Computers vs. Brains

let me guess cinnamon toast. I actually was not thinking, well, now I am, because now you put the thought into my head. No, I was thinking about how our brains are different from computers. And we've talked about artificial intelligence multiple times and we've kind of alluded to that, but that we really wanted to go into a deeper look at the differences between the way computers process information the way our brains process information, and it really kind of

dig down into that. You know, when you listen to some futurists talk about the singularity and the merging of mind and machine, one thing you often get the feeling of is that they sort of think of the brain

as a computer. The brain is a computer, and if you can basically just get them sync up right there, right right, And it's a really sexy thought, especially since many futurists have backgrounds in computer science, and so it's kind of like, oh, yeah, obviously this is a great metaphor us, especially if they're thinking about, like we, with just the right amount of computing power, we'd be able to pour our brains over into computers and then achieve

digital immortality. You know, that's a That's one of the many ways people have envisioned a kind of singularity events. Another thing that ties into that is sometimes they compare the power of a brain and the power of a uter as if they're both both. Basically just like scaling up this same chart with the same value. You know, it's just a number and a computer is higher or lower or something like that. But really, a brain and a computer, though they do some of the same things,

are are in other ways fundamentally different. Sure, It's like it's like saying, you know, if if I'm able to lift a certain amount of weight and a machine were able to lift ten times that weight, that that machine is ten times stronger than I am. That's a pretty simple comparison. But if you then try and convert that over into the ability to process and understand and learn and and adapt, you know, that's that's totally different. It's

it's the it's fundamentally different. Like you said, the way a brain and a computer work, using your analogy, be kind of like, well, yeah, okay, so the machine can lift more weight than you, but you have hands. So maybe you could pick up a heavy but delicate thing that the computer couldn't pick up without puncturing, or that you could realize that that they as a cat without having to be taught over the process of many, many weeks what a cat looks like, and that ties into

what we're gonna be talking about in this episode. So I think we should start by looking at some of the big high profile computers people have seen. Is the really smart ones, the kinds that compete with human intelligence in a very recognizable way, like Deep Blue. So Deep Blue this is an IBM computer that was designed specifically

to compete in chess tournaments. So this was one of those things where various computer scientists had been predicting that a computer would be able to beat an expert at chess. Uh eventually given enough processing power, really and Deep Blue was the first one to really achieve that it was able to beat a world chess champion, Gary Kasparov. Specifically, it was the second of the great Deep Blue Kasparov matchups.

The first one did not go in Deep Blues favor, but they IBM kind of revisited it, did some adjustments. There was a second tournament. It was the best of six games, which I guess if you really think about it, there could have been a complete draw there. But in fact, Deep Blue one two matches against Kasparov's one match, and then they drew in the remaining matches, so Deep Blue

came out the overall winner. Immediately, IBM retired Deep Blue because they said that was exactly what they set out to achieve, was to make a computer that could compete at a world chess champion level. And um, straight from the after party to the burner, Yeah, exactly, like you're just right at least deposit chess master computer into incinerator slot. Well, I'm not gonna say they destroyed it, they just retired it.

They anyway, this particular computer was able to analyze two hundred million moves per second, or fifty billion possible positions within the three minute time limit that chess players get per turn. So that's kind of interesting because obviously I am not a chess grand master, uh and I don't know all that much about chess, but I would think to some extent, chess has to be kind of intuitive, right like a like a very very smart person could think many moves ahead, but the very smart person cannot

think fifty moves ahead. I'd bet most chess champions would not be able to think nearly as far ahead. But the other thing to take into consideration is not just thinking ahead, but anticipating what the other moves are going to be. So, for example, I'm a terrible chess player. I love playing, but I against anyone who's who's competent at chess, I am almost guaranteed to lose. And it's partly because my brain just does not work very well when it comes to planning out moves and anticipating what

my opponent will do. I might anticipate what my opponent will do based upon my particular circumstances at that moment, but I don't necessarily consider weight when I make this move. It's going to change those circumstances, which then yeah, and then they're all also thinking ahead and they may be changing their plans dynamically as I make my moves. The computer was much more adept at looking at all the potential possibilities and picking the most advantageous uh choice out

of all of them. Now, even so, humans are really unpredictable creatures, and uh, there were other computer programs that didn't do nearly as well against human chess champions. Even after Deep Blue, it would take a while for that

to really develop. And even even people who are chess champions who look back on the Deep Blue Kasparov matches say that Kasparov was just kind of off his game on that that particular tournament, and that if Kasparov had played just a little differently, he probably would have come

out victorious. So it was. It was even in this very specific instance where you've got a limited number and it's a huge number, but it's a limited number of moves you can make in any given situation, the computer was very, very good at it, better than the world chess champion. But that was a very specific use case for that. Oh sure, well, I mean from a computer's point of view, Chess is about probability and and logic solving and and those are two things that computers do

incredibly well. Yeah right, Well, I mean, if you've got the space in your memory and the time to look you know, at what was it two million moves per second, that it doesn't really require intelligence. You just kind of have to crank the numbers well. And and again, even with Deep Blue and being so great at chess, if you put Deep Blue up against a human in a different type of game, it wouldn't have done as well.

It wasn't It wasn't designed to do that, right, You mean, if you put it in front of a checker's board, it would just violate the rules and get disqualified or connect four, or it would be terrible at twister. Yeah, okay, but no, the point I'm trying to get at your jokesters is that this these computers are really really good at the specific applications they're meant for, but they aren't any good at anything outside of that, whereas a human

could be good at lots of different stuff. Yeah, programmatic behavior, and we'll talk more about that later. How about another one sort of along the lines of Deep Blue Watson. Okay, so another IBM computer. A lot of IBM is going to be talked about in this in this particular episode, and the next episode will be doing well studies show they make computers. They do, they've been doing that for a while in fact. But Watson, of course, is the Jeopardy Jeopardy UH computer. It's doing a lot more than

just playing Jeopardy. Yes, it's no longer just resting on its laurels as Jeopardy champion. Now Watson wasn't just one computer, No, this was a computer cluster. It's actually a cluster of ninety Linux based servers with a total of two thousand eighty processor cores. So it's built for parallel processing. That's something else we'll talk about with the human brain, about how the human brain works compared to say computer, a

classic computer model. So it's able to solve more in parallel, which was very useful when you had to do things like search all of the data banks for information that seemed to fit any particular clue in Jeopardy uh, and then assign a probability for how how sure quote unquote is the computer that that would be the right response to any given query? And if it if it met the threshold, it would give that response. If it didn't

meet the threshold, Watson would stay conspicuously silent. Now, from what I understand, Watson had some interesting intelligence and that it was even okay at parsing some of the word play, which is pretty impressive, though that was ultimately where it performed the weakest. Right, Yeah, so yeah, I love if

you've ever watched Jeopardy. In case you haven't watched Jeopardy, some of the the clues are in the form of puns or a lot of hominem's things like that, where they'll they'll give a clue that you requires you to think outside of just this is this is the quote unquote question, or here's the answer. What is the question to that answer? Um? And it may be that it involves a little a little trick. You have to think in a creative way, and Watson was okay at that,

not nearly as good as the human champions were. So any question, any category that did depend on that, Watson had to work a little harder, I imagine to come up with an answer. Yeah, I probably think about it, like, Watson was really good at questions that you yourself could

easily solve by googling keywords. Yeah, and when you think about this cluster, this nineties server cluster with two thousand processor course, sixteen terabytes of memory, four terabytes of clustered storage, able to do a hundred eighty thousand gigabytes per second processing power, and it edged out the the Jeopardy champions. I mean some would say it beat them early soundly, but even so, you're looking at look at how much raw power was right, and this was one that that

you know, it's it's still outperformed Jeopardy champions. And just this one way, like perhaps on another day or another week, maybe the champions, the human champions, maybe they would have won against Watson, or you know, they challenge Watson to some other game like hop scotch. It's only they win. Although to be fair, I couldn't have won against those

Jeopardy champions, so well my argument is invalidated. Okay. I think we should break down computers and brains and start to think about them in terms of what each one does and how they do them differently. So we started off talking about the idea of using a computer as sort of a an analogy to think about a brain. It's like, yeah, brain is basically like a computer. Does that analogy hold not at all? And why doesn't it hold? Well?

I mean there are some similarities right in the sense that both can process information, sure, but the way that computers process information the way our brains process information, uh, in fact, saying brains, that's really pretty limited. The way our nervous systems process information is very different. Well, both do use electrical signals to transmit that information. I mean both are able to send messages via those electrical signals. Yeah.

Both they have input and output, right, yeah, yeah, you can. They have memory and you can add to that memory. Yeah, that's true. Um, both require energy in order to do this work. Yes, very true. Yeah, and both can be changed or modified. Yeah, we can learn, we can make our brains stronger, we can we can suffer injury or illness that can affect the ability for our brains to work. And computers you can you can change or modify those

pretty easily. That's pretty much what they've been designed to do. Okay, So, actually, at a bird's eye view, a brain is a lot like a computer. But it's when use zoom in that's when it really the different differences show up. So let's start with electricity. Sure, so a computer has electrical on and off switches, and that's how it manages information. Now, I don't think we even really understand fully how a brain manages information. But what's basically going on when we're thinking?

All right, so you've got you got neurons because your little brain cells, right, and your neurons. Uh, there are several different types of neurons, and what they do is they use electrochemical reactions. You've got these these various ions that get released within the neuron, which changes the electrical potential of the cell membrane. This ends up creating another

way of releasing some ions in a synapse. Synapse is where you have to nerve endings kind of uh, there's a gap, they're not actually touching one another, and you have a chemical message going between one to the other and it propagates throughout the nervous system. This way, it's very, very different from the way a computer works, where it's not an electrochemical response. There's nothing to do with any

kind of ions there. It's strictly electricity using electrons going through transistors to process information to represent numbers, really to represent zeros and ones, so very very different approach. Yeah. Another thing would be that, uh, computer instructions are executed in in binary code. Basically we understand the way information is encoded in a computer. How is information encoded in a brain. Well, that's a great question. So science doesn't

know yet. Think things like memories and stuff. We we've seen that as people think of things, certain neural pathways uh pop up, and in fact, neural pathways may represent things like memories, like the specific pathway represents the specific memory. However, every time you think back on that memory, that pathway may change in subtle ways, which one is another illustration of how memory is not an entirely reliable or infallible

source of information. And to suggest that we got to learn a lot more about the brain to really under stand how it works, because certainly that's not the way a computer works with memory. It's a very different model. So yeah, until we really understand more fully how the brain and neural pathways work, it's hard to really even

you know, describe it in any accurate terms. You're really just talking in generalities the stuff that we've observed so far, but we don't fully understand it's also harder to upgrade a brain. I mean you can't just add a second brain to yours, or I mean you know, you can't strengthen the neural connection. Yeah, yeah you can. You can learn, you can read, you can you can do a lot of things that that essentially nourish your brain. Well, good

nourishments also important, obviously, So that is good. You want some you want some energy in there. But yeah, you can do stuff that will strengthen your brain, but it's not as simple as you know, popping open the skull and sliding in another memory card. We've got an interesting note here that says computers are great at multitasking and brains most of the time aren't. And I think that's

basically true. But that might seem counterintuitive because we were just talking about how computers are far outstripped by brains in terms of parallel processing. Brains can have a lot of different things going on that you're making different connections while computer executes one instruction at a time. Really, it

depends on how you're looking at multitasking. If you're talking about a cognitive approach, like you're actually thinking about things like you are actively engaged, engaged in an activity or task or whatever. Then we're terrible at multitasking. And anyone who says that they're good at multitasking is probably lying to themselves and or to you, because statistically improbable, there's a very tiny number of human beings on this planet

who are actually supertaskers. They're not trying to deceive you on purpose. They're just texting at the same time they're talking to you, and so they they're not really they're they're so. So computers are very good at being able to run multiple programs at the time and continue those process sees. But depending on the type of computer you have, our computers at work, maybe not so much. But you know, typical good computer, yes, But people are not good at that.

So while we are able to do a lot of things in parallel, there's some things we're doing great with multitasking, but it's all the unconscious stuff, right, all the blinking of our eyes, the breathing, the heartbeats, the monitoring our body temperature, stuff that we don't have to think about. Our brains are fantastic at that kind of multitasking, but when it comes to actually stuff that we are thinking

actively on, not so much. Yeah, I'd suggests that computers are good at multitasking because computers are good at keeping tasks discrete. So you have one application running and it sends an instruction to the processor and executes it and comes back with the result, and then maybe another one sends an instruction, and it all happens so fast it seems like everything's going on at the same time, but the processes are kept to discrete. In your brain, it's

hard to keep different activities discrete. They bleed over into each other. If somebody's talking to you while you're taking notes on a sheet of paper, you'll end up writing down the words they're saying by accident, right, Or if you're listening to a lot of you have this confusion listening to a lot of music with lyrics in it, like, I can't. I can't do research and listen to the two songs with lyrics in them or else it. It just starts to distract me while I'm trying to read.

Even if I'm not actively listening, it does kind of been I can't even listen to something with a melody I listened to like a rain Generators online. I listened to the same melodies over and over again, because I've got my music my film score collection on and it has the same three composers, and I swear they just use the same score repeatedly. But anyway, one more thing we're going to touch on more later is that brains have the ability to repair and regenerate to a certain extent,

which computers don't without outside intervention. Yep. Uh. Now, you can, of course suffer irreparable brain damage, that certainly, because certainly in many different ways, but but in general, low levels of brain damage can be quote unquot repaired by by the brain just finding a different neural pathway. Sure, yeah, you can reroute the way that you would normally do a task, depending upon I mean, this is very specific case to case, but computers can't do this in general.

There are a lot of interesting research projects out there that are working on ways of making transistors that can do this kind of thing, where they can reroute a process so that if there were physical damage done to a chip it was it would be able to continue working. But this is not something that is naturally quote unquote naturally built into computers. Um. It's something that you know, people have had to innovate around. Whereas it's just a component of who we are as you know, that's one

of those things that brains can do. Yeah, okay, So how do machines actually work? If you want to look at the way a computer thinks and zoom way in when it's worrying, it's got the little machine brain going in circles, the hour glasses flipping or the little pin wheel, the little wheel, what's how happening there? Uh? The first thing I think we should talk about, I guess, is something called the von Neuman architecture that. Uh, if you have a laptop sitting in front of you, it's almost

definitely a von Neuman architecture machine. You know. If it's not, then you're working for someone pretty cool, possibly the military call us. The vast majority of typical computers are von Neuman. So. John von Neuman published papers in the nineteen forties about the requirements for a general purpose electronic computer, and those thoughts ended up becoming the architecture upon which we base

computers today. Most computers, the vast majority of computers. So he identified four main quote unquote organs for a general purpose computing machine, which were arithmetic, memory, control, and user interface. So ultimately This spoils down to what we think of as the CPU and memory and plus some some sort of user interface. But input and output, well not even input and output you're talking about really the two input

outputs what you get from this. But CPU and RAM, those are the two elements that you need in order to make any sense of anything. The other elements sure sure, sure yeah, control would be in that as well. Yeah, so CPU central processing unit. The job is to execute instructions, so essentially, think of that as some sort of mathematical process.

Increment this value by one, right, delete this value, multiply by whatever you know, multiplied by the closest prime number to that, to that number, anything along those lines, and the instructions will change as whatever it is you're doing changes. The Now, those instructions have to be executed upon something. It doesn't make sense to you for you to just say add You need things to add stuff too. So that's where the RAM comes in. The random axis memory.

This is where the computer holds information that it is going to execute instructions upon. So you've got the instructions that you you need, and then you've got the data that you need, and you do the two together and that's where you get the computation. Yeah. So in the memory, you might have slot A and slot B, and you could execute the instruction that says multiply what's in slot A by what's in slot B and put the answer

in slot C. Yeah. Uh. And so anytime you're talking about any kind of input into a computer, let's say that you're just playing a video game. Every time you're taking an action in that video game, it's sending an instruction down to the processor. Ultimately, actually it's sending lots of instructions. No single task is going to be just

one instruction and then you're done. But it then takes in all the data that needs it, executes the instructions in order, and then gives you the output that whatever the outcome is supposed to be. So when you press a Mario jumps, I think it's a B. I think is run anyway. Uh. That's the basic idea here. And also an important thing to note is that this series means that you are there are accessing information in sequence. Right.

It's not that you suddenly have a task and it just does everything all at once and everything gets completed and your classic CPU RAM relationship, you are executing instructions one after the other, which means that if you wanted a more powerful machine, you would have to get a faster processor. Right, that's the only thing you can do,

because it can just process those instructions and a faster time. Yeah, still one at the time, still doing like okay, we gotta do step one, step two, step three, step four, step five. It's not doing steps one through a thousand all at the same time. It's doing these all in sequence. So in order to have a faster machine, you just have to make a a microchip that can process stuff faster, which means that you need more energy, which also means

you're going to be generating more heat. And here's where we start running into the problem with von Neumann architecture. It's that in order for you to scale it up to really, really really high levels, you have to pour so much inner g into it, and you get so much excess energy in the form of heat coming out of it. You're losing so much energy in the form of heat that it's no longer efficient. It's fine for the lower powered machines, and by lower powered, i'm talking

about the stuff we tend to rely upon. But if you're talking about a truly powerful supercomputer. It doesn't quite measure up. Now, there are plenty of computers that use some degree of parallel processing. Sure, oh, sure, that's that's another way to squeeze a little bit of extra efficiency out of a system. Yeah. You can have a processor with multiple cores, for example, and those cores use things

like multi threading architecture. Multi threading means that you're able to divide up tasks so that each core of a processor can work on part of that task. The way I usually describe this is imagine that you have, uh, you have a really smart math student alright, like like genius level math student, and then you have eight math students who are above average. They're not geniuses, but they're good,

and you give both of them uh problems. So you give first all everyone gets the same math problem, and the genius, you would imagine, would finish that faster because the genius is just able to to process this and to go through the whole uh series of questions or whatever really really quickly. But then you give both of

them a different questions. You give the genius a question that's like a multi part problem, and each problem is independent of the other, so you don't have to solve one in order to know the answer for the number second one the group of eight, you give each person that group of eight one part of this multipart problem that you've given the genius. Now, individually, those eight people might be able to solve their smaller problem much faster

than the geniuses. And that's where this multi core processing comes in. It works great for problems that can be divided up like that. If it can't be divided up, it's not terribly useful. So this is another thing that we talked about with the potential for quantum computing down the line, where you have massively parallel systems which for some problems would be amazingly fast, like orders of magnitude faster than the fastest classical computer. For other problems, it

would not be any better than your average classical computer. Okay, So when when looking at them compared to brains, what are computers really good at? I'd say they're really good at any pre programmed wrote processing at scale and speed. So a human brain will never be able to do something like arrange a correctly formatted data set according to

a specific instruction as fast as a computer can. So imagine you've got like this huge unsorted list of number values and you've got like a hundred thousand of them, and the task is, we need to put these values in order from the lowest to highest, so ascending order. You cannot possibly compete with a computer. Your human brain will just not do it. The computer can do that so much faster and better than you, And it's hilarious.

Even if somehow you had some sort of savant like ability to sort all those numbers instantaneously or what would seem instantaneous to anyone else, you wouldn't be able to express it faster than a computer could. So even even if somehow your brain was able to keep up with it, you could not. At the very least, you couldn't express it. But yeah, I don't think there is a human especially if you're talking about a list that's you know, it

doesn't matter how long it is. If you've got a computer that's got a powerful enough processor, if you've got a list of ten thousand numbers and you sort it, even if it's a list of ten numbers, the computer will do it much faster. Part of this rote processing thing also is that computers are way better at following

rules consistently. You know, they don't get stuck on contextual information, like they're never in a bad mood and so therefore less effective and or or you know, they don't have a lack of caffeine or sleep that's messing up their processing. Or they're not influenced by what else is going on around them, distracted by clicking on links to things. And then and then it's four in the morning and you've got thirty five tabs open. You have it sorted, your list.

But there's a variation on this same scenario where a human might be much better than a computer. So I want to say, imagine the same thing. You've got a list of a hundred thousand numerical values and you need to you need to sort them in ascending order. But the list is not just numerals. It consists of numerals and numbers that are spelled out like five F I V, and the numbers are spelled out in six different languages,

so you've got five and sinko and other stuff. And sometimes there are typos, so it'll be like flive f l I V or clink oh, you know, whatever it is. The computer is stuck. Unless it has been told how to deal with this kind of information, unless it's been given a backup system that tells it what to do when it hits an anomaly, it cannot complete this task. The human brain might take a long time, but you

can do it well. Not only that, but you can also make this even more uh difficult for the computer by having visual representations of the number five that are not a numeral. So let's say you have a hand held up with five fingers spread. For a computer, what how would it know that the number is five versus say one as in one hand. I mean it's you know, this is one of the things where unless the computer is taught what that means, then it cannot it cannot count it. So uh that, in fact, is one of

the things brains are really good at. We're able to with our brains be able to learn a concept and not just that one instance of that concept, but be able to apply that across a more universal plane and recognize variations. Yeah, we are. Brains are very adaptive. I've said this before on this podcast, and I think I want to stick by it. Computers are really good at getting it right and doing it fast. But brains are better at making things work when there's trouble. Well, here's

another example, a very simple example. Let's say that I show you Joe a mug, and I tell you this is a mug, and it's just a you know, plain mug, maybe white in color. It's got nothing in it, it's empty. But then I show you a black mug. What is that? Yeah, you have never seen anything like that. Okay, for most normal humans, they would just say, okay, that's also a mug.

That's just a black mug. But let's say I showed you another white mug that's similar to the first one, but twice as large, it's different shaped handle, or you're showing it to me from a different angle, or there's something inside it, or I mean, there are all these different variations that. But once we learn a concept, we can apply that and recognize it when we encounter it again, even if it's not exactly the same shape, size, color,

has a different witticism about Monday's on. It could be whether whether there's Garfield or no Garfield, maybe there's Heathcliff, who knows, but you're able to recognize it for what it is, whereas with a computer it's a lot harder to do that. You might be able to set certain parameters with a computer and tell it this is what this thing is, but then if it encounters anything else that is that thing, but as a different set of parameters, it could be totally stumped. Yeah, okay, well let's talk

about then, how does the brain work? Now? Granted again the caveat that we don't really understand exactly how information is encoded in the brain or how the brain does information processing. But we have some ideas that at a certain level what's going on when we're thinking and and by wei, Joe means humanity and science, not just the

people sitting at this table. Yeah. I mean again, we're talking about those electrochemical signals right, Specifically, you're talking about sodium and potassium ions within a nerve cell that's changing these these uh the membranes electric potential. Uh. And then you've got the synapse, which is that little gap between two nerves cells they don't actually touch, where the the the exchange of chemicals happens, where it becomes this communication

between different nerves. Our nervous system is made up of these nerves. These nerves are running all through us right, we've got it's we've got our central nervous system. We've got the brain and the spinal cord. But our nervous system is more than just that, and so it's pretty complex system. And and again we don't even we can't

even fully explain what's going on. We know that there are certain regions of the brain that are responsible for specific functions, like you know, you say, like the visual cortex for example. We talk about these parts of the brain that we know are responsible, we don't fully understand

the mechanisms involved or the inter relation between all of them. Well, we know that it doesn't it doesn't happen like a computer where you have memory storage here and then processing here his hands up, which is very useful for radio A. A value goes from memory storage to my other hand where it gets processed, and then back to my left hand. Uh. I just thought I'd throw that in there, and then and then it goes to your ft, which is what people are looking at. That's the screen. Okay, No, uh,

it travels back and forth like that. What's going on in the brain. Um, well, we don't have knowledge of really discrete tradeoffs like that. Instead, what we have is a vastly interconnected system of cells that are all communicating back and forth with each other when the brain is thinking, We've got these cells lighting up with electrical and chemical signals action potentials, and they get traded off back and forth, and something's happening. Yeah. In fact, there's a lot of

debate and philosophy about what is going on. For a long time, the the prevailing model was this idea, a dual model where you had the mind and the body. Right, so the mind is totally separate from your body exactly. It's kind of the material. The mind might be dependent upon the brain, but that would be about as far as they would go with that, saying okay, the brain, Yes, sure, if you damage the brain, you damage the mind, so therefore there's some connection there, but they wouldn't go further

beyond that. But there's also another philosophy, embodied cognition, which is really more about how the body is part of our nervous system, to the point where a lot of our cognition is dependent upon our experience with our bodies. If we didn't have those bodies, we would not think the way we do. The reason we think the way we think is because of our physical forms, at least in some part. So it's interesting how this manifests in

different ways. One of the popular ways that people have pointed out is through the development of our language and metaphors. Oh yeah, yeah. For for example, if you're talking about affection or anger, you you talk about it being warm, um, and that, you know, is theorized that it could be because you know, affection you get when you're a kid and someone hugs you, and that's that's a warm thing, and anger is a physiological response that happens in your

body that that raises your body temperature. Right, Or if you're feeling happy, you're up, and if you're feeling sad, you're down. You know, these are interesting ideas that again without the body, if you if the mind were totally separate, why would we have these different metaphors. Now, there are a lot of different arguments about how whether or not that that actually indicates embodied cognition, but it is an interesting idea and it does mean that again the way

we think would be very different from computers. I think A very interesting way of thinking about it is the classic quote from Marvin Minsky, who said the mind is what the brain does. Yeah, it's a it's a process that's being conducted by this organ. Sure, sure, and you know I do want to put in there. I mean like you've you've got neurons all over your body, you know, you've you've got them in your fingers and and so it's it's really all when you really what binds us

and penetrates us. And I'm sorry, I think of the force it's similar. Yeah, no, very similar, you know it's I mean, really it's it's it's the intrinsic component that allows us to think, even though we can't fully understand exactly what the processes, but what we do understand is

that that's not how computers think. You know, that's partly you know, computers, even if you were to say that a computer thinks it processes information in a very linear, serialized way, if we're talking about von Neuman architecture at any rate. Okay, well, now let's look at the flip side. Earlier, we talked about how computers are much better than human brains it doing preprogrammed sort of arithmetic based tasks or any kind of information processing that has correctly formatted data

sets and needs to be done at scale. Really fast. What are human brains better at. Well, they're they're definitely brains are much more energy efficient. Yeah, so let's take a look. If you look at the k supercomputer, which is one of the most powerful supercomputers in the world. Right, you're talking about a supercomputer that's pulling in about the

equivalent energy that would power ten thousand homes. Yeah. Now, if our brains required that, we would never stop eating ever, and even then we wouldn't be able to think very much. I mean I already never stopped eating. So that's right. The moral of the story is, never stopped eating and you'll be a genius. Um Sources vary about the exact number of watts that it takes the human brain to run on. Yeah, yeah, I've seen estimates in the range

of twenty watts to twenty five watts. I read a Scientific American article that estimated about twelve point six watts. We're talking about not a whole lot of juice here, mean, yeah, yeah. The the number that I've seen given is is that silicon computers take some forty thou times more energy to run than the human brain does. So even though these computers are better at certain tasks, they definitely are not better. As far as as efficiency goes, the brain is incredibly efficient.

It's also I mean it's volumetric. It's a really dense processing power all by itself. But uh so that's one thing we can look at brains being better. Another thing is this whole idea about being able to learn. Yeah, I mean you've heard of the idea of a learning machine or a learning computer. The reason that phrase exists is because that's kind of a unique and ideal situation for a computer. Generally, computers don't learn. They're told what

to do. They seem smart, but they're dumb. No, they have no they have no creativity or initiative or adaptivity. They're just very good at doing what they're told. So, unlike computers, brains have what's called neuroplasticity, which is the ability to for a brain to change itself in response to injury or in a reaction to a new situation.

So simple computers do not learn anything or change their behavior, and less a programmer specifically instructs them to do it, they don't naturally make change just to themselves, and brains do. So what if you want a supercomputer that doesn't just

do exactly what you tell it to do. What if you want to create an artificial general intelligence program that updates its behavior the way we do when we say we learn about something, We learned that something we were doing, maybe the way we talk is considered rude by someone in a way we didn't think about, and we update our brain with that new information and we just don't act that way anymore, or we act that way all the time always, Right, if you're Jonathan Yes, um or

and this is really helpful. What if you want a computer to help you solve a problem when you don't fully understand the nature of the problem or how to solve it. I mean, with simple computers, you're stuck there. They're very useful tools, but they're not going to help you define the boundaries of the problem you're trying to solve, um and some But a brain can do that, right. You can talk through a problem with a person. You can sort of figure out, oh no, here's what we're

really talking about. There's there's innovation with a brain, right, you can you can innovate as a person. If we couldn't, we would not be here talking into microphones, you know, recording a podcast right now, none of that would be possible, whereas with computers, Uh, innovation is not something you would typically see from a computer. A computer would simply follow instructions. They do it really well, but not come up with new ways of coming out a problem with a classic

von Neumann approach computer. Right, here's another one. So dealing with real world data. Real world data, I mean like images and sounds and smells and you can touch. Yeah. Yeah, So computers are really good at dealing with, as we've said, correctly formatted abstract data like numbers or strings of asky characters. But computers are not good at dealing with the kind of data we get through our senses. So we can force them to do it, but it's top heavy and

it involves a lot of inefficiency and hand holding. So imagine a computer is trying to look at a human face and answer a question like does this person have a mustache? Not that hard for humans to do, but for a computer, you could write a program to do this. I have no doubt that it would be possible to make a mustache detecting computer, but it would take a lot of work and it wouldn't be as efficient at the task as just a human would be So to

do it, what what would you need to do? You need to take a picture, convert that into digital data, like a bitmap image. Then you'd have to program the computer to recognize the different parts of the face in terms of the digital bitmap, and then account for differences in how the thing might look, so differences in hair color and skin color, and what if all the pictures aren't head on? And how do you keep it from

thinking an eyebrow is not a is a mustache? And those those kinds of patterns are very difficult to teach a computer, um the definite missions of Yeah, I would suggest that one way of going about it is to build in some GPS capabilities so that if it's in Brooklyn, it's already probability that, yes, oh Jonathan, your hipster jokes could only be produced by a human brain. For that, you should all be thankful. Okay, but seriously, I mean, so, computers can do things like this, but brains are much

better at them. They're much more efficient, and they have much more versatility in these kinds of tasks. This is coming back to that whole idea about showing Joe the mug and then Joe is able to recognize a mug, whether the next mugs he encounters looks identical to the first one or or different, he understands what a mug is, right right, It's it's really hard to teach computers about context.

And I know I said a minute ago that that's on the plus side of machines, you know, not getting distracted by stuff or confused about extra information, but really can go either way depending on what you're trying to do. You know, if you're trying to solve a logic problem, then removing the context can be really useful point machines, But if you're trying to react appropriately in a conversation,

context is everything also. Okay, we talked earlier about how computers are really good at doing things faster than human brains in many different contexts, but the brain is better at at speed in a lot of other ways. You know, you can program a computer to do lots of difficult things given enough time and power. But for example, even a mouse cortex runs about nine thousand times faster than

home computer simulations of a mouse cortex function. Sure, So if you're trying to make a computer behave like a brain, it turns out that's really complicated to do, and the computer is not good at behaving like a brain, at least not on the scale that an actual biological creature is capable. It's going to use way more energy and

take way more time, right right, Yeah. Um, the state of the art in terms of speed is from that k supercomputer that we were talking about earlier, and uh, in the beginning of it modeled one second of one percent of human brain activity in a mirror forty minutes. Yeah, I mean, that's how complex we're talking about here. Granted, also that complexity stems from a large amount of us not knowing what the heck is going on on the

grand scheme of things. So if we were able to make a computer behave more like a brain, it would be really interesting to see what that would computer would be capable of doing. And that's what we're going to talk about in our very next podcast what oh yeah, look right there in the notes. Tune in next time to hear the follow up to this discussion where we talk about how to make computers more like brains. Yeah,

and it's uh, it's an interesting discussion. We'll even talk a little bit about why would you want to do that, so join us for that. Also, if you have any suggestions for future topics, then I highly recommend you get in touch with us and let us know about it. You can say as an email right addresses f W Thinking at discovery dot com, or you can drop us a line on Twitter or Facebook or Google Plus. Our handle it all three is FW Thinking and we will

talk to you again really soon. For more on this topic and the future of technology, visit forward thinking dot com, brought to you by Toyota Let's Go Places