Subscribe
Share
Share
Embed
https://youtu.be/kAYdIpcDnZk
Today is a very brainy conversation. Matt had the chance to chat with Dr. Kyle Wedgwood and research intern Wiktor Wiejak from University of Exeter in England. Wedgwood’s a mathematician specializing in neuroscience, and he’s using a computer from FinalSpark. Matt just put out a full video on what’s special about these computers, but in short, they’re alive. They’re made up of bundles of neurons, also known as organoids, and Dr. Wedgwood is using them to explore what he calls “the fundamentals” of how neurons work.
Watch the Undecided with Matt Ferrell episode, Merging Humans and AI: The Rise of Biological Computers https://youtu.be/DfUkaE7HcnU?list=PLnTSM-ORSgi7uzySCXq8VXhodHB5B5OiQ
Today is a very brainy conversation. Matt had the chance to chat with Dr. Kyle Wedgwood and research intern Wiktor Wiejak from University of Exeter in England. Wedgwood’s a mathematician specializing in neuroscience, and he’s using a computer from FinalSpark. Matt just put out a full video on what’s special about these computers, but in short, they’re alive. They’re made up of bundles of neurons, also known as organoids, and Dr. Wedgwood is using them to explore what he calls “the fundamentals” of how neurons work.
Watch the Undecided with Matt Ferrell episode, Merging Humans and AI: The Rise of Biological Computers https://youtu.be/DfUkaE7HcnU?list=PLnTSM-ORSgi7uzySCXq8VXhodHB5B5OiQ
Follow us on X: @stilltbdfm @byseanferrell @mattferrell or @undecidedmf
Undecided with Matt Ferrell: https://www.youtube.com/undecidedmf
★ Support this podcast ★
Creators and Guests
What is Still To Be Determined?
Join Matt Ferrell from the YouTube Channel, Undecided, and his brother Sean Ferrell as they discuss electric vehicles, renewable energy, smart technologies, and how they impact our lives. Still TBD continues the conversation from the Undecided YouTube channel.
Today on Still to be Determined, we're talking about things that make computers go hmm. And I mean that literally, I think. Anyway. Hello everybody. I'm Sean Ferrell. I'm a writer. I write some sci-fi. I write some stuff for kids and I'm just generally curious about technology and luckily for me, my brother, is that Matt!
Yes. That Matt from Undecided with Matt Ferrell, which takes a look at emerging tech and its impact on our lives. And usually what we do here is we jump into a conversation in which we look at our most recent episode, what you had to say, his most recent episode, what you had to say. But this week we have another deep dive into a long form conversation between Matt and experts in the field.
And who are those experts today? Well, today is a very brainy conversation. Is a pun that Matt wrote for me.
You're welcome, Sean. Matt had a chance to sit down and talk to Dr. Kyle Wedgwood and a research intern, Wiktor Wiejak from the University of Exeter in England. Wedgwood's a mathematician specializing in neuroscience, and he's using a computer from FinalSpark. I'm sure you're wondering, as I did as I read this copy initially, why does the source of the computer matter?
Well, Matt just recently put a full video out on this. It's a special type of computer because in short, they're alive. They're made up of bundles of neurons, also known as organoids. I mean. Yeah. 10-year-old Sean on a Saturday morning would watch the crap out of the organoids, let's be honest. Yes. Yeah. And Dr. Wedgewood is using them to explore what he calls the fundamentals of how neurons work. So on now to Matt's long form conversation with Dr. Wedgwood and his intern, Wiktor Wiejak.
So thank you both for, for joining me to, to talk about your research. And the reason I'm reaching out to you and my team reached out to you was we're, we've been looking into companies like FinalSpark and these organoids and these new kind of like organic computers.
And our first thought was like, what do you even do with these kind of computers? Like what, what, like what's it like to use them? What kind of research is being done on them? And that's when we came across you and we reached out to you. And so I was hoping we get to kind of kick things off of could you both kind of introduce yourselves and what it is that you do.
Yeah, perfect. So I guess, I guess I'll go first. So my name's Kyle Wedgwood. I, I work at the University of Exeter in the Living Systems Institute. Um, as you might be able to tell from, from my background, uh, my, my sort of first love is, is mathematics. So I was, I was trained in, in mathematical neuroscience.
That's what I did my, my PhD in. Since I've been working in Exeter, I've been moving more into kind of experimental science and here it's really trying to ask the question about what can we import from sort of mathematical descriptions of, of things like neuronal networks and use them to, uh, understand how neuronal networks work, how cells communicate with each other, but also how can we influence some sort of exert some sort of control or some, some modulation in a sort of targeted way on neuronal networks, which is part of the reason why we're working with these sort of organoid models, um, with FinalSpark. How about you, Wiktor?
Awesome. Hi, um, I'm Wiktor. I'm an undergraduate student, uh, currently doing my third year. Very stressful time. Uh, exams are coming up and the dissertation's due as well. Um, but I got to meet Kyle, uh, last year, um, September, October time when he was posting some job, uh, applications.
I got very excited about what the possibilities of this organoid, uh, could be. And we had a chat and then an interview and finally I was able to get the job.
Based on like, so what you guys are working on, obviously neural models and mathematics and all that kinda stuff. What, what is it exactly that you're hoping to get outta the research that you're doing?
I guess you're talking still in quite general terms here.
In general terms. Yeah.
So, um, one of the really big sort of questions in, in neuroscience really, and bear in mind. We'll probably talk about this a little bit more later, but I'm really interested in sort of the fundamentals of how computation in the brain kinda works.
Um, things like learning and memory and, and part of this is trying to understand what components of neuronal response are really being driven by things within the cell and what is being driven by the way that the cells communicate with each other. So, um, you know, I guess I don't wanna be patronizing, but if we go back to sort of basic biology, right?
So we have neurons that are doing their own kind of thing. They have these little electrical impulses that we can record and when they, they don't quite contact each other, but they have little gaps between them called synapses and they can transfer information between cells. And so when you look at the complexity of the brain, or complexity of any bit of neuronal tissue, it's really what you're seeing.
Or what you measure, what you record is a combination of of cells, of cell intrinsic processes and synaptic processes. And a lot of the kind of work that I do is trying to understand which bits contribute in which context? If you asked me for your, for my phone number, and I, and I tell you, you need to store the pattern of digits that I give you in your working memory and translate that to something so that gets converted into a, a pattern, a sequential pattern of electrical activity in your brain.
That is sort of made more robust by the way that the cells talk to each other. So then that asks the question, right? So can you understand that process? Can you understand what happens when people get Alzheimer's disease? Right? So memory is something that people reassociate with deficits in this synaptic communication.
And then downstream of that, you can then, if you understand both of those two questions, you can start asking really funky questions like, okay, well how can you intervene in that system to correct some of those deficits? You know, if you wanted it to be really kind of futuristic, you could ask yourself, how can you improve learning, right?
In these kind of networks, is there a way that you can hack the system, hack the biology somehow? So these are the kind of questions that that keep me awake at night.
So you, your research is essentially like the, the building blocks of what could be a foundation for big advancements down the road. So once we understand this, we could treat memory issues, we could improve learning like you're talking about.
So it's like this could unlock, I don't wanna totally future stuff, but like curing diseases or memory issues, Alzheimer's, those kind of things. This kind of understanding could lead to better treatments down the road.
That's, that's the general idea. You know, I'm, I'm really interested in the sort of the, the fundamentals, the, the sort of, I guess you could call it basic or fundamental neurobiology, but down the line.
Yeah. I would love to think that this is gonna be, you know, a, a step on the way to curing diseases on the way to, to thinking more about how do humans learn, um, how do we remember and, you know, I dunno what the future will look like in terms of human computer tech interface. Already yeah, we've seen some really cool things happen, but now people are starting to think about sort of non-invasive ways of achieving the same kind of goal.
So, you know, whatever happens, it needs to be based on some sort of fundamentals. And for me, that even goes down to, you know, when I, when I think about neurons, obviously I think about them as biological cells, but I also think about them in terms of equations, right? I think about what can we abstract away, what are the sort of the key components?
What is modulatory? How can we understand what happens between a human brain, a mouse brain, something like this. Right? What are the things that are shared, which we still don't really know, right? Maybe we don't have that fully mapped out yet.
For as much as we know about the human brain, we really don't know that much about the human brain.
Yeah, exactly. Exactly.
I think that's the, the fundamental truth. Uh, yes.
So that kind of leads to the FinalSpark system because it's organoids. Neurons. What brought about the collaboration with FinalSpark? Did you reach out to them or did they reach out to you?
Yeah, so this. You know, there are some moments in life where things just happen serendipitously, and I think this is, this is one of those things.
So at, at the time that I first interacted with FinalSpark, I was writing a fellowship application, an application for a large body of research funding from the UK government. And I was developing my ideas on around things that we, we talked about at the top of this call, and it was all about technology development.
It's, I, I spent, it's 10 years now working in sort of mathematical neuroscience. I spent a little bit of time developing, um, sort of closed loop systems for putting my mathematical models into real systems, and I was writing a fellowship application around this topic and I dunno exactly what the steps were that, but that the FinalSpark reached out.
But, but they found me somehow, uh, and they said, okay, you are doing some work in, in sort of neuroscience and, and, and tech development. Um, would, here's our system. Would you be interested in using it? I was like, okay, that sounds really cool. And, and very well aligned to, to what I want to do. So we had a few calls with them.
They told me a little bit more about the system. They, they gave me a quick demo. And that's basically what, what kicked it off. And then, so Wiktor is sort of employed on, uh, a research, internship with the, the hub for quantitative modeling and healthcare, which is a, a body of people that I work with here where we're trying to take ideas from, from mathematics and, and use them to understand something to do with, with health and or human health and, and biology.
I, I and Wiktor, I guess you already explained a little bit how you got involved. And since then, you know, I guess when we first started speaking to FinalSpark, they were, I guess in the early days of rolling out their platform. So they'd been developing their, their organoid models, their multi electrode array recording devices, but it was still kind of early days in terms of their platform development.
So it's been really interesting for us to see how that's been expanding, uh, and uh, in, in response to things that we've also been saying. Right. So it's, it's, it's, it's been kind of. Kind of cool from that perspective.
And that, that was one of my big questions for you guys, which was how are the capabilities, like you were talking about a lot of mathematics and what you're doing and now you have this platform that is Yeah.
Kind of how do you change what you're doing from working in the mathematics to when using the FinalSpark system? It's like, how, how difficult is that? Or are there changes you have to do to your, your process?
So I wanna let Wiktor speak about this because he's obviously the person that is, um, is, is doing more of the day-to-day stuff, but I would also like to come back, uh, towards the end.
So I guess the, the di the difference between, between me and Wiktor with respect to this project is. So I'd already come in from the perspective of doing a lot of work in, in neuroscience on the mathematical side and a bit of experimental work as well. Wiktor comes in from a computer science background with a, a very different take, I think on, on what a or neural network actually is.
So maybe I'll let Wiktor take a first stab at that question.
So from the FinalSpark platform, in terms of running experiments and doing, um stimulations, it's not really mathematical, uh, in a sense of equations and specifics. I tend to do, and my job is I stimulate the neurons in a certain way that I look for a, uh, outcome from this.
And this is really interesting because you find these patterns and you find these exhibit these simulations, which are, uh, resembling of what we want to get out in the very end, I guess. Then once you get the results and you have everything compiled, you take it back into the more statistical world and you apply statistical methods to yield the results outta it.
Um, we've done a lot of different things. Um. Still trying to get the statistics under hand and Kyle helps me with that. It's been really interesting.
I was gonna say, but how, how, like this is where my brain like completely breaks trying to think about this. So you've got your mathematical ideas and your theories about what you're trying to test, and then you've got this system.
It's like how do you figure out how you're supposed to stimulate those neurons to get the results you're looking for?
Yeah. That, that had, that took a very long time to work out. Yeah, I mean, obviously when you translate from a very nice bit of equation or a bit of code into, into working with real biology, you, you encounter problems pretty quickly.
The, the first main thing being that cells are all different. The, the amount of heterogeneity you have from cell to cell, from organoid to organoid is, is pretty huge. There are certain things that are, are pretty well preserved, but other things which you, you almost have to take on a case by case basis.
So, um, some of the things that we've been been looking at is exactly what you've just asked Matt about how do we know how to stimulate the cells, where to stimulate the cells, when to stimulate the cells. And actually here I think is where the mathematics really, really helps us do that. So, um, I've been, well together with Wiktor.
We've been designing ways to incorporate things that are sort of known in the, I guess we'll call it neural computational, computational literature about what kind of responses you can expect to see. And we're seeing the same things in the organoids, right? Which is really nice. So, um, so that is kind of directly mappable.
So in that sense, you don't have to change the way you think about it so much. But what you do have to do is you have to learn to live with the, the beautiful amount of heterogeneity and uncertainty you have. And of course, the nice thing about the FinalSpark platform is that they deal with a lot of the cell culture viability kind of stuff.
Okay. So all of the, what we would call wet lab stuff is handled for us because this is a big problem as well, right? I mean, there are people here in, in the, in the institute that I work on, who, who, who just focus on the, the organoid kind of stuff, which is a whole wonderful body of science in itself. How do you design tissues and to have certain properties?
For us, it's more about figuring out what components of our way of thinking about things are exactly the same and, and what we have to build in some, some uncertainty about what we're seeing.
That actually, I mean, earlier when you said it's like. It's different from cell to cell. So that makes me wonder is like, because I know the organoids in these systems don't live forever, they have to be replaced.
Are they, like, is this the kind of thing where you can get consistent results doing the same exact thing every time? Or do you have to kind of like, move with the system instead of is changing? Like, that's kind what I'm wondering is like, because in, in, in, like computers, it's ones and zeros. You put something in, you're always gonna get the same thing out.
But is that the case with this system or is there like this malleability to it that you have to kind of learn to kind of, kind of dance with, I guess?
Yeah, as it stands at the moment, it, it's more like the second, uh, okay. The cell cultures are actually stable in some sense for, for longer than you might think.
But that doesn't mean that functionally the baseline is always stable, right? So the viability might be stable, but the actual computations that these networks are doing is gonna change a bit over time and, and you kind of have to live with that at the moment. I can see a future in which that is not true as we learn more and more about how to design these kind of organoid models in such a way that you get much more reproducible kind of behavior.
Um, I actually work with a group in Australia who are thinking about these questions about how do you design these kind of, uh, we call them just tissue models in general, that will be more reproducible, that will have certain properties that for a long period of time. And I think the reality is that. You really need a confluence of people working on the sort of organoid modeling side of things, building the sort of scaffolds that you need to create reproducible tissues as well as systems for recording stimulating like FInalSpark are offering, and people to do the kind of data science, um, to figure out how to deal with those wandering baselines. So I think where we're at now is, is an exciting time because I think these, these systems are capable of doing some, some really interesting things. If we want to get to the point where we're thinking about them as being a sort of like for like comparison with sort of conventional, sort of composite based computers.
I think there's still a little bit of of work that needs to be done, and that work really needs to cut across disciplines.
Since you've got a computer science background, I'd like. Is it kind of frustrating or maddening to have to work with it? I'm curious like what it's like to work with, like are you, are you, are you scripting stuff in your, like in Python or something on your own and then inputting it into the system and seeing what the results are that come out?
Like how are you going about the, the workflow of this?
So it isn't straightforward. Um, we have to use a dedicated system that is given to us by FinalSpark. So we can't specifically, we can script on our own computers, but then we have to input it into their system. Now, what streamlines their approach is just to code it on their systems from the get go.
And doing so is is very intuitive. We've got access to their database, to their systems, which control the organoids as well. And the API endpoints are extremely intuitive as well. Um, from that said, the API endpoints that they provide are very intuitive to use. And when I code up the experiments, what we tend to use is a classified method.
So most of the, um, step-by-step protocols that need to be done in order to get a result, um, they can be quite complicated and take up a lot of code. To prevent this, we have a set of steps that are predefined and we can then instantiate them time and time again. So, for example, stimulating the organoid is, requires a couple setup features, which can be automated.
Then getting the results back, um, is another task. You can do this in a closed loop fashion. So, um, when you're experimenting, you can get the results, 200 milliseconds of them back while you are performing the experiment. However, what we tend to find works better is when after the experiment post, um performing everything. We look at the database and then we find everything that is, um, in our region of experiment time and then evaluate that further.
Alright. So, and you'd also mentioned earlier that like the statistical modeling, you're still trying to get kinda your, your head wrapped around, I guess, like, so you're getting a lot of the results, but you still haven't quite.
I've been hitting Wiktor with a lot of, um, a lot of theory from stochastic processes, which is, is not quite his, his traditional ballpark. Okay.
Okay. I would like to consider myself. So you're getting, you're getting usable results out of the system.
One of the things that we're really trying to understand a little bit at the moment is, so you imagine that, you know, the system is basically this kind of sphere of neurons.
Embedded into that are these eight recording electrodes. So, and these can be used to, to record electrical activity and also to stimulate electrical activity. And one of the things we're trying to, to work out is how are the neurons that are being recorded from each of these electrodes influencing behavior on other electrodes, which is quite a hard problem.
Right? I mean, I won't get into the, the details of all of this. But ultimately, remember we're trying to think about this question about intrinsic properties and, and network properties. And this is one of the key questions that we kind of have to, to, to address. Um, and coming back to your earlier question about Matt, about, you know, thinking of this as a, as a computational resource, you can think about those kind of network ideas about sort of, um, transfer of information in the same way that you might call a function right? In a programming language. So you kind of need to understand those a little bit and, and we're making progress on, on, on understanding that, and it's really trying to get a combination of how can we perturb the system in a way to, to get more information, which in itself is, you know, there's some work to be done there.
And then this other question you asked about, you know, how stable are those, are those responses over a longer period of time? So there's some element of tool development. As well as the, as well as the actual experiments that we run. So it's always kind of this thing of, we, we design an experiment, we run an experiment, we, we try and work out, okay, did that experiment work?
Um, and then we redefine the experiment. So in, in many ways, it's like a traditional kind of lab based, um neuroscience, except the lab is split across Switzerland and, and, and here. Yeah.
So this, this does sound like the system is almost tailor made for you and your research. It's Right because, because it sounds like you would not be able to do what you're doing just with the computer.
Just with the math, this is actually taking those theories and you're actually able to do actual tests to see what's actually happening.
We do have some capabilities here of, of doing those kind of things. The, the nice thing about this sort of, uh, split resource, if you like, is a lot of the hard work is, is it, like I say, in maintaining and designing these cultures in the first place.
So it's, it's amazing that, you know, you can just log in. You can, they even have a, a sort of camera feed where you can look at the, the organoids. They're not that interesting to look at, right? Because, and they don't really move anywhere, but, but you can do that if you want. And so it's nice to, I mean, there is a booking system, right?
So it's not like you can just log in anytime and do stuff. But it is great to think, okay, well actually it would be really cool to just do this experiment and then, you know, you can book a slot at sort of two o'clock in the morning if you want. Go run it, see what happens. Um, collect data the following morning.
Um, so in that sense it is a bit like a remote server that you can just log into, right? That you, um, you know, you don't have to usually, if you want to run an experiment here, if I wanna do something with my colleagues, we've gotta plan sort of three or four days in advance to get everything set up. And then usually things don't work the first time, right?
So then you lose another few days preparing. Whereas this is a sort of system you can just log into whenever you like that. I mean,
you said you do said, you said you do book, like you have to book a slot to do it, but like, how long does it take for it to run the simulations that you're, you're giving it like?
Well, that's something that we, we determined.
Um, so in, in some of the stuff that we've been doing with Wiktor, well, I mean Wiktor can probably tell you a little bit more about it in a minute, but we, we try and design our experiments on the basis of what we think is gonna happen. So we've been looking at some experiments where we've been trying to a.
Work out what the network structure is between neurons in this, in this organoid, but also try to, to modulate that somehow. And so, um, there are ideas from the literature that tell us about how we should stimulate this, this organoid to achieve that. Um, and so that tells us how long we should design our experiment for.
And so Wiktor goes in and he books it for however long he needs to run that thing.
Well, indeed, I, I, I'm curious, like are there constraints that either of you are hitting with the system and that. Are, are you able to work with their team to try to find resolutions to that?
Well, so you've said about the constraints.
Um, the maximum time we can work is four hours, and this seems like not much time when you could be performing experiments over multiple days. However, I feel the scale of what we're doing isn't reached that point just yet. Typically, we'll be looking at a one hour experiment where we'll be stimulating for, say, 20 minutes, giving it about 10 minutes of rest.
Then continuing the stimulations over another period of 20 minutes. So I guess depending on what you want to do with this, with this idea, you could be looking at a very wide range of stimulation times, something that we don't really reach that threshold just yet of four hours.
I, I think one thing I would say, which I, I think is a, is a bit of a limitation for some of the things that, that we would like to do is at the moment we are using the sort of eight electrode device. And of course there are, there are many, many cells in these organoids. So we have a very, very limited, um, field of view if you like. And there has been in, in neuroscience in the field in general, there's been this big push for um, record more and more neurons simultaneously, which is good In one sense.
I don't think that's the solution to all of the problems in neuroscience, right. Just record more stuff. But it is, you know, it is useful to have some of this stuff. And with the system that we are currently using, we, we are quite limited in that, in that regard. Um, but I know that FinalSpark do have electrode systems that are, have more an electrode, so a finer spatial resolution.
So I think that's one thing I would say about the way that, that we are interacting with the system at the minute, we, you know, we, we have a limited, a limited ability to record a subset of neurons, if you like.
Is it like the resolution of the system? The resolution
is, is that we're using is is quite low.
How do you train the organoid to respond to a specific pattern?
That's a good question.
Um, yeah, I, I guess some of the interest in the use of organoids as a, a sort of computational interface device, if you like, came from this, uh, I think it was a 2022 paper, um, which is a, a collaboration between some sort of people like me, some more like, uh, mathematically minded people and some, some people more on the, the wet lab bio side, where they basically took, um, one of these organoids.
They basically kind of taught it to play pong, you know, the, the old Atari video game. And so what they did was they, they had this neuronal network and they said, okay, some part of the network is gonna control the paddle moving up and down, and some part of it is going to respond to where the ball is going on the screen.
And what they did was they used the combination of what they called predictable and non unpredictable stimulation epochs to basically train the neuronal network to not want the ball to go past its paddle. Over time if you stimulate these networks, they, um, they respond in a way by effectively strengthening and weakening different kind of connections between neuron.
So broadly, this is called synaptic plasticity, and it's one of the fundamental ways that, um, brains learn, remember, acquire new skills, all this kind of stuff, and you know their results. I mean, it's obviously the, the neural network did not become really, really good at playing pot, right? It just got better than it was at the, in the beginning.
But it did show that actually you can study kind of learning in these systems. So what that means for us is trying to determine the stimulation epochs that can reliably promote synaptic plasticity in these kind of networks. And so part of the statistical toolkit that we're developing are ways to measure this.
And ways to then design stimulation epochs that will more reliably hit this thing. So even if the baseline is wandering, we can design a set of stimulation epochs that will drive the system towards where we want it to be. I, I kind of, kind of like a, a bit of a, a, but wavy hand, wavy explanation. I,
I think like, again, where my brain breaks on this whole thing of like the research you're doing and what FinalSpark is doing and how they're so overlapped. It feels like the work that you're doing is gonna help them create a better system. I hope so. For this, right? So you're, you're like, they are not only at the bleeding edge of this, you're at the bleeding edge of this working together. Yeah. I think this is to make a better system.
Yeah.
Yeah. I, I think this is the way that, that everybody wants it to be. You know, I, I see this as a, as a collaborative project with, with FinalSpark. So, you know, we're, I think both sides have been pretty open about this, right. I guess as a disclaimer, they're, they're letting us use the system for free as an academic partner, and as part of that, we are feeding into them as a, as a, just a, a general user of the system.
What's working well, what isn't working, what our suggestions would be. But at the same time, you know, we are telling them a little bit about how we're thinking about the system, right? How, how we would design different bits. And you know that, and I think that's part of the, the collaboration.
Because, because like this is not a, a nitpick at FinalSpark at all, but like the marketing materials I see from Final Spark and others are like, in the future this is gonna power like AI and do all this other kind of cool stuff that you can do.
And it's like, I can't do that right now because you're clearly still trying to figure out the system. Like how does the system work reliably? And when I like logged into the, their website and was seeing what they could do. It was clear, like it would have to, you'd have to be a researcher that knows what you're doing, like what you guys are doing with the system.
'cause you're guy, you're, you're, you're building the baseline for what this could potentially unlock for the future. Is there anything else about your research or the FinalSpark platform that we haven't touched on that you, that might be worth knowing?
Yeah, so, so I think, but just to, to come back on what you said, I think that this is what you've just said is, is exactly how I think about it at the moment, right?
So if you look at where AI is in, you know, conventional computers, that's before we even get to, to things like quantum computers, but just conventional computers, it's clear that things are accelerating at ever, ever faster paces. And where we are with understanding and leveraging biological computation is, is very, very far behind.
There is a hope in the future that that situation will be reversed where we can take advantage of, of sort of paralyzation, sort of implicit parallelization that goes on in biological systems that you can achieve in some conventional computers. But it's much, much harder to do. There are lots of things that need to be addressed before we reach this end goal of AI on Wetware, I guess.
Uh, and I think it's exciting, but you know, we also have to think this is quite a long-term endeavor. Um, and, um, there are lots of lots of different disciplines that I think all have to come together to, to make this work.
I'm, I'm glad you brought up quantum computing 'cause that's the first thing that popped in my head when I was learning about this.
It felt like this is like the very early days of quantum computing. And quantum computing still has a ways to go. Yeah. And quantum, but it's like, yeah. So it's. Is it like there's three basic frameworks for computers today, and this is just another one of those frameworks that's in its infancy and needs to be figured out.
Yeah. So, so Wiktor, from from your side, is there anything we haven't talked about that you think would be worth knowing about? Like how you have to interact with the system or anything like that?
I, I think you've covered all of it. Um, my main area of expertise is interacting with the system itself and, uh, performing experiments, um, based on the theories that Kyle and I generate.
Um, but most of the questions, if not all, have been covered. Yeah.
Actually, there is one more question I have for you. Have they been adjusting the interface that you have to use as you've been using it? Like have they been improving the interface?
I've seen, um, very periodically, almost every single week them change what is going on on their documentation and also what is available to us as well. Um, it's kind of interesting to see the thing you're working on and is almost, um, the very interesting parts of your life being adapted and being changed day to day. So I guess the answer is yes, is they take feedback very, very carefully and adjust the system that they have very quickly.
So they're, they're rapidly evolving the system as you're going, which is Yes. That's really cool. Yeah. That's awesome.
They, they have a discord server where people can, can basically make comments and you can chat to the development team, um, kind of in real time, which is, which is kind of neat. You know, this level of interactivity is, is, uh, quite refreshing, I think.
I mean, that just shows how much of the bleeding edge you guys are with them because it's like you're helping to shape the system. That's, that's really cool.
Yeah. But I mean, also I have to say that I think it's pretty incredible that, and, and this is in no way meant to be a disparaging remark about Wiktor, but, uh, that you can go in essentially knowing how to program, but not really knowing much about anything to do with, with neuroscience and biology and, and go do an experiment just by, you know, by writing some Python code. I mean, you have to understand their system but.
Wiktor, like you can do that. But then the results you're getting out, you, I'm assuming you don't completely understand all the data that's coming out.
Not exactly. Um, I'm getting to a point where I'm beginning to understand very slowly. It takes a lot of time to take the theory that's in papers and you read about it and identifying what it is and creating those patterns. Um, but yeah, uh, it's getting slowly there.
Yeah, it, it's definitely one of those things where I think, um, if you really want to get the most out of these systems, you know, you really have to dedicate quite a lot of time to understanding the fundamental biology as well.
Um, even though someone else is actually doing that wet lab biology. You know, you, you really would, it's difficult to make any concrete conclusions without understanding really what's going on and under the hood, I guess.
Well, I, those, that, that was all the questions I had for you guys. I really appreciate you taking the time to talk to me and, uh, it's been, thanks.
Thanks to Wiktor and to Kyle for sitting down with Matt, and thanks to all of you for taking the time to watch or listen. What did you think about this? Was there an element of this work that made you stop and gasp and say, I don't want my computer thinking at me, or was there something that made you wish it had been explored further?
Drop into the comments. Don't forget liking, subscribing, sharing with your friends, and commenting. Those are all very easy ways for you to support this podcast. So if you enjoy what we do here, doing those four free things is great. And if you have a few coins in your pocket that you don't like the way they jingle when you walk around town, and we know that you do enjoy walking around town without jingles. You could throw those coins in our heads by clicking the join button on YouTube or going to still TBD fm and click the join button there. Both of those ways allow you to give us head wounds, which we greatly appreciate with your quarters, your nickels. Yes, even your dimes sting. Thank you so much for taking the time to watch or listen.
And we'll talk to you next time.