slides. And that's what you typically, the doctors are typically used to seeing. And so this cut typically got a kind of this, this purple and pink color sometimes is what they, yeah, it typically has this purple and pink color. That's what the doctors are used to. And so my task is to take those slide free microscopy images where they may not have any color. So in the case of the QBAM, they're like grayscale images.

• 18:52Speaker 00

  Or in some cases, they may have very different colors. So there's various different slide free microscopy technologies. Actually, in my PhD research, I was working with QBAM, but I was also working with a couple other slide free microscopy technologies as well. And trying to do a similar sort of virtual staining task of converting those images to look more like standard H and E so that it can be easier to interpret. So the issue in terms of collecting data for this is when you take your slide free microscopy system and image a piece of tissue, you're imaging like a fresh piece of tissue, for example. This tissue isn't processed in any way because that's the whole point of the methodology is to take a fresh piece of tissue and, and image it. But then in order to get your ground truth, you have to process it.

• 19:37Speaker 00

  So you have to do this sort of formalin fixation, embed it in paraffin, slice it into thin slices, put it on the slide and then stain it with your stains. And then finally, you can look under the microscope, image it with some sort of scanner, whatever you have, and you get a final image. The problem is all of that processing causes significant changes to the tissue. So the tissue that you see at the beginning isn't going to be looking exactly the same as the tissue at the end. You're going to have different changes in these sort of spatial positions, maybe changes in some of the structures, maybe a little bit different. You won't be able to get a pixel match between the original slide-free microscopy image and the final H and E image.

• 20:21Speaker 00

  A typical process would be like you have your input and you have your output and you can just kind of like learn to transform that. But you don't necessarily have the output that corresponds to the exact same input now. They're slightly different. So this is what is known as an unpaired image-to-image translation because you want to translate from the slide-free microscopy image, like this QOBM image, for example. You want to translate from the QOBM image to the H and E, but you don't have exact pairs of the QOBM and the H and E. So this is why it's called unpaired image-to-image translation. So there are a few approaches for this. The most common approach being the cycle GAN approach. So that's the GAN-based approach for doing this translation. There have been some other novel approaches

• 21:06Speaker 00

  that have been developed in the past few years that I've also attempted, but they haven't worked as well. A cycle GAN is actually a very old approach. It actually was developed back in maybe 2016, 2017. It's actually very old. I guess for AI, it's old. Maybe in other fields, it's not as old. And I think AI is one of that unique fields where like if something's six months old, it's already like ancient. So in other fields, that's not the case. It's like in the field of like medicine, for example, oh, you know, something in 2017, that's still pretty new. So like in that field, people think, oh, a cycle GAN is actually quite new technology. But in the context of AI, a cycle GAN is actually quite an old method. But it seems to work really well for these tasks.

• 21:49Speaker 00

  And in addition, I've tried some of the newer techniques. So there have been another, there was another approach that was actually based on contrastive learning, which is kind of funny, you know, connecting to the previous stuff that we talked about. I tried that contrastive learning approach. It didn't work as well. I've also tried diffusion model approaches and hasn't worked as well. So I've actually tried a few different other approaches. And maybe it's possible with those other approaches that you can maybe design appropriate loss functions or maybe change the architecture or something and get them to to work better. But Cyclogann seemed to work very well, like right out of the box and also with limited data. I'm like not hundred percent sure why that's the case, but like that's something I've observed throughout my research is that you can actually train these models with very limited data. They don't seem to overfit.

• 22:34Speaker 00

  They don't seem to have any issues. They seem to generalize fairly decently. Yeah, as long as you're passing in high quality data into the models, they seem to to be that you can train them quite well and they work well for these workflow staying tasks. So, I mean, that's the kind of the one of the first models I've tried. Most of the models I've tried afterwards seem to not be as good as this baseline. And so that's why I've kind of stuck to this Cyclogann approach. So sometimes, yeah, sometimes the the newest things are not always the best for at least the specific tasks that you're working with. I think there are some different properties of histology images that make them unique.

• 23:14Speaker 00

  And some of these unique, these newer methods make different assumptions, implicit assumptions that many people don't realize that they do. And those sorts of assumptions don't work well. They don't correspond well to histology images. So some of these newer methods don't seem to work well from what I've observed. But within our case, the Cyclogann seems to work really well, which, you know, yeah, I'm still kind of surprised that this really old technology, I mean, we thought that, oh, you know, we, and I thought that, oh, we'd be working with diffusion models already. But like, I've tried diffusion models and they haven't worked as well either. And it's like, kind of surprising to me. Sometimes, yeah, it's worth even exploring some of these older. Uh techniques and seeing if they still may be applicable and maybe even better in some cases.

• 23:58Speaker 00

  Um, and that seems to be the case here. So, yeah.

• 24:00Nathan Labenz?

  Yeah. So how does it help you get over the lack of unpaired images? Cause at the end of this, right, your, your kind of end state is we now have a technology where we can use this new kind of microscopy, you know, even that can be, you know, used during surgery on a live tissue. It does like a 3D kind of imaging, then you can take slices out of that 3D space as 2D images. And then you can kind of apply this mask, so to speak. You know, it's almost like if I went to, um, you know, Playground AI and said, make this an H and E image, except, you know, with a lot more rigor and actual, you know, usefulness behind it.

• 24:44Nathan Labenz?

  But that's kind of the, you know, you're sort of painting it over as if it had been stained. But I'm missing the, uh, the trick that the cycle GAN architecture performs that allows you to do that without having the paired images in the first place.

• 24:58Speaker 00

  Okay. So maybe it's worth talking about, uh, GANs in general, because like you said, maybe it's a kind of an old technology at this point. Maybe some people are less familiar with it nowadays. But basically a GAN is this sort of, um, dual neural network, uh, framework. So you have a, a generator neural network and you have a discriminator neural network. The generator neural network is trying to produce, trying to generate your images, whatever data that you have. It's trying to generate images that look like they're, they came from that data set. And then you have a discriminator neural network that tries to, uh, determine which images are real and which images are, are generated. So it's being passed some random image. That random image could be from the original data set, or it could be an image that was generated by the generator neural network.

• 25:43Speaker 00

  And the discriminator is going to try to figure out if this is a fake image or a real image. So basically what you're trying to do is you're actually trying to train the generator to generate images that fool this discriminator. So. You're training your generator to just try to make images that are more and more realistic such that it fools the discriminator. And the discriminator is trying to tell, it's being trained to try to improve, it's trained to try to tell the difference between the generated image and the original image. So this is sort of like back and forth that's going between the two models. The generator is trying to fool the discriminator and the discriminator is trying to determine which one is generated. And it's just kind of going back and forth. And the idea is that eventually your generator is going to keep, it's going to, through

• 26:27Speaker 00

  this process, it's going to improve and improve and improve and finally give you really realistic images that look like they came from the original dataset. So that's just the general idea of GAN. And this was used very frequently for all kinds of applications. It was kind of like, I'd say, the first primitive form of generative AI, I would say. I mean, these were being used for things like generating faces and generating, we had all these sorts of different GAN architectures, like style GAN was very common at the time. So this was, this was being used quite commonly, I guess, before 2020 or like, yeah, 20, up to 2020 or so. So the additional thing that the cycle GAN itself provides, in addition to this general

• 27:11Speaker 00

  GAN framework, is that the cycle GAN adds, it actually has two generators and two discriminators. So the idea is that you have one generator that is taking in the original image. So in our case, that is a QOBM image, it's taking the QOBM image and it's trying to produce the HNE image. So it's going to produce some sort of fake HNE image. And we also have a discriminator that is going to try to figure out if it's a real HNE image or a fake HNE image. It just sees the HNE image from the generator or it sees some real HNE image. So it's trying to learn how to classify between real and fake.

• 27:56Speaker 00

  That signal is going to the generator and the generator is trying to improve itself and produce more real HNE images. So it's able to produce really nice Realistic HD images. But that doesn't mean that it has to produce HD images that correspond to the original QOBM image. It can produce any HD image. It doesn't matter. As long as it's real, it will fool the discriminator. As long as it looks real, it will fool the discriminator. So that's why we have an additional generator that takes in the HD image produced by the first generator, and it produces a QOBM image. And again, we can have another discriminator that can take in that generated QOBM image

• 28:37Speaker 00

  that is classifying and makes that second generator, it's able to now produce really realistic QOBM images. But what's unique now is that we can compare that last QOBM image to the first original QOBM image. And basically, you want those QOBM images to be the same. You have your starting input QOBM image. It's going to that first generator to produce an HD image. Then it's going to the second generator to produce a reconstructed QOBM image. And you can map those two together, do some sort of like MSC loss or some sort of direct comparison of these two images now. And you can say, okay, those have to be as close as possible to each other. They should be exactly the same.

• 29:17Speaker 00

  And the idea now is that in order to be able to reconstruct that QOBM image at the very end, the content has to be maintained throughout the entire process. The content has to be maintained throughout the entire process. So that means the QOBM image is being converted to an intermediate HD image. And that HD image should hopefully now maintain the content of the original QOBM image. So now you're able to get a realistic HD image that maintains the content of the original QOBM image. And that's what you actually want. And so that's the general idea of CycleGAN. It's not perfect. Like, you're just kind of hoping that it does maintain the content. But it seems to work most of the case. Like, you know, I guess it turns like, there are some cases where people see like, oh, it doesn't actually maintain the content.

• 30:01Speaker 00

  And it actually, there are some cases where it doesn't work perfectly. I won't go too much into that. But in most cases, if you also, if you have really high quality data that you're working with, a good dataset that you're working with, it seems to work quite well. And again, that, so that's, yeah, that's basically in a nutshell, how CycleGAN works. It's this idea of having the cycle where you're going from QOBM to HD back to QOBM. And now you have something that you can compare directly because it's comparing basically to itself. So it's this sort of idea, it's called cycle consistency. So that's the, that's why the CycleGAN actually stands for cycle consistent. Generative adversarial networks. So that's why it's called Cyclogan. So that's the general idea of a Cyclogan. And it's surprisingly powerful, even though I'd say, you know, it's not the most sophisticated architecture, but it's, it works. So, yeah.

• 30:46Nathan Labenz?

  So you've ultimately got four models, right?

• 30:50Speaker 00

  Yes, but at the end, you only need that one first generator that's used for inference. So it's actually, you do have like basically three models that you actually don't necessarily use anymore afterwards. So, you know, it feels like, okay, maybe there's like, yeah, it's like, maybe you're using extra computation. But again, I've tried some of these other methods that seem to be simpler, only use a single model, but they don't seem to work as well. It seems to work, but like, yeah, like it's not ideal. You do have these four models that you're training with. And then at the end, you just use that first generator for your desired virtual staining task.

• 31:23Nathan Labenz?

  Still in there, there's not a conceptual guarantee that virtual stain will have like exactly the same structure, whatever. The thing that is keeping it on the rails is the fact that you're then enforcing that it must be able to be converted back. So if it was like too divergent, you'd lose that ability to convert back.

• 31:43Speaker 00

  The issue can be sometimes like, so for example, I don't, it's mentioned in this paper a bit, and it's actually discussed in another paper, a workshop where I wrote for another slide free microscopy virtual staining, is that sometimes, basically the idea is you want to make your task as easy as possible for the Cyclogan in order for it to be able to do this. Like you want, yeah, you want it to be as easy as possible. So like the easiest possible solution is basically to maintain the content. That's what is just kind of naturally, you want it to be the easiest thing to do. Because what we've seen are some examples where, again, in the case of the QOBM and for some of the other slide free microscopy images that I've worked with, you can have

• 32:24Speaker 00

  sometimes in the HNE, the nuclei are dark because they're, they're labeled with this absorbent dye known as hemotoxine. And the nuclei, the cell nuclei kind of show up as this dark purple color. But the problem is that in QOBM, they're kind of bright in the, in the QOBM. The cell nuclei are kind of bright, especially mostly for like the tumor cell nuclei, actually. They're like kind of pretty bright. So this can cause issues because the Cyclogan is confused that in the HNE, the nuclei are dark, but in the QOBM, the nuclei are bright. So it may make some sort of like.

• 33:03Speaker 00

  HNE image, but that HNE image has the dark regions in the QOBM as nuclei, but they're actually not nuclei. They're just like other parts of like they're like maybe the cytoplasm of the cell or some other region of the cell because the nuclei in the QOM are actually bright, but it's able to do that whole cycle consistency because you know it can still know how that maps then it has its own way of mapping that to a regular QOBM image that incorrect HNE. It's probably better to see this visually and we do have there's a figure in the paper in the supplementary material where you can actually see that really visually. It's harder really it's harder to describe this sort of thing, but there are some cases like yeah, because like there's this difference between the QOBM images and the HNE images. So you have these issues.

• 33:45Speaker 00

  So the solution that we take is we simply invert the QOBM image. And so now if the invert in the inverted image, the nuclei are now dark. And so now the psychogans like, okay, there are dark nuclei in the QOBM, there are dark nuclei in the HNE. The simplest possible thing to do in order to make HNE images while keeping the cycle consistency and the you know consistency going on is just to take those dark nuclei in the QOBM and convert them to dark nuclei in the HNE. So you've got to make it as simple as possible for the psychogan. So that that was kind of one of the key insights that we had when working with the psychogan. It's like it's a bit of like kind of like trying to understand what's going on, the sort of psychology of these sorts of models in a way. It's like, what

• 34:30Speaker 00

  are the things to get this to work? So one of the things we noticed is trying to make it as simple as possible for the model. And you know, there's different things like this that we we have to play around with in order to work with the with the with these models. So actually, a lot of the work that we did was on the data side, you know, data processing, data cleaning. And it turns out there's a lot you can actually do on the data side that can really improve the quality of the models. So I think this has a lot to do with what is known as data centric AI where it's like you're focusing on the data development rather than the model development and trying better and better models, but you try to improve your data. And I'm a I'm a huge believer of that sort of practice in terms of like, yeah, the data is the one of the most important things you should be

• 35:11Speaker 00

  focusing on. You can actually get like really basic models to work really well if you've got really good data sets. And so that's and knowing how those data sets are processed by the models and how the models work with those data sets and having all of that and understanding that can help you to do to do to to to get good results with even really basic models.

• 35:32Nathan Labenz?

  I was interested to see that to the degree that there are any points of confusion, it seemed to be that the system was flagging things as potential tumors even when they weren't. It did not miss any of the actual tumors.

• 35:48Speaker 00

  Yeah, it wasn't something that we explicitly designed for. So again, I think this has to do with sly free microscopy being different from regular HD images and how to how to get an exact match. So we went to the errors, the ones that it failed with, and we noticed a few kind of common factors. One of the factors was like the this cap, these sorts of like blood vessels, you have these small blood vessels, you know, these sorts of capillaries, as they're called. In HNE images, you don't get continuous capillaries because you have these sort of very thin slices. It's hard to get like a capillary like going up and down or something. You may just get like a, you know, if you're going, like, think about a 3D, if it's going like a little bit up and down, and you just like

• 36:31Speaker 00

  cut a very thin slice, you're not going to get the full, you're not going to see the full capillary, right? So you just get like a very thin slice, you're going to get very small segments that you will see. So we don't see like full continuous capillaries. That was one thing that we do see, though, in the Q of BM, because it's an intact tissue, it's not necessarily sliced, you can kind of see more kind of the full capillaries. And not only that, like during, again, some of the processing with the HNE, some of these red blood cells in the in the in these blood vessels may fall out and all sorts of issues. So yeah, basically, you're not going to see some of these structures are intact in the original. QOBM images but not intact in the HNE.

• 37:15Speaker 00

  So sometimes that confuses the psychogan. It doesn't know what to do with those sorts of structures. Sometimes it's able to deal with it, sometimes not. And in this case, we had those capillaries in those sorts of images. Another thing is that we have these sorts of what I call white matter bundles, which are specific structures again in the, in the, in the brain. And again, they look quite different at the QOBM compared to standard HNE. And so because they look different, maybe in a more intact manner, that this is maybe actually how they are supposed to look like. But then in the HNE with some of the processing and some of these extra steps and the staining, it looks kind of different. Again, the psychogan is getting a little bit confused about the differences between the QOBM and the HNE. So those are some of the kind of features that we see that happen.

• 38:00Speaker 00

  And because of these differences between QOBM and HNE, it's, it's, it can be a little bit hard sometimes to map them. And, and so sometimes occasionally the psychogan will get a little bit confused about this. So it's not necessarily ideal, but you know, it's, it's a very difficult problem. And you have things that, you know, look different like this and it's hard to get an exact match. So, you know, it is, it is what it is. But, you know, it's possible that with, you know, collecting more data and different data sets, it may be possible to improve this further. And ideally you would want no false positives and no false negatives. So, you know, there's still room for improvement, obviously. Especially if you're working in some, something like a, you know, with brain tumor. You, you don't, you also like, don't want false positives

• 38:43Speaker 00

  because that means that you'd be removing perfectly healthy functioning brain tissue. And, you know, that's also a, you know, a very difficult and, you know, something you'd like to avoid potentially. So, you know, there's still room to improve. But, you know, it's still, still quite good right now. But, you know, in the future, there may be better approaches, maybe larger data sets or something like this that may be able to help solve some of these issues.

• 39:08Nathan Labenz?

  What does the training process on this look like? How compute intensive is this?

• 39:12Speaker 00

  Yeah, it, it, it's also trained on a single A100. Each of the models are just trained on a single A100. And it takes just a, it's just a few hours. It's not, it's not particularly long again. So yeah, I can just put in experiment for running and after two or three hours I can come back and see. So yeah, it's again partly because the data sets are not very large. So you know, also it's like, yeah, it takes quite, takes pretty quickly to go through the data sets. Um, but also, yeah, I mean, I tried doing these experiments though earlier, even like a V one hundred and things like that. And of course, it used to take much longer back then. So also, you know, part of, part of it is also like now, you know, the, the GPUs are getting better too. So it's much faster too. Um, but yeah, it's, it's pretty, it's, it's pretty fast and it's, it's doable on, on, you know, pretty

• 39:57Speaker 00

  simple, yeah, pretty, pretty limited hardware compared to, of course, a lot of these larger trading runs.

• 40:04Nathan Labenz?

  What do you think is kind of the path to this sort of thing being actually used in a clinical setting? And like, what are the bottlenecks?

• 40:11Speaker 00

  So of course there are certain like potential model developments that could be made to the, the virtual staining. So you can of course try to improve that further. But also it's very important to further validate this technology. Um, we only train, we only played around with, you know, a few human specimens. So, um, you know, for example,

• 40:33Speaker 00

  we only got like one, a single human healthy image, which you can imagine you wouldn't want to be like, there's obviously going to be very little human healthy images available. Uh, because, um, you know, um, you wouldn't be wanting to take out a healthy tissue from a patient anyway. So we, there was, there weren't many like human healthy images that we could play around with. So we, we weren't able to validate how well the model works with, like if it saw a healthy image, how well it would do. Uh, it, for the one single healthy image that we did use, it actually worked perfectly fine. Even though the model was trained mostly on tumor images, that was actually kind of interesting and kind of surprising. It's like, okay, it kind of generalizes to this healthy image, which I thought was kind of neat. But you know, again, those are examples of things that need to be validated more carefully.

• 41:18Speaker 00

  Again, we used only a subset of types of tumors. So we use these sorts of great. Grade two and three astrocytomas, but there are various other types of brain tumors that would would undergo the same sort of tumor removal surgery and would benefit from the same technology, but we didn't get a chance to image and utilize and train our models on those sorts of images and and validate our technology for those sorts of tumors. So, you know, just trying to make it more broad for other types of of brain tumors, for example. So there's a lot there that just kind of scaling this up, basically, getting more specimens, getting more images, training them on these different images and specimens, validating them. And, you know, just trying to do some sort of like clinical

• 42:02Speaker 00

  trial at the end. That's going to obviously be necessary if you wanted to actually get this into the clinic. So you would have to. Yeah, there's there's a lot of kind of scaling up of this research that would need to be need to be done before you can, you know, properly validate it and have confidence that this is something that would be safe to use and accurate and, you know, for for clinical applications. And then in terms of, yeah, in terms of the actual hardware, you know, they're, you know, it's still kind of in the research stage. There is a handheld probe that is has been developed, but, you know, there there may need to be more work in terms of just ensuring that it's reliable and things like this. And and then I guess once that's developed, again, it probably

• 42:44Speaker 00

  have to go through its own form of clinical validation. And the medical side of things, it can be quite intensive in terms of the sorts of testing that needs to be done in order to validate the technologies, which of course is important because, you know, you're using this in, you know, in life and death scenarios. So, you know, you want to make sure it's it's working properly. So, you know, it it would take a while for it to be able to be finally utilized in the clinic. But, you know, the the eventual promise is definitely very, very appealing. And, you know, the eventual benefits that it could have in terms of really speeding up the workflow and and improving the patient care is it's really, really exciting. So, yeah.

• 43:28Speaker 00

  One of the main takeaways for me during my entire PhD has been the importance of of data, of high quality data and the focus of of data centric AI. And that's something that I think people haven't been focusing on. As much, I think we're now starting to see that become a focus these days, even in the context of trading large language models or some of these other fields. I know that, you know, there have been some recent papers that have talked about the importance of data filtering and large for, you know, trading large language models and things like this. So I'm sorry, I'm really glad to see that sort of focus of high quality data. And that's something that, you know, I've, I've learned from my time as a PhD student

• 44:09Speaker 00

  and something that I will continue to focus on throughout any of these sorts of ML projects that I will work on. Yeah, I think that's kind of the, the key takeaway for me. Having the, the most advanced models is not going to do you any good if you don't have high quality data. My, my P I, my, my advisor liked, liked to say garbage in, garbage out. So, you know, I mean, that's a common saying that people, people have. So that's, I think, a very important point to make, I feel.

• 44:36Nathan Labenz?

  You have these affiliations with Luther and with Stability. Anything you would just tell us about those organizations that may not be obvious from kind of a, a public viewpoint?

• 44:46Speaker 00

  Yeah, I mean, I, I, my affiliation with Luther is just being kind of part of the general research community. And I think it's definitely a very, very, again, a great research community in terms of like, again, focusing on that sort of open collaborative nature. And I think Medarc took a lot of inspiration from communities like Luther AI, some of the other communities, Lion, ML Collective. There's so many, nowadays there are lots of great open source research communities and all those have, you know, really inspired Medarc. And now I think we're seeing that this will, it seems to be like this is a really promising approach for medical AI research as well.

• 45:29Speaker 00

  So, yeah, and I'm just, yeah, really happy to be part of some of these communities and be involved. And, yeah, that, that kind of is the extent of, I guess, my Luther AI affiliation. And then in terms of Stability AI, Stability AI, I'm a, I'm a part-time employee of Stability. And they are basically supporting my work at Medarc and, and the research that I'm doing there. And, yeah, you know, they've been really a great. Great support for the community and yeah, of providing compute to run all of these experiments and to support this research. It's been, you know, that's been really appreciated. Yeah, they've also supported my PhD as well. So like they provided a fellowship for my PhD

• 46:14Speaker 00

  as well. So they were supporting my PhD research as well. So yeah, but yeah, this lots of interesting research has been going on at Stability. And so I'm really excited that I've been part of that. And yeah, I'm excited, of course, continue the research at MetaArc in these directions. I think why Stability is interested in these sorts of, in this area is to, of course, see how these sorts of advances in generative AI that Stability AI is making and other, of course, other companies and other research institutes are making, how these can be applied to medicine. And that's kind of the focus of MetaArc is, and the focus of my research is to see how can generative AI be applied to medical applications. Because, yeah, I

• 46:57Speaker 00

  don't think many people realize that that's something that, you know, even worth looking into. It's like, you know, you think like, oh, it's like AI art. How is AI art going to be any useful in medicine? It's like, what's the point of that? But it turns out, you know, these sorts of models can be used for all kinds of interesting applications in medicine. And similarly with language models, it's like, I guess that may be a little bit more, kind of a little bit more obvious why that may be useful for medicine. But, you know, it's still lots of unexplored opportunities there. So that's what kind of what we're focusing on at MetaArc and really glad that Stability is supporting me to be able to do that as well and supporting the research at MetaArc.

• 47:32Nathan Labenz?

  You're not returning anything in an exclusive way to Stability, it sounds like?

• 47:36Speaker 00

  Yeah, that's correct. There isn't anything necessarily exclusive to Stability. And you're just willing to provide their, the compute resources. And of course, I mean, they're providing this not just to MetaArc, but to other, to other academic research projects and, yeah, all kinds of research projects. But I think also being part of Stability also enables us to kind of Take advantage of some of the the research that Stability is able is actually doing. So there are some like projects that we're working on that is able to take advantage of the stuff that Stability is working on, collaborating with Stability AI researchers. So that is also enabled by being part of Stability. But yeah, apart from that, I mean, it's just kind of, it's just a kind of a research organization.

• 48:20Speaker 00

  There's nothing necessarily that Stability is is gaining out of this. And again, this is something that Stability has been doing for various different academic research projects. It's possible, of course, that, you know, some of the developments that we see in the medical AI field may help out for other potential applications. Have you seen this happen in the past? For example, for example, you know, the original UNET architecture was developed for cell segmentation, actually. It was, that was the original application, and now we're seeing it being used for all kinds of segmentation, but also image, any sort of image to image thing, any sort of even for diffusion, it's used in the diffusion models all the time.

• 49:05Speaker 00

  So, and then another great example is the CLIP model. So actually, the CLIP model was originally developed by Stanford for medical AI applications. But it was then scaled up by OpenAI to for CLIP. So CLIP was actually, it was, I think, called, it was, the original paper was called CONVERT or something like this. It was actually developed for looking at kind of image text representations of radiology reports and this sort of, these sorts of data sets. And then I think some researchers at OpenAI discovered that and decided to kind of scale it up, and that's how we got CLIP. So I think, you know, there's lots of like, I think there's lots of interesting opportunities like in the medical AI space because

• 49:50Speaker 00

  The research that comes out of that space could be utilized for other applications too. So I think it's a kind of that, again, going both ways in terms of like the research in general of AI may be useful for medical AI, but it's also possible that the research in medical AI may be useful for more general AI applications as well. So that's the sort of kind of feedback that's going on there that I think is really exciting to potentially explore further. And so that's why I think that maybe some of these organizations like Stability have picked up on that and are willing to support some of this some of this research as well.

• 50:21Nathan Labenz?

  What do you think is kind of the future of radiology work? Is the stuff that you're developing going to change that in a fundamental way? Is it all, you know, just a tool?

• 50:34Speaker 00

  In the short time, it's definitely seems like it's just going to be a tool for various applications. Like I can, I mean, there's so many ways that AI can help in medical applications and even specifically in radiology. You know, there are various questions of like how reliable AI systems are currently. And I think, you know, I mean, we see this, of course, with systems like chat GPT and these sorts of large language models that, you know, are known for hallucinating and and, you know, they have these issues, but they're still extremely useful if you use them as a starting point. And so I think the question is if we can make these systems more reliable. And right now it's unclear if that is, you know, what is, what is the status of that? The sort of current systems, I can see these sort of current systems being used for various

• 51:19Speaker 00

  applications that still require the doctors to be in the loop. So the sort of human in the loop process. I think that's still going to be extremely valuable. I mean, there've been like the sort of general applications include just like, you know, maybe having AI systems provide some sort of diagnosis that the doctors can can see that this is a potential diagnosis that the AI provided, for example. And I've seen some papers that have described, you know, the sort of combined AI doctor system as actually outperforming even the, even the AIs themselves. So this is something that's possible because, you know, the AIs may pick up on certain things and the doctors, based on their years of experience, may pick up on other things as well. And then in addition, it's, I think the AIs could also be, you know, they're very

• 52:03Speaker 00

  Applications for AI to be used in clinical education settings as well. So even using AI to better educate doctors as well is a very interesting scenario that I think will be really promising in terms of, for example, one of the interesting things that we worked on recently was the sort of fine tuning stable diffusion on chest X-rays. So using generative AI tools to generate medical images. And you can use that to, for example, train doctors on certain images or something like this. Maybe there are some images that the doctors are less familiar with. They can utilize these tools to generate some images and use that to help train them better. So there are various applications like that.

• 52:46Speaker 00

  One application I'm also very interested in how AI can be used would be actually discovering things that doctors didn't know were originally there in images, for example. So that's something that I think is really exciting is that there are actually a lot of interesting biomarkers that are hard for people to actually pick up on, but that information is actually there in the images. So I'm coming from this, like I was working a lot in pathology and my PhD research is microscopy and pathology, applying AI to these applications. So one of the interesting applications that I read about, that there's a lot of research in, is actually picking up on molecular biomarkers from HNE images.

• 53:27Speaker 00

  So like, you know, these sorts of molecular signals, like maybe if cells are presenting specific receptors, specific proteins or things like this. You wouldn't think that that kind of information would be available from an image. You think that you have to actually go and do those molecular tests in order to get the information. But the cells show certain sort of different morphology and different shapes and sizes, very minute differences, but those differences are still there that AI systems can pick up on that. So being able to use AI systems to actually pick up on information that regular humans cannot, that I think is also a very interesting application. And again, that's something that the AI systems can provide that information to doctors

• 54:12Speaker 00

  and they can use that for clinical decision making. So overall, there's lots of interesting applications, but again, The question comes down to how reliable are these systems? They can be used to supplement what's already going on and and and kind of maybe alleviate the burden of existing doctors. But if the systems are not reliable enough, I don't see them fully replacing the doctors. But it depends on how this research goes and we'll see how how how the field goes. And it's very hard to predict the developments. I mean, I don't think any of us would have been able to predict where we would be now five years ago. So it's very hard to predict these developments, but based on what I've seen right now, this is where I see things going in the future.

• 54:53Nathan Labenz?

  But I wonder if there was any sort of interesting elements to your early childhood education that maybe were like not scalable or not very repeatable that you think AI could start to make more broadly available to more kids?

• 55:06Speaker 00

  I mean, I think part of it is just like, yeah, a lot of why I was able to do what I was able to do is of course the availability of resources and online. I think there's a lot of great online resources. Like, I don't think this would have been possible 20, 30 years ago. And I think also, yeah, it's part of it is like people learn at different rates. And I think AI, that's something that AI could definitely help with is because, yeah, like going back to what's scalable and what's not scalable is that there's this sort of assumption that everyone's learning at the same rate. That's kind of the implicit assumption of most education systems. And this is something that, you know, has caused a lot of problem for me was the is

• 55:47Speaker 00

  the idea of like, you know, I'm I I learn at a faster rate compared to most people. And then, of course, there are people who learn maybe at a slower rate than other people. So it's like there are people learning at different rates and giving the same homogenous education for everyone doesn't make much sense. So I think that's something where AI could help with is trying to provide more personalized education for people, providing this sort of personalized education programs that meets them at the level that they are at and allows them to progress at the rate that they are comfortable with, whether or not it's a it's a slow rate or at a fast rate. It doesn't it shouldn't matter.

• 56:26Speaker 00

  And and so that's kind of, you know, in my case, you know, I had to do a lot of, I guess, trying to get through the system to be able to to be able to go at the rate that I was. Interested in going at and was comfortable with going at, you know, so that I wasn't bored out of my mind, I guess. And so, you know, that was, you know, I think that there's definitely some interesting potential for AI to help with that. But I also think a lot of it has to do with how people are, how much people are willing to implement those sorts of systems. It feels like, you know, that's again, kind of a kind of complete overhaul of the education system. It's like this idea of like, oh, personalized education.

• 57:09Speaker 00

  You mean, we're not going to like, just have regular classrooms or how does that work? Or, you know, there's lots of these sorts of things that, you know, I think there's also that kind of societal aspect that is definitely hard to solve. So there's a lot of like, again, lots of interesting AI applications. And I mean, this is true for any field. It's like, there could be lots of interesting AI applications, but the sort of societal reaction and how society implements it, that's itself kind of a different question. And, and it's, and, you know, something that's also worth thinking about and worth trying to solve as well, so.

• 57:41Nathan Labenz?

  Neuralink just got initial approval for a FDA approval for a human trial. Let's imagine a future state where they've gone through all that. They're approved and a million people have a Neuralink device implanted. Would you be interested in getting one at that point?

• 58:03Speaker 00

  It's a good question. In theory, that sounds like I would love to get one, but I would be more concerned about the security issues. So I guess it would, assuming, you know, security is not an issue, I would definitely get one. But I think it has a lot to do with, I don't want people hacking my brain. So, you know, that's definitely an important concern, but like, yeah.

• 58:29Speaker 00

  I guess if as part of a hypothetical scenario, like if that's not a concern, security, like I would love the idea of being able to like connect to, you know, have these sorts of branding computer interfaces, being able to, um, you know, augment myself in that way. That sounds, that sounds amazing. But so, you know, I would definitely do that if that was in this sort of hypothetical scenario. So yes, but of course, in reality, it's a little bit more messy than that. So, you know, you know, we'll see how this technology progresses. But, um, yeah, I mean, it's still very interesting and I look forward to seeing how all that kind of progresses and.

• 59:04Nathan Labenz?

  What are your big hopes for and fears for society at large as AI kind of permeates society at large and begins to touch everything?

• 59:16Speaker 00

  Yeah, I guess let's start with fears. Um, with in terms of fears, of course, um, I think my fear in terms of AI in society is just like maybe how, how much polarized our society may become because of, of AI. That seems to be a big concern. The concern of, of biases as well being made worse because of, of AI technology. Um, there's a lot of concerns like this and, you know, just overall concerns of like the reliability of, of AI tools and, you know, what kinds of dangerous circumstances that could, could lead to. In terms of what I'm excited about AI, I think I'm really excited. Well, of course,

• 59:59Speaker 00

  as you can imagine, I'd be very excited about is the medical applications of AI and using AI to improve patient care. I'm really excited about all kinds of applications there. I, and of course I'm excited about using AI for, for education, being able to learn whatever you've, we want to learn. I think AI is going to be extremely valuable for, for that. Yeah. I mean, I think AI is, you know, it's, you know, AI can be used in various different ways. And I think, you know, there may be concerns of AI being used for polarization of society, but I also think that AI could be used for bringing society together and being able to connect with each other. You know, I mean, there's so many ways that it could be able to, I mean, it's happening even now with, you know, very primitive forms of AI, you know, things like.

• 1:00:44Speaker 00

  You know, we have, you know, social networks. We have things like, for example, translation that's able to connect us with people from other cultures. There's so many forms of, of, you know, even primitive AI tools that are being used nowadays that are helping us to connect with other people. So I'm hoping that that kind of trend continues even with the current AI tools. And, um, and I'm also really interested in how AI will help enable people's creativity. And I think we're starting to see this already with some of these, uh, you know, AI art tools. And so I think there's lots of interesting creative applications of AI and just being able to bring ideas that people have into the real world with AI. I think that's really exciting. Of course, maybe that will soon

• 1:01:27Speaker 00

  happen with, you know, mind reading and things like this that will then immediately take it from, you know, the, the idea stage to something that's, you know, physical, something that's happening right in front of you. But yeah, I'm really excited about just how AI can, you know, enable people to bring their ideas and, um, you know, actually augment their creativity. So yeah, I'm, I, I think there's a lot of really interesting applications. But yeah, it's always a, you know, it can go, go both ways. And, you know, it's a, you know, very nuanced, nuanced topic in terms of the, both the benefits and the opportunities of AI as well as their risks and concerns. So.

• 1:02:08Nathan Labenz?

  Tanishq Matthew Abraham, happy 20th birthday in just a few days. And thank you for being part of the cognitive revolution.

• 1:02:15Speaker 00

  Thank you for having me. It's been really fun.