#568 Beyond Silicon: Building the First Living Computer with Ewelina Kurtys


 [00:00:00] 
 Mehmet: Hello and welcome back to a new opposite of the CTO Show with MeMed today. I'm very pleased. Joining me, Ewelina Kurtys Ewelina. She is the strategic advisor at FinalSpark, a company that's doing very interesting stuff, [00:01:00] building next generation bio computers. Of course what we mean by this, we're gonna discuss all this, but before too deep into the deep dive into the topic. 
 As I do always with all my guests. Uh, tell us a little more about you, your background, you know, your journey and what brought you to, to where you are today. So the floor is yours. 
 Ewelina: Uh, so I'm Kurtys. Uh, I come from Poland, uh, and uh, I lived in. Few countries in Europe. Now I live in the uk. Uh, I'm neuroscientist by background, but after some time I left academia and I started to work with startups or, uh, on artificial intelligence. 
 And currently I work with, uh, FinalSpark as strategic advisor. And FinalSpark is building new type, uh, new, new type of computers, uh, living computers, uh, from living Neuros. 
 Mehmet: Great. And thank you again Evelyn to, to be with me here today. You know, even when you reached out to me and I started to read about, um, the company and what you're [00:02:00] trying to do, the first thing that came to my mind, and I think of course, like I come from technical background, I can understand these things, but for most people, maybe when they hear computers made of living neurons, so their mind will go like, this is science fiction. 
 Like. What is that? So what is you, you think is the common misunderstanding about, you know, what you're currently doing and how do you define this space? Uh, Evelyn. 
 Ewelina: Uh, well, uh, we are trying to build computers, uh, from living neurons. Uh, I think it's quite simple topic, uh, you know, to understand on the high level. 
 Uh, we are using neurons, uh, the same as we have in our brains, and we know that this neurons are very efficient in processing information. Uh, there are 1 million time more, uh, efficient. Uh, then, um, uh, then, uh, digital computers, uh, which we have, uh, today. So, um, so we [00:03:00] want to solve this problem of AI scalability. 
 Uh, we want to be able to provide artificial intelligence, which will be more accessible, uh, because it'll have lower price due to less energy. Uh, uh, less energy consumption. 
 Mehmet: Great. Uh, so, so FinalSpark. Originally, the company didn't start, as you mentioned, like as a biology first company, but you know, like as hardcore AI and math driven, uh, you know, you, you, you try to do this. 
 So at what point, you know, the team realized that, uh, you know, it should not be only software, but you know, also biology. 
 Ewelina: Uh, we realized this, uh, when FinalSpark was founded, uh, it was founded with a dream to build thinking machine. Uh, so, uh, FinalSpark did, uh, some fundamental research in digital artificial intelligence for a few years, and then the founders realized that it's too [00:04:00] expensive. 
 Uh, to do this. Uh, and if you are a small startup, you want to make a difference in ai, you have to be more innovative. You have to figure out something new, something different. So they decided, uh, to try living neurons because they are the most efficient known need for processing information. So this is how it started. 
 Mehmet: Eil, can I say, you know, the company is at intersection of AI, hardware and biology, or you, you would define it as something different? 
 Ewelina: No, no. I fully agree with you. 
 Mehmet: It's an intersection of D three, right? 
 Ewelina: Yes. 
 Mehmet: Cool. Now I gotta ask you a question. And again, you know, I don't claim myself to be the, you know, the, the scientist by any means. 
 But I like science. I like to read also about history of things and also sometimes, you know. Coming as they call it, you know, in English, the rationale of, of doing things. So when we think, because you, you [00:05:00] mentioned the word cheaper multiple times, right? Mm-hmm. Um, and of course efficiency. Now when we think about the traditional way and just for the people who are not that tech savvy or they don't know how CHIPS works or just some explaining, and then I'm ask you AvEwelina. 
 So, so, so chips all make. Not made from silicon, which is, you know, like the material which exists in our world, and that's why they call it the Silicon Valley. Anyway. So the silicon chips keep getting faster, right? 
 Ewelina: Mm-hmm. 
 Mehmet: Um, and every time we see like a new chip that is getting like more transistors inside, they're able to fit more power into it in you. 
 From your perspective, let me ask you this way, why you think still there is actually a better way of doing this, especially now in the age of ai, where we need the more efficiency we need, like especially from power perspective and compute perspective, we know that all these large language [00:06:00] models, you know, they rely, they, they are like, uh, CPU, hungry as we call them, or compute hungry, so. 
 To simplify it, why you think what you do is better than keeping with what we have been doing for like more than 60 years, I would say now, you know, trash just relying on silicon. 
 Ewelina: Well, as mentioned, uh, living neuros are 1 million time more energy efficient, uh, than silicon. So, um, we want to be able to provide artificial intelligence, which will be much more affordable. 
 It'll be much cheaper. And, uh. Today, maybe you don't feel it, uh, how AI is expensive, uh, because for example, open AI is losing money. Uh, so that means the price of, uh, uh, large language models like, uh, chat GPT are lower than they should be in reality to cover the cost. Uh, currently open ai, uh, lost, uh. Uh, 19 billion, uh, dollars. 
 So, uh, of course this is because, uh, this [00:07:00] is a phase of adoption, but we can expect that in the future, uh, this price will go up. So AI is actually quite expensive, and, uh, when we use it more and more, uh, this, uh, this, uh, problem of price can be more and more painful. So that's why, uh, we think that using more, uh, energy efficient hardware will solve this problem. 
 Mehmet: Great. Now let me ask you little bit, kind of a technical question. Mm-hmm. Um, Ewelina for you now, when you say programming neurons. 
 Ewelina: Mm-hmm. 
 Mehmet: Right? So what do we mean? Like, of course what, when we think about writing codes, right? We, we think about maybe training a system. So in that context, like what exactly we mean by programming neurons? 
 How does, how does it work? 
 Ewelina: So neurons, uh, they process information through spikes, uh, and generally through chemical and, uh, uh, electrical signals. And that [00:08:00] means that in the lab you can put them on electrodes and you can send them electrical signals, and you can also measure the response. In form of electrical signals. 
 So currently, uh, we put neurons, uh, like that. You can see this, uh, live on our website, finance, uh, uh, spark.com. There is section live. So you can see this in reality, how it looks and, uh. What, uh, what you can do is, uh, you send electrical signals. So this is like an input and you also measure the, uh, output the answer from neurons. 
 So that's basically the programming, kind of the beginning of programming, uh, because you send them some signals, uh, you measure the response and you want that. There is logic relationship between this input and output. 
 Mehmet: Right? Uh, also when I checked the website, I found out that, uh. You know, FinalSparks, what you call the neuro platform is mm-hmm. 
 Described as a first remote access biocomputing lab in the world. So, does it [00:09:00] mean that any scientist or anyone, of course you're gonna give random access, but I mean, how easy is it, uh, to, to access? Um. The, the platform. 
 Ewelina: Yeah. It's very easy. You can access this, uh, via internet browser. Uh, so actually it doesn't matter. 
 Uh, you can, if you do, uh, research in our lab, it doesn't matter if you're sitting in Switzerland, in our lab or anywhere in the world because, uh, every experiment we do is basically writing code in Python. Mm-hmm. Program language and everything is executed automatically. Uh, so that means that, uh, people from anywhere can do this. 
 Mehmet: That's great. Now, uh, in addition to this, and again, maybe now the audience will not like what I'm doing, but I'm curious because also I, I read the document that you shared with me before and mm-hmm. I figured out that it's not only about, you know, the, um. The, the electrical signals that you just mentioned to us, but also like [00:10:00] there are like some chemical signals, like, I mean, I, I, I'm, and excuse me if I'm using wrong terms, correct me if I'm wrong, Alina. 
 So you're using like, you know, injecting or interacting using dopamine and, uh, sterin and uh, uh, other chemical stimulations. Um. What's the reason for that? Like is it, again, to be more efficient? Is it like, uh, to, to, to help in the training? What's that exactly, uh, used for? 
 Ewelina: Yes. As mentioned before, uh, neurons are processing information, uh, via chemical and electrical signals. 
 Uh, so yes. We also doche, uh, use chemical signals in, uh, in this case, uh, dopamine most often, as you mentioned. And why we do this, to reinforce the learning. Yes, because dopamine is believed, uh, to take part in learning and reinforcing, I think the connections between neuros, so reinforcing the behavior. 
 Mehmet: Quick question. These, these neurons, [00:11:00] how do you solve them? What they are? 
 Ewelina: What do you mean? 
 Mehmet: Like be, because you know, if I think, again, in the traditional way to, to have a, to have a processor, a microprocessor, I need to source some silicon. I mean, from manufacturing perspective, I'm, I'm just thinking out loud. 
 So like, where do we get these steroids from? 
 Ewelina: Oh, okay. Uh, then, uh, we get them from stem cells. You can buy commercially stem cells, which are derived from human skin. So basically are transformed into, uh, stem cells and from the stem cells you can theoretically get any cells want, including neurons. 
 Mehmet: How much is the lifespan of, of these, like for how long They can, they can usually. 
 Ewelina: Uh, depends. They can live for months, many, many months. But, uh, on the electrodes they can usually live around three months only, uh, because it's not a natural environment and we still have to find solution, how to make [00:12:00] them live longer on the electrodes and, uh. Uh, we had some success of seven months, but this was really outlier. 
 Usually it's around three months. 
 Mehmet: And again, excuse my ignorance, Sina, because really I'm getting more curious about this. So you have the stem cells, the lifespan, let's say it's between one to three months, uh, on average. 
 Ewelina: Neuros. Neuros, I talk about less one or neurons, yes. Mm-hmm. 
 Mehmet: Or neuros. Yeah. Again, you need to correct me a lot. 
 I'm not, you know, that savvy on this. So when, how, so do you take this, do you store it somewhere until you get the new neurons there, like meanwhile, or is it like. Is it like a kind of a continuation? So you always have the reserve of, of living neurons that would keep the system alive. I'm just trying to understand how you keep the lights on. 
 Yes. Uh, knowing if, if you can explain that to us, 
 Ewelina: [00:13:00] yes, you can freeze the cells, uh, so that you stop any activity, uh, from them, but you can always, uh, uh, bring them back to life. Uh, and you can also just give them a medium. Uh, alive at 37 degrees and, uh, you can keep them for a very long time like that. 
 Mehmet: How do you see the use in. 
 For the mass, I mean, to, to commercialize this project. Uh, do you see that as a replacement for GPUs or like more like a replacement for co-processor? Um, any specific areas? You talked about ai, you talked about LLMs and how, you know, big companies like Open AI and others, they're losing money because again, you know, what they're doing is. 
 We, we are repeating this, but if I want to take what you're trying to do and put it into a little bit, kind of a commercialization [00:14:00] perspective and real life usages, where do you do, do you see this or if you have already that implemented somewhere? 
 Ewelina: Uh, so no, it doesn't work yet. Uh, it's still in the, uh, stage of preparation. 
 Mm-hmm. Uh, we're working on prototypes and, uh, what we want to do is to replace the processor. Silicone processor with living neurons processor. So basically, uh, you will have computer where instead of silicone you have living neurons. And we believe that the best application for that will be generative ai. 
 Uh, because of the nature of this, uh, processing, we be believe that, uh, neurons will be doing this much more efficient and then will, they will be much more suitable for this kind of tasks. 
 Mehmet: Lemme ask you kind of a science fiction thing here, Valina. And this question that just came to my mind, I didn't prepare it really. 
 Um, do you see this kind of also. Would call [00:15:00] it portable technology because now when, when we think about ai, and, you know, of course my audience would know they heard it everywhere or read about it, that, you know, companies are rushing to build data centers and they are building like massive, uh, data centers to, to cater all the need so. 
 If, if, if I think about what you're doing and, you know, the, uh, microscale or even nanoscale of, of the neurons and how small they are, can I think that maybe one day I would have all the capacity computing capacity of a data center maybe in. More small form factor that actually I can port it with me anywhere in the world or this is too much science fiction for now. 
 And that doesn't apply to what you're doing? 
 Ewelina: No, I think, uh, I think that I'm not, no, I don't think that processor of living neurons will be smaller. It's not here about the size, it's more about energy efficiency. [00:16:00] And I personally don't believe in portable equipment at this stage because, uh, neurons they need very specific conditions. 
 So we believe more in central bio server, some central computer, uh, where you will be able to connect like today to the cloud and, uh, that this will be the solution that you will be, have this, uh, remote access. 
 Mehmet: The reason I ask you VEwelina, because now be, you know, when we talk about power here, we don't talk only about our, the energy efficiency of the chip itself, or, you know, the, the components inside the computers, but also everything around. 
 So data centers generate a lot of, you know, heat. Mm-hmm. Because, you know, there's an emission out there. And even we started to see some of the 
 Ewelina: mm-hmm. 
 Mehmet: Uh, you know, known figures. Talking about like, yeah, let's put the data centers in, in space, like, um, like. With that, if, you know, once we [00:17:00] have the proof of concepts for this would make this idea obsolete. 
 Because at that time we don't need to have actually, you know, to, to think about building data centers. Actually in the cloud, quote unquote in the space. And, you know, and will defy, you know, all these theories that they are saying, yeah, if we need the best, you know, we don't need to have the global, uh, warming, uh, effects of the data center. 
 So let's put them in space when it comes to living neuron scan this be kind of a good alternative from a climate, uh, uh, preservation perspective. Do you think? 
 Ewelina: Uh, yes, of course. If uh, they use 1 million time less energy, then obviously they will, uh, be producing less. CO2. Yes, you are right. 
 Mehmet: Yeah, so, so do, do, again, I'm asking maybe to you it would be silly questions, but do these stem cells, like, can they also be put in, in space? 
 I don't think there will be a need in that case to put things in space and, uh, you know, [00:18:00] you can have the neurons here on, on, next to us anywhere. Right. 
 Ewelina: I think it, it doesn't make sense to compare this, you know, the problems with the silicone is totally different. Yes, you have with living neurons, so it doesn't make sense just, uh, to compare this 
 Mehmet: right now, you gave free access to selected universities worldwide. 
 Mm-hmm. Uh, and this is part of, you know, kind of first chairing the technology, getting I'm sure like it's getting feedback. Um, but from strategy perspective. Why was open innovation more important than like locking everything behind intellectual property, like ip? Um, but, and I know that at the same time, like you have, um. 
 We have selected like not many, like out of dozens, university, only handful were selected. So why did you take this approach and [00:19:00] you didn't kind of put a, uh, um, a, a, a paywall behind what, what you're doing? 
 Ewelina: What do you mean? Uh, why pay paywall? What do you mean? 
 Mehmet: So, so you, you, you kept it like an open, open ity. 
 So you gave free access to universities. Like you didn't, you didn't charge them. Like, was there a reason for this and, you know, any, anything that might affect the intellectual property, the ip, uh, moving forward, because you've shared the, the technology with other universities. 
 Ewelina: Well, our technology is pro, uh, protected with patents. 
 Uh mm-hmm. So that's another story. Uh, and we gave, uh, access to few universities. That's more, uh, was like, uh, early adopters. We wanted to see if people are interested and what they could do with this. And the objective was to give them access, uh, at the condition that they can publish. Uh, some scientific, uh, paper. 
 So that was our interest, that they publish a paper and show to the world what can be done, because we focus more on patents, not so [00:20:00] much papers. 
 Mehmet: Mm-hmm. 
 Ewelina: So, uh, so that was the rational, but now we have a lot of actually users who pay us to get the access. Uh, and we saw that our lab is, uh, getting interest. 
 Mehmet: Evelyn, did you face, uh, as company? Of course, any. Concerns from people talking about, yeah, hey, like, this is living neurons. And, uh, you know, we have some ethical concerns about it because like, you're taking, you know, you, you're getting the stem cells and all these things. So were there any red lines especially, you know, when I talk to people in, um, Biocomputing, they tell me sometimes they're sometimes, of course, ethical and. 
 Also some regulations. So if you can just, uh, you know, explain this to, to the audience from, from your point of view also as well. 
 Ewelina: Uh, yes. People ask us question, uh, [00:21:00] sometimes, for example, how we get the living neurons or, uh, you know, why, why we use them. Of course, it, uh, it, it creates a lot of questions and it's normal thing. 
 So that's why we also do communication a lot. We try to talk to press to podcasters, to familiarize people with this, uh, topic so that they don't have to be afraid of this, uh, because we usually are afraid of things which are not known and not understood. 
 Mehmet: Got you. Now, there will be always Alina, some people who we call them the skeptics, right? 
 So who always like to doubt they, um, and they will argue that, you know, anything related to biological computers gonna stay in the labs. If we want to counter, uh, argue these folks, uh mm-hmm. You know, what, what you, you would tell them. 
 Ewelina: Uh, you know, I, I [00:22:00] think, uh, it's okay that not everyone has to like what we do. Sure. And, uh, I don't try to convince anyone about anything. We just like to answer questions and, you know, spread the message about the science. But, uh, if someone doesn't, it's not convinced. That's fine. 
 Mehmet: Do you think there will be a moment where, when, when, when this succeeds, right. 
 And, and you have like e early signs that there is, um, there is a way for this to, to, to succeed mm-hmm. At that time, and maybe it's kind of a philosophical questions as well. 
 Ewelina: Mm-hmm. 
 Mehmet: When all this succeeds, right? Mm-hmm. Um. And when we talk about like different use cases, and you mentioned ai, right? But at that time, you know, I don't know what's the timeframe. May maybe you can tell us, but by that time, do we have to actually redefine the whole AI thing? 
 Because when I think now about it, we're talking about [00:23:00] neurons, right? We talk about something biological, it's not artificial anymore. Do we have to redefine? Maybe it's a little bit philosophical question. Um, am I right about the way I'm thinking? Because it's not artificial anymore. It's like bio intelligence. 
 I don't know what, what, which should you call it by that type. 
 Ewelina: I think that, uh, this is question for philosophers that I would have to decide. Um, we are just scientists and, uh, engineers, so, uh, we actually reach out to many philosophers, uh, to encourage them to work on this topic. And we hope that they help us to answer this kind of, uh, philosophical questions. 
 So I, I really don't know. 
 Mehmet: Okay, that's fine. Completely understood. Mm-hmm. Now, if, if we try to compare. I know it's completely different things, but the attempts of putting, um, chips within humans, we, we, we know like what, uh, uh, Neuralink of, uh, uh, um, that the company, um, 
 Ewelina: yes. 
 Mehmet: Added [00:24:00] by Elon Musk is trying to do, if I want to take what you're trying to do, and can we also imagine that we can transplant? 
 Quote unquote, again, intelligence within the human body. Is this a use case, which is possible in, in, in, in, in the future in your, uh, from scientific point of view? Of course. 
 Ewelina: Yes, definitely. I think it'll be possible in the future to, uh, input some knowledge into the human brain. And our work will, uh, for sure, uh, contribute to this. 
 Although it's not our objective, but, but if we learn how to program neuro in vitro. Uh, we might be able also to affect, uh, human neurons, uh, in the brain. Uh, this will not be done by FinalSpark, but I'm sure others can, can take this knowledge and use it. 
 Mehmet: So there's a lot of branches. I can see EvEwelina from scientific perspective that they can branch out from, from what you're trying to [00:25:00] do. 
 So, and I can imagine kind of, um, hybrid maybe architectures also as well between the traditional compute and uh, uh, bio compute at the same time, which is, as I, as we started, it looks like, uh, um. Science fiction, but it's not because actually you have the prototype for it. Um, you, you mentioned that you, you, you're still in the early phases, so, um. 
 I know like no one can definitely, especially when you're doing all these scientific research, you cannot get the definitive answer. But are we talking here if this succeeds when we could expect it to be mainstream? Like is it like five years, 10 years? Of course you don't need to give me a specific answer. 
 I know it depends, but in, in your mind as strategic advisor, like if you want to put a deadline like by this date, like this is either we prove it works or. It becomes mainstream. Is there a kind of timeframe, uh, in, in your mind, [00:26:00] Ewelina or maybe for the company? 
 Ewelina: Uh, yes, we do have a timeline. Uh, we assume that we will be ready with the computer in 10 years. 
 That's 10 
 Mehmet: years. Okay. 
 Ewelina: Yes. 
 Mehmet: Okay, 
 Ewelina: so that will also, maybe I mention also that will a big part depend also on the investment because co currently we are self-funded by the founders. Uh, but we are also talking to external investors seeking 50 million, uh, Swiss Francs, external investment. And uh, with this money we can accelerate our research. 
 So that's very for us. Yes. 
 Mehmet: Great. Now I gonna ask you also a question, Evelyn, for you personally, right? On a personal level, when, when you think about this project, um, uh. Of course, other than the scientific aspect, aspect of it, does it make you feel like more responsible? Does it make you feel like, uh, uh, like, uh, some, some discomfort? 
 Because, you know, if, if something goes wrong, [00:27:00] uh, uh, you know, with this, uh, or like as a scientist and engineer, you know, it's just mm-hmm. We need to test and then see the results. And based on it, we, we, um. We, we decide how, how to move forward. So personally, how do you feel about this project? 
 Ewelina: Well, I feel that this is fascinating. 
 Uh, science, really exciting. We are a very small team, uh, and I think it's nice, uh, very nice piece of work. We are doing 
 Mehmet: great. What do you tell people who are like. Studying now in the university, uh, or maybe they are like just fresh scientists or engineers listening to us today. Uh, and you want to draw their attention to work in, in, in this field, in Biocomputing. 
 So what, what do you share with them? 
 Ewelina: I think if someone is interested in Biocomputing, uh, good idea is to check our website. We have a lot of resources. [00:28:00] On many different levels . And it's also good, uh, to check our paper in Frontiers Only one we published. And that's a good introduction on which kind of science do you need to work on Biocomputing. 
 Mehmet: That's great. Um. Final, you know, as we are almost coming to, to the end, uh, Lina, if, if you want to, you know, tell us where people can get in touch and whether they are just curious or where maybe they want to get access to what you do, where they can find more information and get in touch. 
 Ewelina: Yes. I think our website is the best. 
 Uh, you can also send us direct message from the website and we answer every email. Mm-hmm. 
 Mehmet: Great. 
 Ewelina: SOCOM is the go to go to place. 
 Mehmet: It's a go-to place. Great. Um, well, you know, that was great, Aina. Thank you very much for sharing. Uh, you know, your insights today, again, I put the website in, in the show notes so people who are listening, uh, or watching us, uh, they can, uh, reach out and get in touch. 
 I really appreciate, it's a really interesting topic. Uh, interesting. And it's, it's the [00:29:00] first time we discuss Biocomputing on the show. And I'm happy that, uh, you know, I'm, I'm hoping that I'm the first one even here in, in the region that we are talking about this, uh, very advanced technology, which is looks like science fiction, but from mm-hmm. 
 What I've checked also already, like there are like. Promising results. Uh, and you know, this is for the audience. This is, uh, how I end my episodes. If you just discovered this podcast, thank you for passing by. If you enjoyed it. Thank you. Give me, uh, you know, uh, a, a, uh, thumb up and share with your friends and colleagues, and if you are one of the people who keeps coming again and again, thank you for following us. 
 Thank you for your support. I really appreciate all that you do to the show. And as I say, always stay tuned for any episode very soon. Thank you. Bye-bye. 
 Ewelina: Thank you.