Making Computers Think Like Brains

a lot of them. Yeah, a lot of times people talk about one sort of using the other as a metaphor for it, like the brain is like a big computer, or computers are thinking. You might say that, and it seems like a really good metaphor on the surface, but when you get down into the nitty gritty of how each one works, they are incredibly different. Yeah. So they have some similarities. They both process information, they both store information, have memory, they both can perform logic, they both uh,

they both use electricity. Let's say that there are a lot of they have input and output, but they're fundamentally different in a lot of other ways. Fundamentally different, and especially in the fact that we really understand how computers and code information, and we don't fully understand all the ins and outs of how information functions in the brain. But we do know a lot about things brains are good at that computers aren't good at, and we talked

about some of those in the last podcast. For example, brains are more energy efficient than computers for the amount of processing they do. Brains are much more versatile than computers. So computer might very be able to very quickly execute a very specific, pre programmed set of instructions. But brains can learn how to do new things, they can adapt to new scenarios, they can reprogram them themselves to do a task better. Computers can't do any of this without specific,

you know, augmentation to allow them to do it. So we want to talk today about how we can make computers more like brains. Yeah, this is not a new idea. Um people have been I mean we've we've had discussions about this. Even even Touring had some thoughts about how a computer could at least simulate intelligence enough for it

to seem like it was thinking like a person. Now in that case, he was really talking about, uh kind of a simulation the idea that if it's simulated so close to human behavior that we might as well assume it has intelligence. We're not necessarily talking intelligence here in the sense of a a self aware of thinking machine. We're talking about giving a computer some capabilities that are uh, that come very easily to brains, but not so easily

to computers. Yeah. I think the difference between this and the kind of artificial intelligence we talked about in the Turing test is that's more like the appearance of intelligence.

This is working like a brain works. So if you were, example, uh, teaching a computer how to recognize certain phrases spoken out loud, and then you had someone else come in who was a different age, gender, and perhaps even a different dialect say that same phrase, the computer would still be able to pick up on it, even though the phonetics might be remarkably different. Yeah. So people were pretty good at this.

Not depending on where you are, you might if you're in Yorkshire and you happen to be from the South, you might not understand a word anyone is saying. I maintained that no one in Yorkshire understands what anyone else is saying. But but if you're a computer, I mean, these issues can become difficult even with subtle changes. So

it's well, it's not just sounds like that. One of the other things I should have mentioned at the beginning that human brains are very good at that computers are not very good at is dealing with real world data. So computers do great jobs with data that's like a formatted, structured set of number values. They do not do a great job with say a visual image or a sound or you know, the feeling of a of a touch. All that has to be translated into digital language for

the computer. It just doesn't work intuitively like it does for us. It's also really bad at making associations between things without you having built in entire ontology to teach it what things are in their relationships to each other. So making a computer think more or perhaps work more like a brain has a lot of a lot of of appeal to it. And like I said this, this appeal goes back quite a way. It is in fact

John von Neumann. In our last podcast, we talked about von Neuman architecture and the fact that the their your basic computer today is based on von Neumann architecture. This idea of essentially a CPU and RAM. That's that's your basic when you ole it all down, that's what it comes down to. Um. He actually was working on a book and the book was titled The Computer and the Brain.

He passed away before finishing it, but even then he was sort of looking at the differences between computers and brains and how could you perhaps start to close the gap between the two, Uh in which way? Well, and in the way of making computers more like brains, not brains, more like computers. He wasn't suggested, He wasn't suggesting that we all have positronic brains installed in our heads. Um, some people would like that world. I just thought we

should be clear. Yeah, that's fair, it's fair. The clarification has been made. Then. We also have a professor Carver Mead of the California Institute of Technology who began to experiment with designing computer chips that were inspired by brains, and he called them neuromorphic chips. Will be talking a lot about those today. And one of his collection of transistors mimicked how our brains processed visual data, and it actually allowed them to create a computer that could recognize

the edges of objects in an image. So being able to tell where one object ends and something else begins, whether that's just space or another object or whatever, which again very easy for humans to do, not necessarily easy for a machine to do. You know, you have to figure out how to how how do you have the machine identify that actually very difficult for machines to do.

Uh So, his work emulated brains but didn't simulate them, and he also admitted his work was much harder than he anticipated and that achieving a brain like computer system was a non trivial challenge that was going to require a lot of innovation because you could brute force it and as in other words, you could you could use a lot of classic processors to carry that heavy load, but that defeats the purpose. The whole purpose of making a brain like computer is so that you can take

advantage of the brain's properties, including that I'm amazing. Energy efficiency doesn't be any good if you are able to brute force your way into behaving kind of like a brain. But you're, yeah, yeah, you're you're. Every time you turn it on, all the lights in the neighboring houses go womb where they fade out for a moment in the cup.

Some crunching numbers. Again, that's not useful, So that would be uh, you know, that was in the nineteen eighties and there were some like I said, some success skipping way ahead. Uh, there are a lot a lot more work has been done in the fairly recent past. Back in two thousand eight, DARPA announced the Synapse program with the goal to break free from von Neuman architecture and

create a new model of computer designs. So this is about as dramatic a departure as you can imagine for a computer science You're talking about not just getting away from that CPU and RAM model, but possibly getting away from binary data entirely. So that's I mean, that's that's so different from what we've been doinging that it's uh,

it's it's completely revolutionary. I mean there's really hard to it's hard to imagine it just because we've been relying on computer so long that are stuck on this binary system binary just zeros and ones or off and on switches, if you want to think of it that way. Um, So in a broader category, you could just think about it not in terms of binary but just digital data of any kind, right, Well, I mean with binary specifically you're talking about your basic unit has two states, right,

That's that's what you're limited to. You can't have any more than two states. And that's one thing that Darka was saying. Let's try and get away from that. Let's look at a way of doing computation where we're not limited by this too state of your basic unit, which is incredible. I mean, it requires a total different approach. Now, two thousand eleven, IBM unveiled at the first synapse chips that could play Pong. Well, come on, now, they couldn't

just play, actually learn how to play. Right, So you sit your little brother a little little sister down at a video game, and because you're mean, you don't tell them how to play, and you just keep beating them over and over at Mortal Kombat or whatever. Eventually they figure out how to play. They might not be real good, but they figure out what the buttons do and how to win, and because they're younger than you are, they've got amazing twitch skills and eventually they dominate you in

every game. And then you just don't want to play. That's the killer instinct, right, Yeah, that's what this computer did at Pong. They sat it down they were mean to it. They didn't tell it how to play. They just let it know when it was doing good or doing bad. Right if it if it blocked the ball, then it got essentially a little bit of a reward like good job. And if it let the ball go by and uh and fly past and the opponent got a point, essentially got a message saying, you're not doing

so well. And so it began to learn that what it was supposed to do was block the ball from going in, and so it got better and better and better at doing that. So in in human terms that sounds kind of trivial because we can all do that, but in a computer that's a big deal. The idea of learning computers could very well revolutionize the kinds of

jobs computers can do. And another example they had, which you know is less trivial than learning to play pong, is they had this these kind of chips as a guidance system, part of a guidance system for a flying drone that could just follow a yellow line, specifically yellow lines on a road, and so it could actually follow along a road and stay on track because it new to recognize this particular yellow line and that that was

the course it was supposed to follow. So it was able to, you know, observe an environment and identify the important feature of that environment and follow that feature, which again something very easy for humans to do. You tell a human as long as they have their senses, and you tell human, hey, you need to do when when you see or experience this, you need to do this. It's very easy for a machine, not necessarily that easy. Well,

like us in navigating the real world. Computers like these are learning, not just being told what to do, but figuring out what it is they need to do. Yeah. So two thousand thirteen, so very recent past, HRL built a chip that can alter the connections between neurons synthetic neurons in this uh in this chip, and it's similar to the way the brain can change synapses. So it learns quote unquote like a brain. It's able to make

these connections that are emulating what happens within a human brain. Now, keep in mind, as we said before, we still don't fully understand all the complexity that is the human brain and the human nervous system and and how we think. We're still learning quite a bit about that. But by emulating some of the physical processes. We're starting to see some improvements and things like energy efficiency for specific types

of computer tasks. Yeah. Well, there are also very large scale adaptations that are being made in computers to make them more like brains. For example, you've got neural networks. You're familiar with this concept. Yeah, yeah, So in your brain you have a neural network. You have a network of neurons, and they're these nerve cells that are connected in tons of different ways all throughout your brain, sending signals back and forth to each other. And somehow this process, uh,

that's a very simple representation of it. There there are other kinds of cells and things going on to somehow this whole process creates thinking, It manifests itself and thought right. And artificial neural network tries to use some of the same basic concepts to make a computer or network of computers work more like a brain. So, neural networks are systems of interconnected nodes that simulate Each node simulates the

behavior of a neuron. In basic terms, neurons in the brain receive a signal, and if the signal is strong enough, they pass that signal along to other neurons. Artificial neuron nodes receive quote weighted signals, and they received them from lots of other nodes, and then the waiting system determines whether or not that signal gets passed along to the next layer of nodes or to a final output node.

And so these networks are very good at things like pattern recognition and learning how to deal with new types of data sets, and when spread out across lots of different processors, artificial neural networks are very good at processing huge amounts of data very quickly. Yeah. So an example of this is Google Brain. You may have heard of that. That was a hush hush Google x project, right, yeah, one of many, uh neural network approach just like you

were saying. So it uses thousands of computers to mimic the synaptic connections in our brains and rely is on simple sets of rules also called algorithms, right, So it's it's relying on that to learn on its own. So the example everyone loves to mention, you'll see it all over the internet because it's the Internet, is that it learned what a cat was by scanning thousands of pictures

of cats. Now it had been told the term cat, it had to be it did not realize it didn't see tens of thousands of pictures of cats and is a cat like it could have just as easily said that is a bucket. That's a different meme entirely. Yeah, no,

walrus is Yeah. No. The amazing thing about it is that that it was able to isolate the visual concept of a cat and then be able to identify a cat, even if it was a cat that was not among the pictures that already scanned, a totally new cat in a in a video or a picture exactly so you can show it a picture of a cat and had never seen before, but because it now knew what the concept of a cat was, it could identify that as

a cat. Yeah. One of the people behind this project was gave a quote saying that it basically invented the concept of a cat, which is kind of true. Like it it had to just I thought of it, and I wrote this in the video. Don't know whether you said these words or not, but I thought about it sort of in the in the terms of like a naturalist surveying new animal or plant species, you would you would look for patterns, and you're not being told what

new species to look for. But by observing enough of them, you'd say, Okay, I'm seeing all these similarities with slight variations, this must be a species. And here's another species that's kind of the same but different. And you're creating these categories in your mind without direction from the top. You're just taking in lots of data and figuring out what to do with it. And that's what these neural networks

are really good at. So now what they're not necessarily good at as being super energy efficient, because we're talking thousands of computers. Was sixteen thousand computers to figure out what a cat was something like it was, So we're talking about a very energy hungry approach here. Because each computer is behaving as a neuron. It's not like we're we've you know, distilled the neuron down to an element that's on a transistor. No, we're talking about individual computers

acting as a neuron. So you can also simulate them within a computer, but this is far less powerful. Right, if you simulate within a computer, you better have packed a lunch and for the next like several months. No, it's incredibly slow simulating these things, making virtual brains, which people have tried to do even on a small scale requires an enormous amount of processing power, because again, that simulation is resting upon a classical computer approach. It's not

something that's broken free of that Von Neumann architecture. It's just resting on top of it. Um. One of the guys working on, or at least working with some of the people who are working on Google Brain is our friend R. Kurtzwell Singularity Dude. Yeah, so, yeah, a pioneer in the in the field of voice recognition software as well as other artificial intelligence developments, and he's working with Google,

not necessarily on Google Brain. At least the interview I read, it seemed to me that he was working, you know, some of his colleagues were working on Google Brain, and perhaps he had worked on some things that are are related to it. His natural language work is something that could be very useful in UH and certain approaches with a neural network, so you know, it's related to but

not necessarily directly overlapping the project. Another project that I wanted to mention along these lines is an open source one called NEST, which stands for Neural Simulation Technology that started coming together in the mid nineteen nineties and and studies large at works like the brain and and tries to write flexible emulators for the brain for for general

research use. It's something that I ran across while I was in the rabbit hole of all of this research, and I thought that it was really neat that it's you know, we're working towards giving researchers better tools for

studying this kind of stuff. That's pretty cool, totally. Okay, let's let's zoom in a little bit, because we were talking about these sort of artificial constructs like neural artificial neural networks and you know, take place plus across a lot of computers or in some virtual simulation inside a computer. Can we make the actual hardware in your device work

physically more like a brain does? That is the million dollar question, and in fact is the thing that I think is the most important if you want to capture the energy efficiency angle here, right, if it works more like a brain, then it's going to require far less power than trying to simulate what a brain does using this other architecture. Well, based on some really cutting edge research, I think the answer is, yeah, overall, there's there's a

project at IBM right now. Yeah, the neurosynaptic chips. We mentioned that just the beginning in the timeline, But we're looking at things like collections of six thousand transistors copying the electrical spiking behavior of a neuron. So six thousand transistors is nothing. I mean, you look at a micro processor right now, and there could be more than a billion transistors on that single one inch silicon chip. But usually transistors are not modeling spiking behavior. No they aren't.

But what I'm the only point I'm making is that we're talking much smaller now. Before before we were talking massive networks of individual computers. Now we're talking about discrete elements that you would find within a computer that are mimicking the behavior of a neuron specifically because of the way they've been engineered. It's not like you know, you've hit a little switch like, oh, now, the transistor things like a brain. No, it's because it's been specifically engineered

that way. So yeah, it copies this electrical spiking behavior, and the collection of transistors are wired together to make a system of synthetic rons. So using these chips was a little complicated. When they first showed them off. The team had to simulate a physical chip with a virtual chip inside a computer program then port that configuration over onto the physical hardware that they had built, and that chip allowed programmers to create protocols so the computer could

learn how to recognize handwritten numbers or play pong. This is the pong playing chip we talked about at the top of the podcast. So, but they could also learn like if if you show a computer, like you've got some optical recognition software, this is something else Kurts while worked on optical recognition software in your computer, and you show it a picture of a six, the number six that has been printed in a specific font, and then show it another number six that either someone has handwritten

or it has been printed in a different font. The computer may have trouble identifying it because it's not exactly what you showed it the first time. But let's say it really really wants to leave a spam com in the comments section on your article. How can it get

past this problem? Well, using this approach with neurosynaptic chips, it could really help because, like I said, they were able to use this this network to create a program that was able to recognize digits that had been drawn, even if they had totally different people drawing them in different styles, they were still able to have this machine

recognize what that number was. So really, when you look at it, this particular type of machine was end end of which, by the way, the people who created Capture are pretty much okay with because they said that the yeah and only it was a stop gap. It was something that pushed forward the development of artificial intelligence because they said, we had to create a thing that was difficult for machines to do but easy for humans to do. And really, what that does is it points out, hey,

machines are really bad at this one particular task. How can we make sheens better at doing that? And so while it's kind of irritating for anyone who wants to have controls on their site so they don't get hit by spam content, uh, it is very reassuring for anyone in the artificial intelligence field because it shows that progress is being made. So double edged sword, I guess we

could say, but at any rate, Uh. The whole idea was to emulate the brain's ability to make lots of those synaptic connections at once, and the current state of the art involves organizing neuro synaptic cores onto a grid to create a synthetic cortex. So IBMS neuromorphic programming architecture is based off modular blocks of code called core lets, and programmers can choose core lets from a predetermined menu and choosing the type of program for specific tasks like

image recognition or audio recognition. So that stuff, you know, the stuff has already been built by IBM, so that people who want to work with this kind of technology don't have to invent from scratch. They can actually take the stuff, tweak it to their needs. It's own, you know. It's it's a very well, I mean, it's a very modular approach. It's really innovative. Yeah. Qual Calm was also working on neuromorphic chips for for sort of learning machines, right, Yeah,

they were looking at it. FEXTELR. There's a program called the Zeroth program zero, after the Asimov Zeroth Law of robotics, Yeah, which is cause no harm to humanity, nor allow humanity to come to harm. That's sort of the principle behind law number one, which is don't harm a person. Yeah. Yeah, this in this case, it's saying don't harm people in general, like all of people don't do any harm to them, so qual it comes approach isn't necessarily to protect humanity.

That that's not the that's not the issue here. A little bit grand. Yeah, their their their goal is to make a large scale commercial platform for neuromorphic computing. So they want to incorporate this into things like handheld devices so they can recognize certain things or maybe even wearable stuff like medical uh phenomena, like anything that it could detect that something's going wrong with a patient, it could alert you or just you know, context aware sort of stuff.

Was it Qualcom that created the learning rolling robot? I think it was along the same lines as the Pong game, So you had the computer based on neuromorphic technology that that learned how to play pong without being told. Uh. There was a company, I believe it was Qualcom that used neon neuromorphic technology to get a robot to figure out where to go in a room without telling it

where to go. It's the same kind of principle. They just let it roll around and when it went to the right colored tiles on the floor, they give it positive feedback and say, yeah, that's right, that's what you

want to do. So in this case, you're looking at like a device that could give you a notifications and and advice in certain situations, like if you've ever heard of Google Now, which is a an option on Android phones, android devices where you get Actually, I think I Os might even have it now as well, but I know Android does because that's what I have. It brings up little things based upon things like you're you're browsing history.

So as long as I'm logged into a a Chrome browser and I'm browsing around, it starts to keep track of that stuff, and it might let me know contextually when that stuff would become important. Let's say that I did some research on a place that I want to go and visit sometime. It might pull up more information that I could use to look into that. It's kind of in that semantic web approach. But this is all done on the back end for Google, right. This is

all done in the cloud. So what qualcom is looking at is creating these neuromorphic machines that would actually be doing it natively on the device itself, not being related out to some cloud computing network of sixteen thousand computers, but rather happening in real time in your hand at that moment. Along this a lines. Bioengineers at Stanford University have developed a circuit board based on the brain synup

structure that they're calling the neuro grid. It's it contains sixteen custom design chips that together they say can simulate

one million neurons and billions of synapse connections. The whole thing is about the size and energy efficiency of an iPad um and they they say it's about a hundred thousand times more energy efficient than a regular computer simulation of of that many neurons, but still way less efficient than the human brain um, you know, which contains about eighty thousand times more neurons than that and runs on

just three times as much power. So, but it's really cool and and right now they're they're focusing on the goal of giving researchers who are working in robotic press theses tools that they can use to to simulate human reactions without needing to know intimately how the brain works, you know, Let those let those people focus on the robotics programming that they're really good at, and this machine

translating that into a quick and efficient commands. Interesting. Yeah, so so the robotic prosthetic can actually learn that's kind of interesting. Yeah, or you know that that it it can think the way that your brain does, so that it can you know, help you solve that problem of picking up a cup, which is an incredibly easy thing for a human person to do, but really difficult for

a robot to determine. Maintaining your balance if it's like a robotic leg which again something that as long as uh, you know, as long as you're not impaired in some ways, fairly easy for a person to do, assuming you're on level ground and everything, but can be tricky if you are, feel if you have to wear a prosthetic. So interesting stuff,

I mean, really taking over the fact. You know, we talked about embodied cognition in the last podcast and the idea that that are are thinking could extend beyond our brains, that it incorporates the entire nervous system which stretches through our entire body. Well, if you are missing at them, clearly that would mean that you know, with a prosthetic you're replacing that. You have to build that kind of intelligence into the limb if you expect to have the

same sort of utility as the functionality. Granted, you know this this one single device that they've got right now cost some forts to build, but the team thinks that it could be mass manufactured for one that like four hundred bucks of pop. That's pretty impressive. Well, how about we try a different approach. So we talked about neuromorphic machines. What about synaptic transistors? Have you guys heard about these? So this is super Harvard project, right, yeah, Harvard project.

It's some light reading. If you want to read through the Harvard project, it's it's a little dense, but here's the here's the takeaway. So the neural networks and the neuromorphic transistors are both attempting to do the same thing, just at different scales, which is to simulate the relationship between neurons in the nervous system and to try and take advantage of that relationship the way that they communicate with each other through. What do those neurons communicate through

the synapsis? Uh? The synapsis synapsis is again the the little gap between one neuron and the other neuron where you're having the information pass from one cell to the next. So a synaptic transistor isn't just simulating neurons. What they're trying to do is simulate the actual process through which neurons pass information. So it's not the relationship that they're they're aiming at. They're not building synthetic neurons that will

behave like biological neurons. They're building synthetic synapsis, and they are still dependent upon things like you know, ion exchanges, except we're talking about oxygen now instead of sodium and potassium. So, uh. The other thing that's really interesting in this, and this is where you can really break away from the von Neumann architecture, is that you're no longer dependent upon a on or off stage for your base unit. You know, your your binary unit. Your zero or one can either

be of those two states, and that's it. Um This could actually have a whole spectrum because it's dependent upon the conductance. So it's like the strength of a connection. Well, not strength, we're talking you know, conductance, it's it's it's the ability to conduct electricity, and the conductance of these synaptic transistors can actually change and it's an analog continuous change.

So think of it like a sign wave, and it could be the value of that particular state could be anything along that sign wave, So not just a zero or one, but anything along that entire wave. It's kind of the same way if you think about it, that the cubit, the quantum bit of quantum computers is said to be both zero and one at the same time and all values in between, which doesn't make a whole sense, you know, intuitively, But it's kind of similar to that. This idea that it it allow us to have a

completely different approach to computer science. It would it would require a lot of different work to really make this useful. I mean, it's the possibilities are are pretty phenomenal, but it does mean that you're not programming the way anyone has programmed up to this point using the traditional computer sciences. So very exciting, very weird, and uh yeah, we'll we'll

do a link. I'll try and write up a blog post with a link to the Harvard article because you know, it's it goes into a lot more depth um, and it's very difficult to summarize in a way that doesn't trivialize it um. And I certainly don't have the expertise to go into a lot of depth on the subject without stepping in lots of holes and revealing my ignorance, uh speaking of holes. Yeah, I think we should talk about whole brain simulation. I'm so proud of I shave

my head and I'll teach you the ukulele later. Uh. Yeah, So there are quite a few simulated brain projects that have happened, but but two or that are in the process of happening. There's actually been a few different attempts at simulating the brain. Most of them are ongoing. Um, i'd imagine, I mean, this is not something we can do yet. No, I mean we've seen some projects that simulated part of a brain, as in like a collection of neurons that would represent just a fraction of the brain.

But even then you're talking about on a time scale that stretches into days for that would match up two seconds in an actual brain. Yeah. I know one that I read about years ago, it must have been going on for a long time was the Blue Brain project. Yeah, it's a reverse engineering effort. Really, it's meant to understand more about the brain, uh, more than anything else. So I think a lot of these whole brain things are Yeah. So it's not necessarily let's make a computer that can

think like a brain. It's that's that's a far feature step right now. They're kind of like, let's learn some stuff. Yeah, let's learn what the brain. Let's learn what the brain do. That's really what it comes down to. So they're looking at synthesizing of mammalian brain, specifically human brain down to the molecular level. You just the whole idea of like recreating a whole brain and what it does. It's like, congratulations,

you have synthesized compulsive behavior and confirmation bias. Exactly, good job. Their goals also include reconstructing the human brain and building a virtual brain in a supercomputer. It began the project began back in two thousand five, and each simulated neuron requires the power of a laptop computer, not necessarily meaning that a neuron is a laptop computer, but that's the it's the equivalent to a laptop computer to simulate a single neuron. U So it's successfully created a virtual model

of a rat cortical column. That's a collection of about ten thousand neurons. Now, these are the basic unit of a cortex. So according to the team, a single rat cortical column might be devoted to a whisker, So every whisker a rat would have its own dedicated cortical column. But that's progress, whisker. We have solved the equation of life. Yeah, now another project. Sorry, that didn't mean to downplay their success.

I mean that's pretty cool. It is cool. Is it does illustrate how far we have to go to get It's one of those one small step kind of situation. Yeah. Yeah, one of those things where you really realize how big a problem this is, how big a challenge it is. The Human Brain Project is another one. This one launched in two thousand thirteen at a conference in Switzerland, and the project is expected to cost a total of around one point six billion. Do it's a joint effort of

the entire European Union, Yeah, yeah, it's not. It's not something that is just being taken on by a you know, a guy with a guy in a dream and a song in his heart and a brain on his drawing board. Now, at any rate, it's um. The very first decade of this project is just dedicated to learning more about the human brain and how we see, think, learn and all that before we try and build the the computer model

of it. Because the argument is that we could try and go about building the building blocks and then putting it all together, but without this understanding of what's going on inside our brains, it's not very meaningful. It's almost you know, it's not quite as bad, but it's almost like you open up a a a door that's going to be your new office in your new house, and you throw in a bunch of transistor chips and and chords and a monitor and a keyboard and you close

the door and you think, I have a computer. Now it's not not No, it's not doing anything meaningful yet. So over here in the U s um, we've got the Brain Initiative. That's the Brain Research through Advancing Innovating Neurotechnologies Initiative. Yeah, rolls right off the tongue. Um. It's a it's a one billion dollar project with a primary focus on the ways that individual cells and and complex

neural circuits interact in both space and time. Um So, So it's talking about the architecture and the signal paths within the brain right now. Um uh. The initiative was announced in April, and it's being run by collaborative efforts from the National Institutes of Health, DARPA, and the National Science Foundation. They've got a kind of crazy collaboration coming up with the Human Brain project, Like they just announced it this March, and they haven't announced what they're collaborating on.

But it's cool that everyone's playing together. So I think that's really cool. But I also have a kind of weird, spooky, semi philosophical question about what it means to create or simulate a whole human brain, because we've talked before on this podcast about the question of whether an artificial intelligence program that seems to act like a person actually has consciousness. What we mean is, like, you know, an experience. Is there such a thing as what it's like to be

that program? And should it have rights? Should we have considerations for it? Right now, we don't generally worry about this. It just doesn't you know, we have no way to prove it doesn't have a conscious existence, but it just doesn't seem like it intuitively to us. Yeah, it's not like when that IBM neuromorphic transistors playing pong that it suddenly gets mad if you beat it. It doesn't like slam the paddle to one side to the other. Well, and even if it did, I mean you could think

about that, well, yeah, that's its programmed behavior. You know that that's how it's behaving, But you don't think that these programs are actually having an experience that deserves recognition. I wonder, though, if we're able to simulate a whole brain, especially if we are able to simulate it physically. Somehow, that seems to be getting closer to the realm where I'd wonder, Wait a second, are we creating something that's having a conscious experience? I mean, we don't know, but

could that emerge from that kind of hardware? Yeah, that's exactly what I mean. I mean it was so we don't know what the origin of consciousness is. We think that, you know, mammals with complex brains, uh and more advanced thinking seemed to exhibit it. We don't really know anybody has it except us. We just have to assume they probably, I'm not sure about myself some days, honestly, And by us, I don't mean like our species. Yeah, I mean like the the solipsism problem. It's like, I know I have

a conscious experience. You can't prove anybody else does. You just have to assume they do, right, Um, So, yeah, what do y'all think about that? If we build a brain that works in a one for one way, exactly like a real organic brain. Is there such a thing as what it's like to be that synthetic brain. That's an impossible question to answer right now because we have not done that and so and we don't fully understand consciousness as far as it applies to us, So being

able to answer that question it would essentially be total guesswork. Well, although I mean, to to create something like that, you would technically be having to create a body like like we talked about in our first podcast, your your neural network is a lot more than just your brain meets um. You've got a lot of other neurons and and it's tied pretty intrinsically or or so a lot of thought goes into into what you touch and what you taste

and how you feel in your body temperature. So in order to create a one for one simulation, we would need to create a body for it. And at that point, I think absolutely there is an experience of being that machine. Well, and I mean you could you could have it have an intelligence that's just not a human intelligence or not human like. So if you did make it a completely disembodied brain type thing, uh, you could have it where

it has some form of intelligence. It just wouldn't necessarily be be be comparable to a human experience because, like you point out, we have the body the whole embodied cognition theory approach that would not apply to something that doesn't have that kind of input, which might just mean that you have a computer that won't stop screaming. But

I mean, I don't. I honestly don't know the answer because personally, I think if I'm getting down to like what my personal beliefs are, I think that the mind is completely dependent upon physical matter is the whole nervous system approach, and so therefore, if you were to be able to completely simulate uh, the nervous system of a complex living being, uh, I don't. I don't see why there couldn't be some form of experience or consciousness within

that synthetic being. Yeah, I I think that's sort of entailed by the belief if you have a basically physical idea about what the mind is, as long as you don't believe in wonder tissue or you don't think that there's something magic happening in the brain you have, you really do have to wonder, like, man if if we're synthetically putting together a material construction that's pretty much like what brains do. It may very well be having an experience there, may very well be what it's like to

be this computer. Maybe. I mean it's hard to say. Like, I think we wouldn't be able to identify it until it was as close to us as we could possibly make it, including that entire body experience all right up until you know it says like hey, stop poking me, or hey I am self aware, or hey, share your pizza, or hey, really, seriously, I don't want to go in the maze again. I know the way out. Stop play.

You share your pizza with the synthetic brain fairly, because that sucker is going to know if you end up giving it the small piece. Please separate my hemispheres and cram that hot pizza right in between them and then seal it back up. Honestly, that's what I want to do with pizza sometimes the brain right, so the cheese gonna pools up over the cerebellum. No, I just have a I have a different point of of that I would like to talk about before we sign off here,

which is, do you think is it possible? And I expect it is that building computers that can work like a brain, you know, whether you call it thinking or not, they work like a brain, that they're more energy efficient. Do you think that we'll end up seeing those type of computers being very good for things that we talked about, like recognizing visual representations, audio, the sensory stuff, recognizing patterns, that kind of thing. It will be good at that.

Do you think that they will lag behind traditional computers with the things that traditional computers are really good at, in other words, like a quantum computer. A quantum computer is really good at solving parallel problems, but not a classical computer serialized problem. So I wonder, because you know, our our brains are really good at this other stuff and computers aren't, I wonder if the computer version of our brains will also be really good at what we do,

but lousy what traditional computers do. Yeah, I can totally see that. I mean, computers are the way they are, or one of the reasons they are the way they are is that they're they're quantized. You know that they're dealing with specific quantities of digital data, and our brains just don't work that way. I mean, we can do math, but we're still juggling concepts. Yeah, it's all really poly up in there. It's not a series of equations that we're solving. I mean, I mean it is, but well

it's fuzzy. It's fuzzy, and in that fuzziness is a is a good thing, um, because it is what allows us, like we were talking about earlier, to deal with like your brain can actually deal with the picture. It doesn't have to turn as far as we know, it doesn't have to turn the picture into a bit map. You can just look at it and and for all we can tell what's happening is okay, that's a lamp. Well, I mean not just a picture, but looking at the

world around us. I mean that the same sort of thing is that it's not it's not turning everything around us into one's zeros um. I mean, that's the way any sort of computer that's using a camera ultimately that's digitizing the information. Uh, you know, it's it's the same sort of thing. So you know, yeah, I would say that we're you know, this is not going to be a replacement for the kind of computers that we have

right now. But but the computers that we have right now are really good at you know, playing cat videos and uh, running tax software. And that's not what we would ask this other kind of computer to do. No, we would ask the other computer to be our friend or at least to help us recognize more pictures of you know, whatever it is we're interested in, to wench with us about taxes and find the cool cat videos for us, right, you know, at least to invent the concept of a cat over and over and it does.

Maybe wonder if that Google Google computer just became obsessed with YouTube and just started watching cat videos all day long. Um, if so, what a terrible thing we have brought into this world. It must be stopped. This is what an angry young man ranting into his webcam looks like. No shortage of that. Well, I'm gonna I'm gonna say that we've wrapped up this topic for the sake of my sanity. But yeah, if you guys have any suggestions for future

topics that we should cover here on forward Thinking. Maybe there's something about the future you've always wanted to know more about, or you're curious of a particular science fiction technology has any basis and science fact, let's know, send us an email or addresses f W Thinking at discovery dot com or drop us a line on Facebook, Twitter, or Google Plus, or handle it. All three is f W thinking and we will talk to you again really soon.

For more on this topic in the future of technology, visit forward thinking dot com, brought to you by Toyota Let's Go Places,