Biocomputing on human neurons (Interview)

0:04What's up friends welcome back this is the change log we feature the hackers the leaders and those building biocomputing. Yes, today we're joined by Dr. Evelyn Kurtz. She's leading the scientific research for FinalSpark on the next evolutionary leap for AI, the leap for biocomputing. This is neurons. This is serotonin. This is dopamine. This is all the things and a live thing, computing. We learn about digital processes versus bioprocessors, the role of rewarding these things with dopamine and serotonin, how they measure input and output, reading data, exotic use cases, general purpose, who's using it and why, and why it takes 10 years to get to a useful product.

0:49Of course, a massive thank you to our friends and our partners at fly.io. That is the home of changelog.com. Learn more at fly.io. Okay, let's talk about biocomputing.

1:10what's up friends i'm here with kyle galbraith co-founder and ceo of depot depot is the only build platform looking to make your builds as fast as possible but kyle this is an issue because github actions is the number one ci provider out there but not everyone's a fan explain that i think when you're thinking about github actions it's really quite jarring how you can have such a wildly popular CI provider. And yet it's lacking some of the basic functionality or tools that you need to actually be able to debug your builds or deployments. And so back in June, we essentially took a stab at that problem in particular with Depot's GitHub Action Runners.

1:53What we've observed over time is effectively GitHub Actions, when it comes to like actually debugging a build, is pretty much useless. The job logs in GitHub Actions UI is pretty much where your dreams go to die. Like they're collapsed by default. They have no resource metrics. When jobs fail, you're essentially left playing detective, like clicking each little dropdown on each step in your job to figure out like, okay, where did this actually go wrong? And so what we set out to do with our own GitHub Actions observability is essentially we built a real observability solution around GitHub Actions.

2:26Okay, so how does it work? All of the logs by default for a job that runs on a Depot GitHub Action Runner, they're uncollapsed. You can search them. You can detect if there's been out-of-memory errors. You can see all of the resource contention that was happening on the runner. So you can see your CPU metrics, your memory metrics, not just at the top-level runner level, but all the way down to the individual processes running on the machine. And so for us, this is our take on the first step forward of actually building a real observability solution. around GitHub Actions so that developers have real debugging tools to figure out what's going on in their builds.

3:03Okay, friends, you can learn more at depot.dev. Get a free trial, test it out. Instantly make your builds faster. So cool. Again, depot.dev.

3:44today we're joined by evelina curtis or kudis or curtis third time to charm a scientist turned entrepreneur with a PhD in neuroscience with 20 plus peer-reviewed papers. So you're, you're the real deal, Evelina, real deal. Thank you. You're welcome. Thank you for showing up, coming on our show and talking about neurons, neurons. This will be an interesting conversation. I'm a little bit out of my league here. I'm not going to lie. Cause I saw on your website, finalspark.com on the neuro platform page, it says instant access to human neurons. And I was like, what does that even mean? I have no idea what it means.

4:29So please demystify a little bit that, and we can dig into the science as well. So that means that our lab is available remotely. So everyone from all over the world can access our laboratory, the website browser. They can log in and they can write Python code to do experiments because everything is connected to real neurons. Because we are trying to build computers using living neurons, we want to use neurons as a processor because they are very energy efficient. So that's, that's the reason. And, um, at the moment it's still R and D of course, and you know, we don't have these computers yet.

5:16Uh, but, uh, for the moment it's possible to do experiments only to try to program neurons, but it's not a possible yet to process information like images or sounds, uh, or videos, but we hope to do this in the future. so we recently had greg osuri on the show talking about the ai energy crisis and all these ways that we can potentially power this new compute demand which is burgeoning and it sounds like those ways were hard and maybe if we figure this way out it's way better exactly how much more energy efficient is it to compute on neurons versus silicon so neurons are one million times more energy efficient.

6:02Of course, this is all estimation because we can have some idea about this by looking at human brain, which is built out of neurons, and we can have some idea what will be the processor, what will be the efficiency of the processor. Okay. So when I think of a platform where you're provided instant access to human neurons, or I just added the word human in there, maybe they're not human. No, they are human, absolutely. Human neurons. They are human. So I'm thinking about a bunch of brains floating in, you know, water or some sort of formaldehyde. No, actually, there are no brains. No brains. I told you I'm out of my league here.

6:42I'm just, everything science fiction is coming out of me here. No, sometimes you can see on social media such pictures of brains enslaved in the lab, but that's not what is happening. Okay. So we are just using the same building blocks, which are in the brain. but it's like bricks you can build a house or you can build something else so we just use these building blocks but we don't want to build brains in the lab we would like to build computers which will be totally different probably much much bigger because we imagine these neurons can have huge structures in the lab as we don't have to have make it so small as a human brain so these are only building blocks so we are not trying to reproduce a brain it would be actually very difficult and impossible at this stage, actually on science, because brain is very, very complicated.

7:33There are a lot of little structures. So we don't try to make this. We just use the living neurons and our human neurons indeed. And they are derived from the human skin. So you can reprogram the cells of the skin so that they become stem cells. And from this, you can have any cells, theoretically, any cells you want. Okay. So no brains, but human skin cells and the neurons that are in them? No. Skin cells, which later become neurons. Okay. They become neurons. They are kind of like... Adam, do you know any of this stuff? I'm over here like a... Kind of. I mean... In the soul. What we know about human anatomy and why there is so much curiosity and why she's studying neuroscience and now this you know science fiction era is just simply that our brains can compute so well with such little power requirement that's why there's the lore right 20 watts is what i read 20 watts to power the human brain right very little comparative to chai gpt or something like it right and to simulate human brain you would need a little nuclear plant.

8:52So you don't need the full cognitive brain. So here's what I understand about the brain at least. And tell me if this even maps to the science that you're doing to discover this stuff is that you've got this humanity, which is your frontal lobe. That's what helps you have rationale, reasoning, et cetera. If I don't have my frontal lobe, I'm angry at them. I'm not nice at them. I don't make good choices. I make very poor choices. how do you get to this level of compute without the full brain how are these cells able to do so much without what i would typically call like the human brain i guess well so human brain is actually for many things not only thinking it also runs all our body it controls everything so that's not always necessary for the computer what is the most interesting for us is indeed this cortex part, which is responsible for thinking, for processing some abstract information.

9:50So we are most interested in this. So we would like to, in the future, process information through the neurons. Information only. So we don't try to, for example, to control human body or stuff like this. So there are a lot of things in the brain which are not really related to the biocomputing project. So you're effectively using the neurons just for like as logic gates, like you're just doing ones and zeros at the end of the day, they're not doing. Well, yes, we would like to try to play to reproduce their logic gates. However, neurons work totally different. And that's why it's so difficult actually to build the computers, because indeed in a computer, in the computer, you have zero and ones.

10:34And this is one of the reason why actually they use so much energy. but the brain is encoding information totally differently in space and time so when we have neurons in our head it matters when and where in exact location they are active and this is information so this is totally different type of encoding so no zero ones actually but there are a lot of ways how we can look at the activity of the brain for example how often you have spikes or what are the time in between the spikes. So this electrical activity of the neurons. So we know for sure it's totally different. And that's why this project is so difficult because we have to learn totally.

11:19We have to figure out totally new way of programming, totally new approach. It's the same actually as in quantum computing. It's the same situation that you have totally different hardware which is working differently. So it's necessary to figure out new way of writing algorithm. And that's why it's so difficult because actually someone has to come up with some idea which would be totally different. But indeed, at the moment, when we do research on living neurons or sometimes on some simulations of neurons in silico, people usually try to follow the rules of digital computers, like try to reproduce logic gates, which is really actually not really correct, but that's the best way what we can do at the moment.

12:09Let me see if I understand this. I'm grokking some of the stuff from what you're sharing and then also from your very awesome website, final spark.com. It says they transform stem cells into mini brains that learn and adapt growing neurons in an orbital shaker, which I have no idea what that is. Sounds so cool. Sounds cool. Yeah. Over a three month period. And these many brains, organoids, not sure if that's a term y 'all came up with or not, but that sounds cool too. Sounds like I mentioned it. 0.5 millimeters in size with about 10 ,000 neurons that function as real brain tissue. So you've found a way to take stem cells, grow them over a three month period.

12:50They used to have a half-life that was even shorter, like a few hours now that you can actually live 100 days. and they get connected to this neuro platform with 24 seven access. So essentially the same way we treat a CPU in AWS, you're doing with stem cells, turn neurons, turn mini brains, turn organoids that can be compute platform. Yes, absolutely. Although now it's for experiments, so we cannot really process information the same way as in digital, but yes, it's available remotely. And we imagine that actually in the future, our lab or our biocomputer will be available remotely as a cloud service today.

13:34Right. Not quite AWS yet, but working your way there. Yes, absolutely. How did you get involved in this? Where are you coming from? So I come from Poland. I was always on the medical side, let's say. I studied pharmacy and biotechnology, and I always wanted to be a scientist. I enjoy a lot working in the lab. I was very fascinated, you know, by cracking my brain, you know, teasing my brain with some ideas and challenges. So I always wanted to be a scientist. And I realized at some point that brain is the most interesting part to study. So I did a PhD in neuroscience. I was working on brain imaging.

14:15and later when I moved to industry because I also wanted always wanted also to see what is outside academia outside this academic world I started to work with startups on actually initially on imaging and then there is a lot of AI in the imaging medical imaging in industry so this is why this is how I learned about AI and I become fascinated by that and I started to discover which opportunities it brings beyond imaging. So I started to work on the commercial applications of artificial intelligence. And after I started to work on Next Frontier of AI, which is actually closely related to neuroscience.

15:00So on biocomputers. Can you talk about the imaging? I think you mean when you say imaging, you're probably referring to like MRIs, like brain scans. Is that right? Yes. So actually I did my research on a positron emission tomography. So this is something what you do using radioactivity. You put some radioactive substance in the body and the substance goes to some specific places in the body and you can detect this non-invasively. So you can get a picture, for example, of the brain, which parts are active, for example, or you can visualize some receptors. receptors without opening the brain. But it's actually similar to MRI.

15:44MRI is just a little bit different. So you don't use the radioactivity and you can see a little bit different things. But the idea is always the same, to look inside without opening the body. Right. Yeah, the MRIs are a little different. Which one's more accurate? Is the imaging or the MRI more accurate? It doesn't matter. Always imaging. so I think it I'm not sure it's good comparison because I think it really depends on the protocol because there are different types of MRI and also there are different types of PET so it really depends on the parameters and also they measure different things because PET is always functional so there is always some when you use the radioactivity there is always some chemical substance like even glucose everything is actually chemical substance every everything what is flowing in the body so you always observe some process biological process and in mri it's not always like this sometimes you just observe the tissue and you know when when you have different tissues and it's static and it's not always functional it's not always observing this imaging that's why i asked this question because this imaging is really kind of like the rage i would say and my version of the rage may be way different than your uh scientific version of the rage but what i mean by that is that there's a lot of study around uh mental health uh adhd add you know trauma you name it that folks are trying to image brains in these scenarios is that kind of what got you into this curiosity of like how the brain operates from different trauma levels or different, you know, prescriptions or descriptions of health concerns, mental health concerns, whatever it might be.

17:37Is that what got you interested in this imaging process to understand more clearly how the brain reacts to, I suppose, life? Well, actually I was working on something a bit different to what you are talking about, because you say about different activities of the brain during different maybe diseases or maybe different tasks, cognitive tasks. But I was actually working more on inflammation. So I try to visualize microglia. Microglia are a type of cells which are around neurons in the brain. So they actually take care of the neurons. And sometimes they become very activated, which means inflammation.

18:13And it's believed that this process is actually involved in many neurodegenerative disease and also depression. So when you have an inflamed brain that you can develop some disease. Yeah. Alzheimer's. That's interesting. Yeah. Inflammation is like the number one issue for most people. Yes. Right. Inflamed everywhere. My research was actually about the effect of nutrition on inflammation. And it's funny because I started to put attention on everything or what I eat after I started to do this research. It's really interesting because I started to look totally different on my groceries because actually diet can be pro-inflammatory or anti-inflammatory.

18:56It's very important what you eat and it can affect also your brain health. And I think now after, you know, it's maybe almost 10 years since I did the studies. Now I see that there is more and more, you know, more and more people are talking about this, about anti-inflammatory diet, about how much is important, what you eat for also your brain health. How did that lead you into discovering? I mean, it kind of seems obvious, but how did that lead into AI and your discovery there? Were you leveraging, you know, trained models, express how you get cured with AI? No, actually, I started my first job in industry.

19:35I started in the company, which was doing medical imaging. And actually, it was a very good start because there was at least one thing which I understood at the time because I had absolutely no idea about how companies work and, you know, anything about this industry world. So there was at least one topic which I understood well, which was medical imaging. And this company was doing a service of analyzing images from different medical studies. And they talked a lot about AI because when you have imaging data, you know, when you have a high number of imaging data, you can analyze them automatically.

20:13In some way, you can use AI for that. And actually that's the way how I learned about artificial intelligence. Because actually when you go for different industry events, you see people talking constantly about AI. And I was very lucky also because at that time I was in London. So this is a very good place for learning new stuff and for networking. So I could get a lot of exposure and to see what people are doing with AI. And I could discover that there is much more beyond imaging. So that's why I was I started to be interested with anything, what you can do with AI. Well, being able to scan a lot more, no pun intended really, but just grasp a lot more of these imagings that you're doing to see the anomalies and see the connection points that you can't really see individually.

20:59I mean, that totally maps to me because the more you can see across different scans is good. Yeah, this general thing about AI because it can see much more than us and can scan a lot in a very short time. So how long have you been working on this problem? On the final spark, I met the founders in 2019 at the conference in London. So I started to work with them initially about some other projects. So actually on final spark, I could say I'm working like three years around. Okay. And there's a platform right now for experiments. You're hoping to get to compute down the road and a service for that.

21:39Is there a straightforward path towards that? or are there like breakthroughs that still need to happen to get from where you are right now to where you guys want to go? No, it's very difficult, very challenging project. That's why we expected to build this real computers in around 10 years. Okay. So it's a bit challenge when we talk with investors potential because it's quite long-term project and it's very, very difficult because nobody knows how really neurons and code information. So this is the biggest challenge. So we know that neurons are active electrically, you know, they are spiking.

22:14Spike means that, you know, there is electrical activity. And we know quite a lot about this, how it happens. However, we cannot really translate this into some specific information. So, for example, you have text or image. Maybe about imaging, there is some understanding already in neuroscience. but you know when you have for example words text it's hard to say how how some word can translate to specific activity of neuron so at the moment as i as i said many people you know do random a lot of random experiments also us we do a lot of trial and error so this is why actually we build automated laboratory initially was the idea was to just be able to do as many experiments as we as we can.

23:02And also a lot of research on neurons are often inspired by what happens in the digital world, which is not really correct because neurons are working totally differently, but it's still at the moment the best you can do. So this is the biggest challenge that we don't really know what activity of neurons mean. And also another thing very important is that brain or neurons, any kind of form of also our neurons in the lab, they are not stable systems. So a computer, you can consider as a stable system. It's a dead matter. So it works today in some way. Tomorrow will work the same way. But the living tissue is not like this.

23:49It can change. So the dynamic inside can change. So for example, today we do some experiments. we send electrical signals to neurons and they can react in one way and tomorrow they can react to the same signal totally differently. So that's also a big challenge. The fact that that living matter is plastic. So it changed behavior actually also like us. And our brains, we also change during time. We can be completely different people. Can you talk about how you get them to compute? This is actually a challenge. So what we do, we try to send them electrical signals because neurons are placed on the electrodes.

24:32You can see this on our website, finalspark.com. There is section live. You can see the readout. So we send them electrical signals and we measure how they respond. So how they change the activity, how they change electrical activity. Another thing how we try to compute neurons is also by sending them some chemical signals. At the moment, we can send them dopamine or serotonin. So programmatically, we can program in Python that neurons will get dopamine at some point. So this is why. So at the moment, this programming is not really to do some specific task as with computers, but actually to change the behavior of neurons.

25:16So this is actually the first step. So we want to be able to consistently control how neurons behave. So how is the electrical activity of neurons? So you may have one lazy organoid and one very non-lazy organoid. Yes, absolutely. It is biological tissue sometimes can vary. Sometimes they can just die. So, you know, we are still learning. And yes, so there is also variability. Uh, yes, absolutely. A lot of things can be also lazy organoid. Yeah. Or depending on the day, like yesterday it was really productive and then today it's lazy. Yes. As I said, you know, every day is not, is dynamic. It's not a stable system.

26:02Uh, however, we have some success. It's not only so bad. Uh, we were able to store one bit of information. So just to give you an idea about the stage at which we are, uh, we, we store one bit of information in neurons. that was quite consistent and we were able to reproduce this many times. So we are happy. There is some kind of progress, but yes, it's very challenging to get something. So when you store a bit of information, how do you read it back out again or how do you get it back or how do you know that it's stored? Yes. So actually that's quite technical because I can tell you every blob of cells, because there are such a blobs, these neurospheres, organoids, they are 3D structures and each is placed on the eight electrodes.

26:52And all these eight electrodes, they measure activity from neurons. And depending on how strong activity is at each electrode, you can mathematically calculate something what is called center of activity. So this is quite a hard science approach and we were able to shift the center of activity. So yes, no, that was one bit of information. And this is quite complex as you can see. But yes, we were able to have this consistently, this kind of results. Consistently across different neurons and across different times. Yes, and different days because that's always in general in bio science, in every time when you work with biological tissue, is actually this is important that you have to be able to repeat because many things work once or twice, but it's important to be able to repeat your results.

27:53Yes, on different days, on different neurons. And is the process slow? Well, I cannot, to be honest, I cannot tell you how much time it took. No, actually, I don't know. But I guess it's in seconds or milliseconds. But I don't know. But generally, I can say that in general, neurons are slow. And in general, neurons will be good for tasks which don't have to be fast. Because also when we look at the human brain and we look at computers, we can see that computers are very good in speed, in doing repetitive things very, very fast. and we will never be able to compete with the digital on that. However, brain is better in complex tasks because we can solve complex problems using very little energy.

28:41So that's where is our strength. But definitely speed or, for example, memory is not something where neurons are better because also when we look at our brain, they are very limited. Actually, you know, computer can remember 20 books very easily and for us would be difficult. to remember every word in 20 hours. But the computer is easy. It's easy. What does this wetware look like? Like I'm programming my Python experiment. Yes. And I'm sending it over the internet, I suppose, or some sort of VPN connection to you guys. Absolutely. And so, of course, it's going over copper wires and Wi-Fi and whatever, backbones, and then back into your interface, which eventually translates it into, I'm imagining there's like a needle at the end of a thing that like sprays some dopamine.

29:30I don't know what happens. Like what happens at the end, the last mile of this API? Yes. So when you send electrical signal, you have a digital to analog converter. So you have to translate the things from digital word to analog word because neurons are analog. So you have this digital to analog converter and it is translated into electrical signals, which goes through the electrodes. So basically the stream of electrons are going, flowing through the electrodes. And when you want to send a signal with dopamine or serotonin, then it's connected to the lamp. So we have UV lamp. And when the UV light is open, the dopamine is released because dopamine is closed chemically.

30:21It's called, it's encaged. in the chemical so that it's not active, but when it sees the UV light, then it's released. So it's a way of releasing very, very quickly dopamine to the neurons. So this is how it works. So then it's connected to some controllers, which are connected to the lamp and then the lamp switch on. Okay. Same thing for serotonin? Yes. But for serotonin, we have different wavelengths. I don't remember which one, It's not UV, but then we have different wavelengths. Yes. So basically, so that they don't overlap. So you can have both in the medium. And we, of course, we plan to have more of this and, of course, much more neurotransmitters.