 I can tell you that I feel this excitement and creativity around the new technologies in biology like it was Silicon Valley seven, eight years ago. I remember having really fun geek-outs when I was there like, oh my gosh, we could use AW, like Amazon web services just came out, like we can use this and train a big network and then we can solve this and we'll make a company. People felt that. Now the conversations I have here are the same thing. People are like, wait a second, you could use CRISPR to knock down this transcript and oh my God, you've solved a disease. That's really, and then people just like start piecing together these different parts. And so if there's one thing that like is important to get people excited about is that shift of biology is now this engineering substrate, not just like a descriptive thing. Boom, what's up everyone? Welcome to Simulation. I'm your host, Alan Sokian. We are still at MIT in Cambridge in Massachusetts. We are at the Brain and Cognitive Sciences Building. We are gonna be talking about developing technologies to capture the complexity of biology, the brain and so much more. We have Daniel Goodwin joining us on the show, hello. Dude, thrilled to be here. Thank you so much for coming on, I'm super excited. And huge shout out to Adam Marblestone and Alex K. Chen for making these introductions happen. Good people. Good people for sure. Daniel Goodwin is a PhD student at MIT Media Lab making technologies to explore the brain. Previously he was an entrepreneur, co-founder of My Like You and entrepreneur in residence at IDEO. And you can find his website link at Danielrgoodwin.com as well as his LinkedIn profile in the bio. All right, Dan, let's start things off with our question that we love asking. We find ourselves as stewards of earth. What is your current take on the state of humanity? Oh man, start on the small questions, huh? I have to answer that from three different perspectives. And so I guess as we get through this conversation, it's gonna come up. There's different ways of viewing the same thing. So I'm gonna start with the perspective of the designer. And the designer starts with the empathy first for the individual. And what I say is that that says everyone's beautiful, everyone's got a story. So the story of humanity right now from that perspective is that a lot of things are changing, people are concerned, but humanity's beautiful. I'm then gonna contrast that with the perspective of somebody that in his career has bought hundreds of thousands of dollars of ads online. And that gives me this perspective that's a little darker where I think about the social media and I think about the persuasion machines because I really think that's the right term for those things. And when you think of it that way, the future of humanity is scary because we have these new technologies. And I've seen those things work. I was one of the first people to buy Facebook ads all the way back in 2010. And at the beginning then we were amazed by how good they were. And there's a reason those things doubled in price every year, 2010, 2011, 2012, all the way to like 2016. So anyway, from the persuasion machine side, I think there's a technologies that are very scary and the only way to really fight against that, I think, is to improve our education, improve our minds. And so that's why I love what you're doing. The third perspective is the scientist of what I do now. And the scientist of what I do now is I work with technologies to change biology. We're both exploring biology, but then as we go into it, we realize that as we find these new things, we can repurpose them and we can use those to fight disease, to explore the brain deeper, to explore other organs, to start doing multi-scale maps. And so from that perspective, I'm extremely excited because I think that what we're sitting on now is kind of very analogous to kind of where we were with electronics 40 years ago. What we're doing now with the, being able to recombine all these different biological tools to start building applications on top of that. Oh man, it's a very, very exciting time for humanity in terms of the technologies and biology that are gonna come out. Yes, yes. Well, those three perspectives, the designer, the persuasive technologists, and then the one that's actually building the technologies to better understand the body and give us the tools to increase our longevity, our working memory, our intelligence, augment, eradicate disease, augment ourselves. Yeah, your radar, it's so funny because you'll end up explaining this, but thanks to a lot of mentors and other influencers, you've been able to align yourself with that higher purpose and meaning to pursue what you're doing now. And I'm excited to be able to get into that. Yeah, there's a really great quote that I heard recently which is that, the goal is to fall in love with a problem, not a technology. And so if you've got, and I love that, there's a bunch of different ways to slice it. There's a really good read which is Mansearch for Meaning. And what he says all the time is he quotes Nietzsche, who says like that if you have a good why you can survive anyhow. Anyhow, baby. And it's a great idea. And it's the same thing like when you've got a battle flag that people can get behind, it's night and day. And so, yeah, I mean finding, build it using biology to make humans better is one of the most inspiring things. But now, but you also need to start diving into that even further, right? I want to make the world a better place, doesn't mean much, right? I want to understand the brain to cure disease, that's there and then if you can go even one step further then you start getting into a tangible problem. Yes, and being technology agnostic as you tackle that problem is critical. And because there's always going to be new frontiers and technology pushed to be able to tackle that problem, you need to adapt and yeah. And that's one of the things I'm really excited to chat with you too, because I've noticed that a lot of the biggest technologies that end up coming out started as a sub-problem of a sub-problem. And you look at Google, right? Google was structuring information in the digital library initiative from NSF, right? That was a sub-problem of a sub-problem. And by the way, for any kids listening out there, do your linear algebra homework because PageRank is just an eigenvalue problem. Anyway, PageRank is a... PageRank is like actually a very simple linear algebra problem. It's brilliant, but the point is, yeah, anyway, that is getting hugely tangential, but the idea is being exposed to what the valuable sub-problems are. Yeah, and so to think about these huge things that totally change our life, if you trace back their lineage, they start as a sub-problem of a sub-problem, turns out it was really big. CRISPR, you just talked to George Church, is the same idea in terms of programmable, nucleic acid editing, right? Who would have thought that was important? Well, biologists would have thought that was important, but the average person on the street 10 years ago would have meant nothing to, right? Oh, we already know things like can cut DNA, that's not important. Oh, but if we can make it programmable, anybody can use it to cut anything. Next thing you know, you're solving monogenic diseases, you're giving kids, like, solving forms of blindness, solving forms of, like solving sickle cell anemia, like there's these very interesting applications that are coming from what originally was a sub-problem of a sub-problem. And then all of a sudden it's the family members themselves that say that my child, my parent, my relative got their fix. Is now saying the phrase CRISPR, right? Yeah, the scientists. Where can I put more money into that? Yeah, yeah, into that. So, and that's what I love so much, is like, you know, for my journey as a PhD student is to be exposed to these sub-problems of sub-problems. The things that I get most excited about now, I never even would have known were a problem, or I never would have known were a challenge. When I started this PhD four years ago, I thought that it was gonna be taking machine learning algorithms and processing image data, and then that was what was gonna be to solve the brain. Four years later, and we can, we're hopping around here, I think it'd be helpful to tell you that the narrative in chronological order, but the punchline at least where I'm at now is what I find myself being so excited about now are things I wouldn't even have known were an issue before I started doing this. And I think that's also one of the things that I can sell something here, is to get people excited about biology. Because when I took biology in high school, it was just, it was a descriptive science. We had to study and memorize all the different things, and who cares about it? But now I see biology as an engineer. And when I see biology as an engineer, it's no longer a descriptive science, it's a physical science with applications that are super wide ranging, and that is really fun. Damn, that is a profound pivot in thinking. Yeah, it is, yeah, it's no longer just labeling the cell, but it's now become a, I'm applying this as an engineer across the world in agriculture, in healthcare, in X field, Y field. It's just endless where the applications are. So this is only the beginning of whatever's quoted to be a hundred trillion dollar industry over the next hundred years or something crazy like that. So yeah, that means there's a lot of entrepreneurship and a lot of creative potential to be explored in the field. Let's do the journey. Okay, so you've had a crazy journey, born in the Bay Area, then raised in Idaho, and then down in LA for college, and then up to Stanford for brain computer interfaces. Teach us about this trajectory. Yeah, so I work on the brain now. My undergrad was I set off bombs on bridges. Whoa. Yeah, so it's been a journey. How do we- To rebuild that? No, to characterize them. So, and actually simulation-wise. No, no, no, we actually set off, we called them cold gas thrusters. We went out to remote bridges in Maine and we set off these things and they create these impulses throughout the structure and we put these really high resolution accelerometers all throughout the structure and you can characterize it. And then you can characterize the resonant frequency of these bridges and that gives you an idea of if a terrorist was to place a bomb here, where would the weak point be? Anyway, so how did that go from that? Yeah, it was really fun. Professor Daron at Harvey Mould College was my mentor and he was phenomenal. So how did I go from that to the brain? Well, the thing, all of that research was signal processing, right? So we got all these accelerometers on a bridge. We used those to create a model of the bridge. Well, in some ways that's similar to the way I thought of the brain initially, right? We take electrodes, we can put them in the brain, we can measure high resolution electrical activity and then we could decode that to move robots. And that was my dream. And so I went to, I shall say, it's also when I started getting involved with entrepreneurial activities. So the end of college started doing some web stuff. But really, my goal was to become a neuroscientist. So I enrolled at Stanford as a PhD student in electrical engineering. And I got talking to a person in the lab that was doing these brain-computer interfaces using the robots. And the mode there, and it still kind of is the way it works, is that you put these electrodes in an animal and then you create a machine learning algorithm to learn what that activity is in order to articulate the degrees of freedom of a robotic arm. And then I remember I had this very clear moment where I was a first-year grad student, super wide-eyed, and I met this guy in the hallway, or it was like the social, and he was working in this lab that I really, really wanted to work in. And I just said, blah, this is me, like I really want to work in this stuff. And he looks at me so sad and he's just like, I've spent four years training an algorithm on one animal, and the animal died. And it had taken another four years to retrain the algorithm to do the robotic task it was doing. And I remember this because I remember feeling so shocked and so sad for him. And that's where I realized, okay, that created my first crisis of confidence in neuroscience, where I was like, okay, clearly there's gotta be a better way. I don't want electrodes shoved in my brain. The exceptions are for people that are locked in or have a severe spinal cord injury. There was no way to translate the data. You can't translate it. To another animal? Not at that, now this was 10 years ago. And all those people that work in those fields are fantastic and I'm sure they've solved it. But the state of the art at that point was pretty much like one animal, one algorithm. Yeah, wow. So that's when I took a step back and I said, okay, well, what really is the problem here? Like, is the problem electrodes in the brain or is the problem the data processing? And also at this point, I was a 20-something dude. I was in the middle of Silicon Valley, the mobile wave is taking off. And I realized that the data processing is the biggest challenge just across the board. So another fantastic mentor of mine. So I got to work with Fei-Fei Li, superstar and really nice person. And so I worked in her lab for a while and we were doing natural scene recognition. Natural scene recognition being like, is this a photo of a classroom? Is this a photo of the woods? The funny thing about this is this is right before the Alex neck came out, right before the phrase deep learning really just like hockey stick. So we were working on things before the deep learning what craves. So in fact, like everything I worked on then was pretty much just wiped out by the Hinton's work and yeah, the neural network's work. But the amazing thing about that was to start doing image processing. And then that's when I realized that, you know, like this is when iPhones had been out for about two years. So the idea of like all this data coming out in the world and the need to process it was like, this was like a really like a big inflection point. And I fell in love with that, the data processing side. So the first thing I did was I did FMRI analysis, thinking that would be an interesting way to study the brain. So just to say this is so critical because it's something around 90% of all data is created is good created in the last two years. It just keeps happening with the IOT era and with all the brain mapping that we're doing in the biology mapping. It's just, we need to be able to process this data in coherent ways that can be passed down to the Gen Z and past them. I regularly talk about terabyte data sets. Like I will use terabyte data sets and like I'm 33 years old. I started life with the three and a half inch poppy. So it's crazy to be thinking about a room built. Yeah, exactly. Okay, cut from where I... So anyway, I was doing the image and processing analysis but what I realized is that the deeper I went in the image and like the data processing side, the further away I was getting from touching people. And the thing is like, I think an important part about my background is my mom's a nurse, my dad's a professor. And I really liked, and I kind of have that perspective of like I want to work on hard problems but it's got to be rooted in people. And so I found myself drifting, like just chasing harder and harder problems but then kind of getting lost on what my battle flag was. And I got really lucky by enrolling in a class of the Stanford Design School. And it was by complete fluke that I even got in there. I really should be like, I should not take any credit for this. I was dating a girl at the time and she's like, I really want to take this class. It was called From Play to Innovation and it was taught by Stuart Brown and Brendan Boyle. Amazing, Stuart Brown by the way is the guy who coined the idea of play as a primal need. His, oh, it's such good work. And then Brendan Boyle is a partner at IDEO and he teaches the D-School. And those two guys- We need to get them on the show, yeah. Oh my God, yes. From play to- From play to innovation. To innovation, that's beautiful. Yeah, and it was the perfect time for where I was in life because I had now like wandered down, like kind of wandered into the dark searching for hard computational problems and I was distant from like the why I do it. When you go into the Stanford Design School and the IDEO way of thinking, the first thing you ask is, who are we trying to touch? You know, and then why? You know, it was like, what's their problem? What are their needs and then how can we solve it? The very last thing you get to is like, what is the technology we actually need? Before you even get to that, you've already prototyped things, you've come, like you've got some confidence and that's what you need to build and then you like do the hard part of the technology. And for me, like I can't tell you what a 180 that was. And that's also when I realized like, oh man, I'm not made for the PhD right now. And so that, you know, once again, like trying to figure things out and I had this like big moment of like, okay, I want to like, I want to recenter on people. And so Brendan Boyle at IDEO gave me the great opportunities. Like, look, IDEO is, we have a role for entrepreneur residents. You've got some background in mobile apps at that point. I spent a summer working with a friend at the business school, created a company called Pocket Gems, which is now a gigantic mobile gaming superstar company. And so I got to, so I spent some time in startups. I'd already, when I said like I bought Facebook ads, like this is when I bought Facebook ads. And that was the start of that. And realizing like, whoa, you know, like people are quite modelable when it comes to like getting people to click on things. That's like, that's kind of like when I, when I say things that are cynical, it's based on some of this background. People are modelable to get them to click on things. So, you know, statements like that, where are ethics, morals? Oh, I get it. I feel this. And like, Yeah, that's huge that you felt that. And then the fact that you've tapped in to feel and talk to people about it, versus just doing it without recalibrating and thinking, yeah, yeah, yeah, that's huge. So the conversation I had with IDEO was that they were interested in mobile apps. They were interested in trying to explore things outside, you know, like the fee for service. Like, how can we think, like IDEO has got such a great background. I should describe who IDEO is, because I'm just, I'm throwing around this buzzword. For me, it's been such an important part of who I am. But it is just, you know, it's a name. So I, yeah. Such a good design firm. Such a good design firm. Yeah, fantastic place. Been around for 30 or 40 years. The foundation, like founded by four people who are all giants in their respective fields. And it's got the background of being involved with the early days of the, like they invented the mouse, they invented the laptop. They're, and what they preach, or what they sell is a process of solving problems. Right, like, that's kind of a really important thing. Like you don't guarantee an outcome. You guarantee doing a process as well as you can. And so their whole thing is called the human-centered design process. And that starts with understanding people, prototyping towards something that solves their problem. And then you can figure out how you want to scale it up. Keeping the human, yeah. Focusing every human-centered design. That's the phrase. And so, it would end up being nine rounds of interviews. You know, it was like, they took it very seriously. And so when I showed up there, they have this policy there. Like everyone at IDEO is really interesting and awesome and yada yada. So we don't care. It's all about like, welcome to the place. Like, what do you do first? You know, like, it's all about like, what credibility you can build inside the firm. And so I was coached like, hey look, we have this great opportunity. It fits your background perfectly. We've got this joint venture with Sesame Street. And we want to explore what it means to do, like mobile apps. And specifically, like, we started with this idea of potty training. So now this is like a very funny point to just pause, right? So like, I'm a 26-year-old guy. I just went from a windowless office in the computer science building, you know, like where we took a photo out of Fei-Fei's office, like just to see what sunlight looks like and we put it up on the wall. Like that was like the windowless office we worked in. And then now like I'm in this beautiful design firm where like people are well-dressed and well-slept and like I'm sitting on the bathroom floor like learning about potty training with like a parent and like, you know, like two-year-old and her parent. But it was also like such a great experience for me, which is like to say that this process works. You know? Like we went through this really long journey of like we had a hypothesis, we tested it. Everything that I do in science is actually like, you can map it one-to-one. And I think that's where the human-centered design process like probably started from. So... The translational application of it. Yeah, but we're not talking about like, we're not solving some part of the spice zone. Like we're solving a human need. And so... Getting kids potty trained. Getting kids potty trained is where we started. Sesame Street partnership. We were prototyping ways like, you know, because yeah, so there was a partnership and we're trying to figure out like what the best opportunity would be. And so it started with potty training, but through that process of prototyping, we realized that it actually wasn't potty training. It was the opportunity for it to use Elmo as like the good big brother. Because that's what Elmo is. Elmo is a three-year-old older brother. Yeah. You know, who sets a good example. Yeah. And so... Oh, that's great. Yeah, so from that opportunity, we realized that kids just like, parents take up after yourself. Exactly. Say thank you. Things like that. Brush your teeth. Brush your teeth. Oh, that's cute. Yeah, yeah, so in the end, like we made a product that sounds very simple, but is based on these insights of spending hours with kids and parents, specifically focused on potty training. But what we realized is it's just like having Elmo on demand is like the good older brother. Elmo on demand, yeah. So we made an app where kids get a FaceTime from Elmo. They get a FaceTime. They get a FaceTime call from Elmo. That's exactly it. They're kissing and Elmo's like, hey. Hey, it's time to go brush your teeth. Wow. So, yeah. Wow, that's great. Yeah. Becomes a trusted older brother. We're so far away from neuroscience, but it's gonna come back. It's coming back, it's coming back. There's a David Ewing Duncan talking to robots is coming out soon. And he talks about human robot interactions of the future. And one of them is the Teddy robot. And the Teddy bot is the trusted robot that works with the kids. And all of their kids' developmental needs and the love and compassion and the recommendation in this big brother sense. And so, this is a huge part of our future, I think. And I love the idea of getting from a trusted source like an Elmo. Yeah. That was a very crazy transition from the basement with the windowless room with no light and not having translational applications. Yeah. And I was living in a garage before then. I mean, I'm just every cliche rolled into one. OK, so that's how you do it. Well, so this is the idea thing. And what I should say is what we did is with the close Elmo stories that had to go through the whole Sesame Workshop educational thing. They take their thing, their work very, very seriously. So did it satisfy the educational criteria that they set out? Then I got to go to New York and actually be on set because we got to write the words and see Elmo perform it. And so anyway, it was a really, really great story. And then I got to see its scale. So that was the second app that I was involved in that did well. So first it was social mobile gaming, then it was Elmo calls. But the goal of being at IDEO was to start a company. That was from the start, that was what I was supposed to do. Entrepreneur in residence. Entrepreneur in residence, found a company. And so what I was focusing on, and this actually goes back to the second perspective when we started the State of Humanity question, is I looked around at the way we use the internet. And I looked at every Google search we do is used to service ads. Every interaction we have on Facebook is used to optimize our mini feed and then service ads. So I started being curious with, well, what can you do to give people their own data back? That was the design prompt, actually. How can you give some people their value back? I'm wearing an activity tracker watch and that. And so I guess there's a decently long story there. But I was prototyping with different ways of just capturing phone. And you know what? The data you produce from your phone is this data exhaust. It's just there, right? And so the question was, how can we give this back to people in a meaningful way? And so I started, like everybody does in 2012, building a photo sharing app. And the great thing about being in a good creative environment is people will tell you if it sucks. Ed, you know who I'm doing my PhD with, he'll tell me if I'm doing bad work. When I was at IDEO and I was doing this project, people told me, like, I don't get it. Why are you doing this? And that's when I met. You have to be careful, sometimes people can't necessarily also see the imagination. Yeah, it's kind of double edged in a sense. But yeah, I agree, though, you need to be around people that'll tell you why it won't work and be very clear and stuff. And yeah, you go back and forth. Yeah, that's a good point, though. Yeah, I mean, like, well, so this is where a lot of people have different perspectives. Brendan Boyle has a great idea that people are the most creative when they're the most comfortable. And I like that idea. I would also push on it. I think, for me, I do some of my best work when I'm under pressure. Likewise. And so I don't know if there's the right answer for each person. But I do know that when I'm really doing something that's hard, I don't want hugs. I want people to push me, and I want to be better. So when I was showing this photo sharing app around people, I don't get it. I was like, thank you, because that's the person I'd be trying to ship this thing to. And then I talked to my mentor at the time, a guy named Dan. And Dan, 10 years older than me, same height, same build. I think of his brain like, if it was me after 10 cups of coffee, and what we were talking about is the most interesting thing in the world, that's Dan's brain. He led a lot of the venture thinking at IDEO. And the way he could connect ideas and bring people together and see 10 years ahead, it was just everything you'd want from a mentor. And I showed him this. I'm like, hey, look, I did this thing. It tracks you everywhere you go. You can take photos with it, and you can match. And he's like, yeah, I don't get it either. But he said, look, does this tell me how I'm moving? I was like, yeah. He's like, could you do this to tell me whether I'm driving or not? And I realized immediately, I was like, oh, man, OK. I see what we're talking about here. When you work in the consulting business, you drive to different places. Every time you drive, you can expense it $0.55 a mile. I thought, oh my gosh, that's a really good business. So we will automatically track every mile you drive using the phone in your pocket. Very, very simple business. Now, Dan has a great quote, which is, don't confuse a clear view for a short distance. So just because we had a very clear view, we're making a simple product. We just needed to track mileage. But getting that thing out, so we left IDEO. We started the company, just him and I. Oh, wow. And that's what we were working on. And it took us about a year and a bit and five iterations before we could ship it. And so I was the founding CTO. Dan was the founding CEO. Dan has a phenomenal product sense. And so we had something ready, like, OK, we can put it in the app store. He's like, no, it's not good enough. Sure enough, we did some user testing. Yep, not good enough, scrapped it all, rebuilt the interface. Anyway, five times later, we ship it and immediately it starts succeeding. Because we had something like we had something based on a need. We'd done all the same IDEO user testing. Like, we had the insight. Gosh, that's huge to be patient until you iterate on it until it's beautiful enough to actually hockey stick up the user experience, and it'll actually share it with other people and get people using it that way, yeah. Yeah, and so I was living off street burritos in the mission of San Francisco. Oh, man, yeah, it was a tough time. And so that's actually when we got involved with StartX. So we went from a situation of being in many different, you know, like in a big, vibrant environment, IDEO, to basically kind of working out of our apartments in the early days of the startup. And I think it's really important, and I love what you're doing with this educating people. And part of what a big, important thing to say is I think of myself as a creative, right? And I've learned what it means to be on a creative journey. And a big part of the creative journey is being depressed and being bummed and feeling like crap and working through that I think is just part of it. That's like part of doing something that's hard. So when we first started this company and I wasn't leaving my apartment for four days at a time just sitting in my underwear programming, I felt like crap. It was a really hard time for me personally. And so it also worked out that StartX was a few years old at that point and they were looking for entrepreneurs and residents. And there was my background at that point, there was Dan's background. And so we went there just a few days a week to hang out. And it was all these, like, Stanford founders doing these companies and a lot of them have succeeded, by the way. Five, six years later, the success rate is super high. But it was great to just kind of be in a community again. And this is before My Like You started growing and all that. So that's also a good example of a creative community. I think we'll come back to that, but it's definitely worth putting on the side. So the point is My Like You started working very early on. We had something that people totally got. We had a product that worked, and then we were at the scaling up phase. And this is kind of funny to say, but you work all day on a really hard, big problem, right? All right, we've got something working, but now we need to make the next iteration, the next iteration. And it's hard to get that thing that makes you feel good. Like, it's hard to get that instant gratification. And so for me, years into the company, the two most gratifying things for me to do was customer support and reading neuroscience papers. That was like, those were my gifts to myself at the end of the day. And you were teaching me the customer support was because you could immediately get gratification from helping a customer. Yeah, it feels good. It's like, hey, I happen to miss this drive. And I look at the data, I was like, yep, I see what happened, solved it, and thank you so much for being a customer. And like, here's a perk. And I'm like, oh, thank you. You also see some weird cases in customer support. Like, oh man, there was this one guy who was just being so rude, you know. And every time I'd be like, oh yes, you know, like, you know, we'll like happy to help. Stay happy. Yeah, it's like so rude back. And then one day he just emailed me back like weeks later. He's like, look, I was going through a divorce. You were like the only person who's nice to me. I appreciate it. It's like, wow, man. So yeah, humanity at scale, right? Like the weird emails you get. And then you actually were successful enough and ended up selling to Microsoft. So that happened later, yeah. So the thing is I hit this point where I realized that I was like, what I, the dream I had was to make a successful company. And it was succeeding. I was working with one of my heroes, you know, that the team was growing, that the user base was growing. We had a product that people loved. But I was feeling that something was missing. Like something like, and I was like, and that was coming at a cost of my personal productivity. That was coming at a cost of, you know, like the light, like satisfaction, whatever you want to call it. So I started thinking like, oh gosh, you know, what is it? And then I looked at like, well, what am I doing for fun at night? I'm reading neuroscience papers. And I had this feeling that like I should be there. Like that's what's bothering me. Like I needed to be in it. Yeah. And so we started the phase out. So along with one of our financing rounds, I started, like I started phasing out. And this is credit to the people at My Like You, right? Like I was so lucky just to work with the best people. Chuck and Dan, Chuck became the CEO. You know, like that, like a founder leaving, like you just got to call it out. Like that is a hard. And that can be a very bad thing, right? But if they're moving on to something else they're really passionate about and you get the right team filling the spot. Right. And you do it collaboratively. You know, so like when I say this, you know, like I'm the product of really great people investing me. Like Dan, like it was a phenomenal guy to invest in me. And likewise, I would say the same thing. I think we all are to a certain extent. The product of great people investing in us. Yeah. And then now you give it back, right? Like, hey, I know how to do things. Like let me invest back in you. That's right. So, and I think that's the best way you can show respect to the mentors, right? Is like show that you can be a good mentor too. Yeah. Which is also why like I think it's important to be really open about like, hey, look, I'm not perfect. I've screwed up so many times. The point is that's when I started transitioning back into neuroscience. And so this point I'm 27, 28. And there was a heroic neuroscientist named Sebastian Song. And I'd met him all as an IDO. And I reached out to him. I said, look, like your work doing deep neural nets on very, very large scale image data is like, I think the biggest thing in neuroscience. I really want to work with you. Like, what can I do? And he offered me a job to come work with him out on East Coast. So I moved across, this is when I started working at the Simons Foundation. That's when we started working on the paper we were talking about. Yeah. So you got that paper was published in Cell in 2015. Yeah. Yeah. That's huge. That's a big deal to have that done. And I'm glad that you made a reach out to Sebastian to make that happen. I strongly suggest if you get the chance to talk with him. Sebastian Song. Sebastian Song, yeah, a really, really magnificent guy. He's a physicist turned into a machine, like computer scientist turned into a neuroscientist. Very, very, very cool mix. The point of being there, and we can talk about that work is it's going to get into the idea of complexity in biology. But part of me being there was also like, this is working at the Simons Foundation. Was me testing hypothesis. Nah, like, look, I know how to program. I know how to do machine learning stuff. Just give me the data. I'm already a neuroscientist. I've read papers before. And that's when I realized working at the Simons Foundation being surrounded by some of the best neuroscientists in the world, that I was missing something. I was fine. When there's a seat at the table, people were talking in a different plane than I was. And I realized, darn it. Hypothesis disproven, I need to get my PhD. And so specifically what I became really interested in is the source of new data. Because this is the Peter Norvig out of Google saying the idea that the best data is better than the best algorithms. And so with the brain and biology, we're at this point now where in combination with the genetic modification tools and our ability to understand more and more organisms and mine more and more insights out of all these different kind of metagenomic data sets, we start getting more tools to get more signal out of biology. And I guess the way I would describe it, thinking about biology now, is if we're in this building now, we think we studied the complexity of this building with our eyes. And we see that there's this brick in the wall and there's a million of these bricks in the building. Okay, that's how what the building's composed of. But when you get all the way down to a single cell, you can't use light anymore. You can't say there's a million bricks in the membrane of the cell. It just doesn't work. It's like the protein is so much smaller than the wavelength of visible light that we don't have anything to throw at it to bounce back. So what you start doing- Protein's smaller than a wavelength of visible light. Oh yeah, our visible wavelength is 500 nanometers. Proteins are tens of nanometers. Yeah, one nanometer, yeah. It's like orders of magnitude, so you can't see it. But the thing is, your cells are filled with millions of these things. But you can't image it other ways. You can image it other ways, but here's the trick. You have to pick a subset of the proteins that you want to explore. That's the work that we did in 2015. And so the thing is, when you want to visualize protein, you start with an antibody. An antibody is taken from your immune system that can be programmed to bind to a protein. And then once you've got something that you've built that can bind to protein, you can modify that to have some fluorescence. So that's how you can visualize proteins. I care about actin beta, like a part of the cytoskeleton of cells. So you take an antibody for actin beta and then it'll stick to every actin more or less and then you can put a fluorescent tag on that. Then you see all of actin. Then the question is, how many proteins can you do like that in a single image? If you want to take a picture of a cell, I'll go right to the punchline. It's four. More or less, it's four colors that you can resolve out of a light microscope. Why? That's it. Because it's part of the physics of the way the light excites from these proteins. You can only pack so much into the visible wavelength of light. So if you get them too close together, they start cross-bleeding into each other and then you can't resolve them. Oh, we can't resolve them. So really, you're looking at only about four colors. And how does that work? This is the antibody method. This is the antibody method, yeah. So where we've been to biology up to 2014 is you pick your four colors. You pick your four bricks that your cell is built out of. And that's super powerful. Don't get me wrong. We've made so many fantastic discoveries there. But if you want to really tell me how a cell works from the ground up with four pieces out of a million, nah, it's like it's limiting. Totally. So how do we do it? How do we do it? This was the paper, though, that you were just describing. Thank you for that. Yeah. So what we showed in that paper, this was Sebastian Sung and Quenghun Cheng, who's actually in this building. Quenghun, yeah, KC is amazing, yeah. Things really big. And so what KC had developed in his lab was the ability to do multiple rounds of this antibody stain. So now instead of doing four different antibody targets, four different proteins, you can now do four times, for each time you hit the microscope. So you do two imaging rounds, you can now see eight proteins. I think in the paper we showed 20 or something like that. And that's really big, because now you start seeing a lot more complexity in biology. Because 20 starts becoming a lot more discriminative than four, especially when you start talking about the brain. Like one of the biggest things that people argue about in neuroscience is the idea of cell typing. Like how many different types of neurons are there? Or glia as well. Glia, but even just neurons. Yeah, like even just neurons. Oh, just the neuron. Yeah, people say there's more types of neurons than any other cell type in the body. So what are the other neurons broken down into? What are their subsets? Yeah, well it depends where in the brain. It depends whether it's an excitatory neuron or an inhibitory neuron. Oh, interesting. Yeah, yeah, OK. Yeah, and then it's a deep world. So then you have these two warring factions. You've got the lumpers that just want to put them all together. And then you've got the splitters who are like, actually no. Excitatories one, inhibitories one. Oh, but then subunits of that. And then occipitals one, and cortexes one, and amygdala's one. Is that, they actually think that the cells in the amygdala are different than the cells in the cortex. That's what, yeah, well then, yes. And so people are able to do that from two different molecular techniques. Well, because one's been evolving also much longer, limbic structures have evolved longer. So we're talking about different regions of the brain. Yeah, and the cerebellum has these beautiful, like if you look at Purkinje neurons, they're just these big, beefy, planar neurons with all these big, beautiful spines. But yeah, so the point is, yeah, it's very hard to, like, that's something you have to show, not tell. There's a lot of different cell types. We'll have some nice embeds here as well, but yeah. Because there's been some gorgeous fiber that you can see the neural architecture, mapping that and being able to show that, at least for now. And that will have some embeds here to see. And then furthermore, as you're saying, all the way down from where we go up from the spinal column all the way up to the very tippy-toppy neocortex is all potentially different neurons. Yes, and then you start getting like the different parts of each neuron. Then you have like the subcompartmentalization of that. And so the thing is, like, yes, there's two different molecular techniques that people have been doing so far to characterize different cell types. And I'm not trying to avoid going too deep in the cell type world, but it is a good example of where we are with biological tools. So one thing is you can look at them. And when you look at them, generally you do different antibody stains, right? Hey, you know what? Like, if it's an inhibitory neuron, it's probably expressing GABA. So like, let's do an antibody against GABA, right? Or like, and then, or if it's inhibitory, it's probably got V-glutes, like vesicles for glutamine. Like, it's gonna be, we're gonna do that. And sure enough, like, if you antibody stain the two, you'll never see those two cross-talking, right? So that's one type of cell type. But then there's the other side where we've made so much progress thanks to one, like, some of the people you've interviewed with sequencing, right? So now sequencing is very interesting because now we're starting getting into a type of data that we can understand. We're getting into digital data. Is inhibitory and excitatory neurons of different sequences, too? Well, they have different expression patterns. The expression patterns. Yeah, so this is fun because like, I gotta keep in mind that like, I'm four years deep as a biologist, right? And I'm a child. Yeah. And a lot of other people are also somewhat children. Some are really intelligent. So we have a big audience, very. Look, when I first started this PhD, I didn't even know how to use a pipette. Yeah, yeah, for sure. So, and I think it's also worth pointing out like the creative journey that when I first started, I was like, look, I'm a good computer programmer. Like, I know how to do design. Like, I'll figure it out. It's just slightly different. You came to the right place with your background, though, of computer science. The media lab. Yeah, the media lab, yeah. Well, there's an important part of, I think my personal journey, which is fun to say is like, when I first, I went straight into experiments, like whatever, like, I'm smart, I'll figure it out. I did really bad work for months. And I did work that was so bad, I was in the bad track of, I knew what I was doing was bad, like low quality, but I didn't know how to make it better. And the result of that was I was getting heart palpitations. Like, I was missing heartbeats. Just like, I was pipetting, I'm like, I don't know why I can't do better work. And then that's what, once again, like, come back to mentorship, right? I got really lucky, and I worked under a phenomenal postdoc named Shahar. And when I started working on this project, he started giving me the structure of thinking about biology and thinking about these challenges. And so together, we've been working on this great project that's several years in the making in collaboration with George Church's lab and Adam who connected us and Ed's group. And what we're trying to do is we're trying to actually see the sequences in space. The sequences. See the sequences in space. Let me live in space. In space. Space between nucleotides? Or space? Space means a lot, yeah. Yeah? Space here? Where would space? Space inside the original tissue. Okay. Oh, okay. Yeah, and so this is fun because like I'm like talking to myself four years ago. Like when we have this conversation, it's like what is this stuff, yeah. You got the structure from a mentor of how to see biology. Shahar? Shahar, yeah. This is so critical is that when we get the mentor's multi-year long journey perspective that can augment our so we don't need to go through the same treacherous process of figuring it out. But we can get kind of a cheat code in a sense. Can help a lot. Yeah. Yeah. There's nuance there too. There's so much nuance because there's a time where it's best to be a student, right? Like there's times where it's just like tie on the white belt. Be like I trust everything you tell me and I'm gonna do it because like you're gonna teach me how to be better. But then after a while as a student you have to start developing. Like I don't believe everything you tell me. I'm gonna start pushing back. That's good. And the question is like you need to figure out what your optimal rate of growth is, right? And so for me it's been this interesting thing where I started to be like screw everybody. I'm gonna figure it out. My rate of learning was flat. I was doing nothing. And then I started working on a great mentor. And every time we got into technical argument he was always right. And then I was like all right, white belt on, you know. And then we started working through it. And then now like it's fun to have these conversations and it's like this nice peer. And I think like all my mentors in the past I think like it grows to that. You know it'll always be that balance of like you're my mentor and I appreciate you. But it's fun to start being able to push back. It's like hey, you know what if we did this? And so space. So let's go into space, yeah. So the amazing thing about biology recently is the idea of complexity is now expanded with the idea of sequencing. So now we're able to get into a world we understand. And I mean by that is like now we're getting into digital data. We understand the idea of ACTG, right? We think about like you know like when we think about our genome, right? We, it was sequenced when 2001 and then Barack Obama went on the stage in 2013. And he had this wonderful status like for every dollar that went into the human genome project 140 came out in terms of startup companies and add like overall additions to the U.S. economy. We need more things like that as soon as possible. Yeah, we need more things like that. And people I think begin to get the idea of ATCG. Right, like they started getting okay we were six billion base pairs, three billion base pairs. And somehow like that long code makes us who we are. Right, so let's talk biologically quickly of like why that makes us who we are. That genome, that's three billion base pairs, six billion base pairs. If you stretched it out by the way it would be a meter and a half. But somehow that meter and a half gets coiled all the way into a tiny little cell that's just five millions of a meter big across. So first of all that's a biological miracle. Now that did. Whoa, okay so. We're gonna wander all over the place. But it's fun. So the six billion bases straight out is a meter and a half and it can get coiled up in a cell that's only five microns. Five, 10, depending on the cell. It gets quote on the nucleus, yeah. In the nucleus, yeah, yeah. Whoa, okay, yeah, yeah. So this is how a biology over the billions of years has figured out how to become very, very efficient. Okay, continue. Yeah, so some, okay, so you've got this ball of DNA. The ball of DNA will selectively open up. When it selectively opens up that DNA can start being transcribed into mRNA. The mRNA is what then encodes for proteins. And the proteins are the bricks of the cell. They are like your whole cell is proteins doing things massively parallel. Occasionally the nucleus which houses the DNA opens and gives a little messenger RNA to make protein. And then the protein does the further processes, catalyze further processes such as digestion or different types of processes. Yeah, I mean, a protein will do everything from being structural to being functional. Like it'll, you know, it'll cut, I mean, when we say CRISPR-Cas9, we're talking about Cas9 being a protein that cuts DNA. So when I put food in my mouth, it's almost as though my body is going, ah, it's time to express for digestion. Yep, that's exactly right. That's exactly right. And people do studies exactly like that. Like what's the genetic expression difference of when like when you're hungry or you're thirsty versus you're normal. Yeah, that's exactly what people do. And so what do people do there is they'll just take a sample of those cells and they'll sequence it. But they're not sequencing your DNA, they're sequencing your RNA. And so this is where, you know, it's a very, it took me quite a while to like really grasp this because I come from the world of thing about source code, right? I come from the world of thing like I wrote this thing and then it does stuff. The thing is like the DNA is the source code. And by the way, like it is a huge, like not faux pas, but like it drives biologists crazy to speak in these metaphors of like, oh, it's like the DNA is like the source code. Well, it's not really, but like for the very, very like beginning of getting used to thinking about it, it's helpful for a bit. So the DNA is the source code that runs things. It produces copies of itself, which are actually the ones that produce proteins. And so the reason that that's important is that you can't really tell the, like with exceptions, you can't really tell the difference between the individual DNA inside a cell. But you can see that what we call the expression profile difference between cells. And so when we talked earlier about like GABAergic or you know, like inhibitory or excitatory neurons, neurons that do functionally different things, the DNA is the same. They're just behaving differently, which means they're expressing the different RNA. And then so you're trying to sequence the messenger RNA. That's exactly it. How do you sequence messenger RNA? Aha, well, so this is what Illumina did. I mean, this has been done by many people for a while. But the idea is that people like they would, they would take them all like, well, the first thing you do is you take this beautiful piece of biology, right? And the first thing you do is you chemically homogenize it, which means you blend it up. And then you flow the DNA or the RNA that you care about across this piece of glass. It'll functionally attach to the glass. You will then amplify it. So one copy of the same thing becomes 100, 1,000 copies. And then you can do this chemistry to make it glow, just like we were talking about with antibodies earlier. You can make it fluoresce. And what's- That's a polymerase chain reaction? PCR is a way of amplifying DNA. So it goes from the one, mRNA goes from the one to the- To many copies. To many copies through PCR or other methods. Yeah, that's sure, yeah, that's right. Or Illumina does it in a different proprietary. No, no, it's the idea. I mean, you have to prepare your sample to go in the Illumina machine. But the point is it's able to read it out base by base. By base by base, yes, yes. And so what it looks like is a bunch of dots. It's just a piece of glass that has a bunch of little dots. And then you take another image of it, you do some chemistry. And then that one dot is now fluorescing a different color. It's now fluorescing a different color. And then after you do enough rounds of that, you can now actually piece together what that was. Hey, you know, like green, green, green, purple, green, green, red, red. You know what? That's the actin' beta again. We found this transcript. And so what we can do is we can create these molecular profiles of cells. So like, keep on, like, a goal is to understand the brain. We're now like far down, we're talking about cell typing, right? You're creating molecular profiles of cells to better understand the big picture of the brain. And so what have you been learning with the molecular profiles of cells? Well, similarly, what have I been learning personally is that- Because that's the direction of the- That's what I'm, yeah, that's where I'm getting into. So we're not talking about two different things. We're talking about the complexity and sequence space. And then earlier we were talking about like viewing things in physical space. And so the goal that I've found to be really worthwhile is to be able to combine the two. Yeah. So why have we been limited for four colors, four dimensions of biology when there's so many more? And so what we're doing in the lab is we're combining these ideas. We're doing, we're capturing the physical information and we're capturing the nucleic acid information together and then we're reading them out. So what does that mean? That means we're able to see a single RNA, or sorry, we're able to see like hundreds of RNA, of different types, inside the space of a single neuron. And then we can start learning from that. And we see really weird things. This is like one of the other things that's really interesting is that biology from my perspective is on one hand is filled with these incredible experimentalists who go into these very hard problems with very little tools. Like you can see so little out of biology at first, you have to know exactly how to slice the question, right? So when Phil Sharp discovered the Intron here at MIT and won the Nobel Prize for that, it blew people's mind to be like this guy saying, hey, wait, you know what? These mRNAs that people are producing actually have junk DNA that just gets cut out and it's useless. And the whole field was like, that's the dumbest thing I've ever heard. Why would biology develop this thing to produce junk that gets cut out? And then he fought it, he proved it in the 80s with very like, what we look now is like very primitive tools and then he won the Nobel Prize for that. So my goal is to take biology to make it accessible for dummies like me. And me, yeah. Who like, maybe I can't design these like brilliant experiments to like be able to get these insights with very limiting tools. Maybe if we just made the ability to see this complexity in the sample itself, then we might start getting there. So democratizing the way that we both view and begin to probe at an abstract level some of, but you gotta know the micro level then. So you have to understand the micro level and then be able to run these hypotheses as at a more macro level to enable more of the creative potential but that people don't get some of the micro stuff. Yeah, and the scale of look, I think what you're touching on is a really important point is the scalability of a tool. Like if you, there's a really great idea, which is that if one person can do it, it's an art. If a few people can do it, it's a science. And then if a lot of people can do it, it's a technology. Well, that's a good way to put it, yeah. Right, so I've done art. I've also accidentally done art. The worst thing you can do is make something that you think is gonna be scalable and then nobody cares. What do you think are the best ways for us to now be developing technologies to understand the complexities? Yeah, okay, so this is the thing. I've now been four years in biology. I can tell you that I feel this excitement and creativity around the new technologies in biology like it was Silicon Valley seven, eight years ago. Right, I remember having really fun geek outs when I was there, like, oh my gosh, like we could use, you know, like Amazon web services just came out. Like we can use this and train a big network and then we can, you know, solve this and we'll make a company. Like people felt that. Now the conversations I have here are the same thing. People are like, wait a second, you could use CRISPR to knock down this transcript and oh my God, you've solved a disease. Like that's really, and then people just like start piecing together these different parts. And so if there's one thing that like is important to get people excited about is that shift of biology is now this engineering substrate, not just like a descriptive thing. I should say really quick, like, I took one class in biology, two classes in biology and undergrad. The second class I took, which was my last one, was ecology where we counted crabs on a beach. And I just left. I was like, that is, that's not my science. I don't care about counting crabs. I don't, that's descriptive science. I want to be able to do things with it. So your question was developing the best technologies to help us understand the complexity of biology in the brain. Yeah, so this is what we want to be able to build for. I want to be able to like pull a sample out of a meteor. And I want to say, I've only got one of these. Let's fully describe it. And right now we can't do that. We can say, okay, like we want to like, this is a alien brain and we can put it in an electron microscope and we can throw electrons at it rather than photons. And electrons can resolve smaller things and we can eventually piece together all the little connections in this brain. That's one thing we can do. But then the sample's dead. Like when you do that, you lose nucleic acids. You lose the proteins. Or we can do what people have been doing, which is these antibody stains, right? And we can do, you know, 30 proteins, right? This is a really like one-of-a-kind meteor brain. Maybe we could, you know, see 30 proteins, you know, or we can grind it up and then we can sequence it. What we want to be able to do, my goal is to be able to do like an hour. Like this is shared with people in my lab and like a lot of people in this field is we want to be able to do all of those in one sample. Yeah. And so to do that, you need this highly multiplexed RNA readout. And the reason is that people have not solved how to do the protein readout. Like highly multiplexed yet. So we can't do that, but we can read out nucleic acids. So how do we do the complex protein? That, those are some very good people working on that. That is like, so protein sequencing is a huge open challenge. Interesting. That is very much a holy grail. That's very hard to do. Yeah. But also like the thing is like, from my perspective, like I've done hard things that people didn't care about. You know, I think people will care about RNA sequencing in context of like the original tissue. And if we can do it in a scalable way, then I think that's what I'm laser focused on right now. Interesting. Okay, RNA sequencing in the actual tissue. Yeah. In the space. In the space. Yeah, so how do we preserve space? Well, like what we do in our lab is we can anchor everything in a gel. So this is like, this is one of the things. This is the hydrogel? This is the swallible hydrogel. Yeah, expansion microscopy. Expansion microscopy. Yeah. It's like the stuff that's in diapers. Yeah. Yes, that's right. That's exactly right. You've done your research. Yeah. Yeah. And so first of all, like to admire that original work. Because when I, that had just come out when I joined the lab. And also like kind of when we say like sub-problem of the sub-problem, I didn't get it. I was like, I don't care. Like I want to solve the brain. Why are you making things like gels that expand? Well, it turns out that one of the problems is the limitations of light. But also like having a scaffold to explore biology with. Like in this case, it is a physical scaffold. It will anchors all the molecules you care about. You can pull them apart and then you can do whatever you want with them. And turns out like that idea, which is like kind of esoteric to the outside world is both super scalable because it's diaper polymer and it costs nothing to do. But then also like you can do additional chemistries on that intact sample. So whereas like people have formed, like formerly if you wanted to read out lots of different RNA, you have to flow them across the glass cell. Now we can actually preserve the whole thing in the original context. So instead of grinding up that brain, you can now look at the brain and you can see hundreds, thousands of different RNA molecules. So you're adding an expansion microscopy polymer to the space between the tissue. And then you're able to make it easier for yourself to go in and do the antibody binding, enabling you to then sequence more easily the proteins and the mRNA. And I should say like, you know, like sequencing means a lot of different things. The point is that you can, we use the word multiplex, which is like a very fancy way of saying read out lots of. Multiplex. Multiplex is what we say, but that's like kind of like a, that could be like- It's like a multi-omics as well. It's kind of like you're trying to do a multiplex read out, which means sequencing. Yeah, sequencing, capture protein, you know like- So like RNA- So there's a few major classes of biological molecules, right? You've got the nucleic acids, you got the proteins, you got the sugars, lipids. Oh, interesting. So in the perfect world, you see all of them at once, right? Like that's, if you want to talk about like multi-omics, like get all the biomolecules, but turns out that's hard, right? And so like, and I think, but if you want to take that meteor brain and you want to understand it completely, you do need to be able to see all molecules. And then at least the first thing you can do is you can start by preserving them. And then eventually, you know, like if they're all physically in this thing, you can get the information out. Some wrapping thoughts are, what about then other technologies that you are just roaring with your imagination, but that are really hard to like ground and actually implement, but what is the field thinking with its wildest imagination could help? Well, I think about two things on that. I think first of all, like a big part of my background is design like scalable consumer products, right? And then deep science, like I guess whatever you call this, like doing a PhD in neurobiology, tool building for neurobiology. The thing is, the way science is done and communicated has a lot of room for growth. And specifically the funding side of it is very interesting. And so the one thing I have a particular interest in as well is like how can we start rethinking or can we start redesigning the structure of scientific funding? And that's like, so this is like, same thing, I'm answering this in the context of like if you could design any technology, I'm framing this as like, if I could design any system, I would like really want to explore this. And I think there's, I think there's interesting ideas and there's really like good opportunities. But that's like, that's one thing I think about. Yeah. Shout out to the brain mind ecosystem. Brain-mind.org. I mean, they're really rethinking financing for the brain. They're doing a great job. The ecosystems of neuroscientists, philanthropists, all that type of stuff. So yeah. Okay, interesting. So that's the first one. What's the second thought? Yeah. So the other thing I would think about is, okay, this is gonna sound jargony. So we're gonna unpack it. Is non-monolithic artificial intelligence. Yes, teach us. Okay. So I'm now at the point where when people talk AI, blah, blah, blah, like for me, the discussion's over. Like it may not be exactly what we imagined for the movies, but it's gonna hit. That's my philosophy. Yeah, totally. I think it already happens. Like this is like, you know, and we see it everywhere. We see it in the success of ads. Like that's kind of why I come back to that. Like there's a lot of money being poured in that. Those ads work for a reason because those things are perfectly targeted. How are they perfectly targeted? They perfectly understand you. Yeah, correct. There's a reason that Facebook spent $19 billion targeting WhatsApp to buy WhatsApp. Do you know why? It's like, because you now have this enormous corpus of conversations, real human conversations. A couple gigantic papers have come out in the past three years that like to me, like I left the lecture being like, oh my God, like this is why am I like Facebook bought WhatsApp. One is something called once again jargony, deep learning on non-Euclidean data by Bronstein and Jan Lacune. When the non-Euclidean meaning? Graph structures. Graph structures. Yes. So what we do when we process image data is we learn these filters on these filters or just a bunch of linear operations done on what you call Euclidean data. Euclidean just means there's some distance function. So like, different strain these two pixels. Euclidean, when you talk about graphs, that didn't really like, there's not really like that distance function metric. These people solved it. Jan Lacune runs Facebook AI research, right? So the thing is, you know what else goes like a graph, our conversation. Like every time, like back and forth you and I have, like it can go in three or four different directions. You and I are just choosing different paths throughout this graph. The tree of possibility. So when I think about like these AI things and what they're scary about, oh, we'll talk about the other deep learning paper that blew my mind. But when I think about like the AI and the things that are gonna come from it, like it's gonna happen. Like, and it's gonna at least happen in a resolution that's like a lot of people aren't gonna be able to tell the difference. So if you know what's gonna happen, how do you react to it? You react to it one way realizing that very few people are gonna be able to replicate it. Great example, Amazon Go Store. I don't know if you've ever been there in Seattle. Nobody else, but Amazon has that image processing capability right now. There are a couple others that do now. We have a standard cognition office that's right in our office in San Francisco where our recording studio is and they're doing it at a pretty close level right now. That's good. And there's apparently some competitors as well, not only in the United States and even in San Francisco and stuff, but also in China as well. But anyway, yes, continue. So the idea of the non-monolithic AI is you need to start finding ways that you don't need to have a warehouse full of computers to replicate that same performance. Interesting. So you're talking about replicating performance with non-warehouseful of the data. So in this case, the monolith is the Amazon web services cloud, the Google cloud, the Facebook, like we'll process every conversation. Yeah, non-monolithic, what does that mean? Well, it means lower power. Lower power, yeah. Right? It means something that you don't need these like, a lot of this AI stuff that's been done in many ways and I think it's, I'm saying this out of respect, it's brute force. And now the deep mind work is incredible. So like, but the thing is like, if you look- Could you send a computation through potentially with the power of something like a quantum compute, just hypothetically, through a swath of data and then be able to come out so something really small could give you something really profound through a couple petabytes or whatever of data. Yeah, a couple petabytes. I think, well, so humans are amazing because we can learn from pretty much no data. Oh yeah. Well, don't we get a lot of data though? We get a lot of data in, but the thing is like, you get bit by one alligator, you're gonna stay away from alligators the rest of your life. Like on the current, like state-of-the-art neural networks, they need to get bitten by a lot of alligators. Well, that's an interesting one. So, and you look at this, like our brain requires the same amount of power as the lamp in your refrigerator. It's just a few watts. Whereas like the power that it took to beat Lisa Dahl in AlphaGo was, I don't know, thousands, like that multiple orders of magnitude. So if you scale this up, you think that like very few people can replicate a Google, Facebook. So to me, one of the most motivating things also about really understanding the brain is to start thinking about like how more people could have access to that kind of computational power. Because this is the punchline. I think it's gonna happen. The question is like, is it gonna be in the hands of a few? Or is it gonna be in the hands of anyone who can like get their hands on it? We agree, we've had lots of conversations about the substrate monopolies and the way to potentially put it in the hands of the many and democratize it for maximal creative computational purposes. Yeah, so, this is a whole nother conversation. When we do this, the Bay Area, visit next time to the recording studio or next time in Cambridge, Dan, we're gonna have you back on. We gotta keep unpacking this. Also, just as an interlocutor, just a little thought for you is that this has been one of my favorite conversations, trying to put the recency bias aside. Just because I love your energy, I love the way your cadence, I love the way that you make really profound connections across fields and you make it relatable. Your analogies and metaphors and stories are what really helps people get it easier. So, huge shout out to you. This has been such a pleasure. Thank you for coming on the show. Dude, likewise, brother. Yes, quickly, do you think we're in a simulation? Ha, no. Okay, and then what do you think's the most beautiful thing in the world? My wife. All right, all right. This has been so, so, so fun. And we'll get to unpack those answers on the next time we do the show. Huge thank you everyone for tuning in. We greatly appreciate it. We'd love to hear your thoughts in the comments below on the episode. Let us know what you think about developing the technologies to capture the complexity of biology in the brain. Share it with your friends, your families, your coworkers, online and social media. Get talking about it. Get inspiring and engaging more people to build the future. Check out Dan's links below. Also check out Simulation's links below. Support the artists and entrepreneurs and organizations across the world that you believe in. And go and build the future, everyone. Manifest your dreams into the world. Thank you so much for tuning in and we will see you soon. Peace. Brother, that was fire. I love it. Good hustle. Good hustle.