WEBVTT - Ep78 "Does your brain have one model of the world or thousands?" 0:00:05.320 --> 0:00:09.399 What is special about the wrinkly outer layer of the brain, 0:00:09.480 --> 0:00:11.640 the cortex, And what does this have to do with 0:00:11.680 --> 0:00:15.720 the way that you come to explore and understand the world. 0:00:16.079 --> 0:00:17.520 And by the way, why do you. 0:00:17.480 --> 0:00:20.760 See a whole image when you open your eyes even though. 0:00:20.680 --> 0:00:22.799 Each part of your visual. 0:00:22.480 --> 0:00:26.720 Cortex has access to only a tiny bit of the image. 0:00:27.280 --> 0:00:30.040 And for that matter, the brain is divided into different 0:00:30.080 --> 0:00:32.639 areas for sight and sound and touch and so on. 0:00:33.159 --> 0:00:36.559 And so why when you're petting a cat, why does 0:00:36.560 --> 0:00:40.040 the cat seem unified? Why doesn't the site of the 0:00:40.080 --> 0:00:44.159 cat seem separate from the purring and the feel of 0:00:44.240 --> 0:00:47.640 the fur. Can we build a new model of how 0:00:47.640 --> 0:00:50.720 the brain works and in what ways is what the 0:00:50.760 --> 0:00:55.640 brain doing something very different than what's happening in current AI. 0:00:59.040 --> 0:01:02.680 Welcome to Intercouse with me David Eagleman. I'm a neuroscientist 0:01:02.680 --> 0:01:07.080 at Stanford and in these episodes we sail deeply into 0:01:07.080 --> 0:01:10.040 our three pound universe to understand. 0:01:09.600 --> 0:01:12.280 Why and how our lives look the way they do. 0:01:24.040 --> 0:01:27.720 Today's episode is about a new model of the brain 0:01:28.240 --> 0:01:31.160 developed by my friend and colleague, Jeff Hawkins, and we'll 0:01:31.160 --> 0:01:33.920 get into an interview with him shortly but let me 0:01:33.959 --> 0:01:38.200 preface by saying that for centuries people have stared at 0:01:38.200 --> 0:01:41.240 the brain and tried to figure out how this thing works. 0:01:41.280 --> 0:01:44.360 Because when you stare at it, it's just a huge 0:01:44.840 --> 0:01:45.960 lump of cells. 0:01:46.160 --> 0:01:46.880 You can see that. 0:01:46.880 --> 0:01:51.400 There's a wrinkled layer on the outside. And when people 0:01:51.400 --> 0:01:54.480 dissect that, they can see that that part is about 0:01:54.600 --> 0:01:58.520 three millimeters thick, and it looks a little different, looks grayer. 0:01:58.920 --> 0:02:01.280 And so that part is called the gray matter. And 0:02:01.360 --> 0:02:05.520 we call this the cortex, which means bark, like tree bark. 0:02:06.000 --> 0:02:09.160 And the stuff below that thin layer is called white matter. 0:02:09.200 --> 0:02:13.040 And it looks white because the tiny data cables coming 0:02:13.080 --> 0:02:16.240 off the cells, the axons, these are wrapped in a. 0:02:16.200 --> 0:02:19.120 Little sheath called myelin, which makes it look white. 0:02:19.320 --> 0:02:22.799 Okay, Now, what you immediately noticed by looking at brains 0:02:22.919 --> 0:02:26.960 across different mammals is that all the stuff you find 0:02:27.120 --> 0:02:31.520 under the cortex, all the sub cortical stuff, looks essentially 0:02:31.600 --> 0:02:34.960 the same. Horses and elephants and mice. They all have 0:02:35.080 --> 0:02:37.799 the same architecture going on that we do. 0:02:37.840 --> 0:02:38.440 They all have. 0:02:38.400 --> 0:02:42.840 A thalamus and hippocampus and cerebellum and so on. But 0:02:42.919 --> 0:02:46.560 there's one thing that really distinguishes us from our cousins, 0:02:47.040 --> 0:02:51.200 and for that we return to the gray matter, the cortex. 0:02:51.360 --> 0:02:54.800 It's not that our cousins don't have a cortex. What 0:02:54.960 --> 0:03:00.760 distinguishes us is the absolute enormity of our cortex. We 0:03:00.919 --> 0:03:05.400 humans have a ton of this stuff. So take four 0:03:05.440 --> 0:03:08.800 pieces of paper from your printer and place them next 0:03:08.800 --> 0:03:11.120 to each other to make one really large piece. 0:03:11.560 --> 0:03:13.440 That's how much cortex. 0:03:13.080 --> 0:03:16.760 A human has. If you were to spread out the wrinkles. Now, 0:03:16.880 --> 0:03:20.040 our nearest cousins, the great apes only have about one 0:03:20.080 --> 0:03:22.799 piece of paper worth, and most mammals have a lot 0:03:22.880 --> 0:03:26.880 less than that. So something about the story of the 0:03:27.080 --> 0:03:30.600 runaway human success has to do with the fact that 0:03:30.639 --> 0:03:34.200 we have way more cortex for our body size than 0:03:34.320 --> 0:03:38.120 any other creature. And side note, I'm really talking about 0:03:38.200 --> 0:03:42.280 what's called the neocortex or new cortex, because we also 0:03:42.280 --> 0:03:45.720 have a little bit of paleocortex or old cortex. But 0:03:45.760 --> 0:03:48.320 the thing that really makes us outstanding is the amount 0:03:48.360 --> 0:03:50.280 of neo cortex that we have. 0:03:51.400 --> 0:03:54.040 But what is this neocortex doing. 0:03:54.920 --> 0:03:58.280 Well, if you look at any neuroscience textbook, you'll see 0:03:58.320 --> 0:04:01.240 that this part of the brain, the cortex, is often 0:04:01.720 --> 0:04:05.760 drawn with different colored regions like this red region over 0:04:05.800 --> 0:04:08.000 here is devoted to vision, and this green one is 0:04:08.040 --> 0:04:10.760 devoted to hearing, and this yellow one to touch and 0:04:10.800 --> 0:04:14.320 so on. But something I've been obsessed with and write 0:04:14.360 --> 0:04:16.640 about in my latest book, Live Wired, is that this 0:04:16.720 --> 0:04:19.000 is the wrong way to think about it, because the 0:04:19.000 --> 0:04:22.240 neocortex is remarkably flexible. 0:04:22.279 --> 0:04:23.599 It's not a fixed map. 0:04:24.160 --> 0:04:27.440 If you are born blind, the part of your cortex 0:04:27.520 --> 0:04:30.000 that we would have thought of as visual cortex gets 0:04:30.040 --> 0:04:33.720 taken over by hearing and touch and so on. Now 0:04:33.839 --> 0:04:35.400 let me just be really clear what I mean by 0:04:35.480 --> 0:04:39.679 taking over. The neurons there are the same The cortex 0:04:39.760 --> 0:04:42.479 looks exactly the same from the outside, but the function 0:04:42.600 --> 0:04:46.760 of those particular neurons is now not visual. They have 0:04:46.839 --> 0:04:50.120 nothing to do with visual information anymore. Now that same 0:04:50.240 --> 0:04:55.200 neuron instead of firing when it detects a moving object, 0:04:55.600 --> 0:04:58.880 now it responds to a touch on your. 0:04:58.720 --> 0:05:03.479 Toe, or hearing a B flat note or whatever. So 0:05:03.600 --> 0:05:05.000 the little labels that. 0:05:04.920 --> 0:05:08.240 We draw onto the brain, these maps that we impose, 0:05:08.960 --> 0:05:09.400 these are. 0:05:09.320 --> 0:05:11.240 Actually massively flexible. 0:05:11.480 --> 0:05:13.320 And as you may know, I gave a talk at 0:05:13.480 --> 0:05:15.520 TED about this a while ago, where I showed that 0:05:15.880 --> 0:05:19.680 you can feed in new kinds of information, let's say 0:05:19.680 --> 0:05:22.440 through the ears or the skin, and the brain will 0:05:22.480 --> 0:05:25.320 figure out how to deal with that data. It will 0:05:25.760 --> 0:05:30.200 flexibly devote part of its cortical real estate to that. 0:05:30.839 --> 0:05:34.360 And this line of thinking led some scientists, like Vernon 0:05:34.360 --> 0:05:37.920 mount Castle some decades ago to realize that the cells 0:05:37.920 --> 0:05:39.839 of the cortex are. 0:05:39.839 --> 0:05:41.120 A one trick pony. 0:05:41.800 --> 0:05:46.840 No neuron is inherently a visual neuron or a neuron 0:05:46.880 --> 0:05:50.120 devoted to hearing or touch or smell or taste or 0:05:50.160 --> 0:05:54.039 memory or whatever. All parts of the cortex are perfectly 0:05:54.120 --> 0:05:58.120 capable and willing to take on any job. So that 0:05:58.240 --> 0:06:02.200 suggests they're all running some sort of basic algorithm. And 0:06:02.200 --> 0:06:05.400 it doesn't matter what kind of data you feed in. 0:06:05.440 --> 0:06:08.479 Different parts of the cortex will say cool, I'll build 0:06:08.560 --> 0:06:11.479 a representation of that data. I don't care if it 0:06:11.480 --> 0:06:15.520 comes from photons or air compression waves or temperature or whatever. 0:06:15.960 --> 0:06:19.520 I'm on the job here to build an understanding of 0:06:19.560 --> 0:06:24.680 whatever is coming in locally. Now, it's not individual neurons 0:06:24.680 --> 0:06:28.560 that are building models, but instead groups of many tens 0:06:28.680 --> 0:06:33.760 of thousands of neurons arranged in a six layered cylinder. 0:06:33.920 --> 0:06:37.080 So think about this like you're a geologist and you 0:06:37.200 --> 0:06:40.480 drilled out a cylinder of rock and you saw six 0:06:40.600 --> 0:06:43.120 layers in it, six sedimentary layers. 0:06:43.480 --> 0:06:45.359 That's what the neocortex looks like. 0:06:45.520 --> 0:06:49.680 Six layers. And it's built out of these columns which 0:06:49.720 --> 0:06:52.839 have the same types of neurons with the same connection 0:06:53.000 --> 0:06:56.760 patterns in each column. And so think about the cortex 0:06:57.160 --> 0:07:00.440 as being made of lots of these columns, like taking 0:07:00.880 --> 0:07:03.600 hundreds of thousands of grains of rice and standing them 0:07:03.680 --> 0:07:05.760 up on their end and packing them all next to 0:07:05.800 --> 0:07:09.560 each other. Now, people have known about cortical columns for 0:07:09.600 --> 0:07:13.600 many decades since Vernon Mountcastle first discovered these in nineteen 0:07:13.680 --> 0:07:17.880 fifty seven. But recently someone has pulled together several different 0:07:17.960 --> 0:07:22.640 threads to propose how this could underlie what the cortex 0:07:22.760 --> 0:07:26.800 is all about. And that's someone is Jeff Hawkins and 0:07:26.840 --> 0:07:30.040 his team. And so I met with Jeff in my studio. Now, 0:07:30.200 --> 0:07:33.000 Jeff is one of my favorite people because he does 0:07:33.600 --> 0:07:37.000 theoretical neuroscience. He really tries to figure out the big 0:07:37.080 --> 0:07:40.360 picture of what the brain is doing. Now, Jeff has 0:07:40.400 --> 0:07:43.160 a very interesting history, so I'll just mention that in 0:07:43.200 --> 0:07:46.400 the nineteen eighties he was a graduate student at Berkeley, 0:07:46.400 --> 0:07:50.119 where he proposed a PhD thesis on a new theory 0:07:50.160 --> 0:07:53.560 of the cortex, but his proposal was rejected, and so 0:07:53.640 --> 0:07:58.000 he ended up pursuing his vision for mobile computing instead, 0:07:58.080 --> 0:08:01.640 and in nineteen ninety two he launch the company Palm, 0:08:01.840 --> 0:08:05.160 which made the Palm Pilot. If you remember that, this 0:08:05.320 --> 0:08:08.720 was this little handheld device and you could write on 0:08:08.760 --> 0:08:11.880 it with a stylus and it would translate your handwriting 0:08:11.880 --> 0:08:14.520 into text. And you can use this for your address 0:08:14.520 --> 0:08:17.119 book and your calendar and your contacts and note taking. 0:08:17.440 --> 0:08:21.400 This was the first entrant into the world of portable computing, 0:08:21.480 --> 0:08:21.720 and it. 0:08:21.680 --> 0:08:22.679 Really changed the world. 0:08:23.480 --> 0:08:27.320 Anyhow, A decade later, Jeff returned to his original love, 0:08:27.360 --> 0:08:31.440 which was theoretical neuroscience, trying to figure out what's going 0:08:31.520 --> 0:08:34.440 on with the brain, and he wrote a book in 0:08:34.440 --> 0:08:37.800 two thousand and four called on Intelligence, which was very 0:08:37.880 --> 0:08:40.640 influential on me and lots of other thinkers I know. 0:08:40.920 --> 0:08:43.360 So I was very excited when Jeff recently came out 0:08:43.400 --> 0:08:47.319 with his next book that represents his last decade and 0:08:47.400 --> 0:08:51.040 a half of research. It's called One Thousand Brains, a 0:08:51.080 --> 0:08:54.559 New Theory of Intelligence, and it describes his framework for 0:08:54.640 --> 0:08:57.880 thinking about the brain. So, without further ado, let's dive 0:08:57.920 --> 0:09:06.560 into a very cool new model of the brain. Okay, Jeff, 0:09:06.600 --> 0:09:09.079 So you are a theoretician. You think about the brain 0:09:09.080 --> 0:09:11.840 from a high level. We're in this era now of 0:09:12.000 --> 0:09:14.400 AI where AI is doing all kinds of things that 0:09:14.440 --> 0:09:17.200 are amazing and no one expected. But you see the 0:09:17.280 --> 0:09:19.920 brain as being very different from what is going on 0:09:19.960 --> 0:09:22.520 with let's say, large language models. So tell us about that. 0:09:22.840 --> 0:09:26.480 That's absolutely true. You know, the current AI wave is 0:09:26.480 --> 0:09:29.200 really amazing, but those models don't work at all like 0:09:29.280 --> 0:09:32.160 the brain. And I think you could start with one 0:09:32.360 --> 0:09:37.720 really fundamental difference. Brains work through movement. We move our 0:09:37.760 --> 0:09:40.640 bodies through the world. We move our hands over objects 0:09:40.679 --> 0:09:42.520 to touch and learn what they are. We move our 0:09:42.520 --> 0:09:45.600 eyes constantly, so the inputs of the brain are constantly changing, 0:09:45.640 --> 0:09:48.800 but mostly because we're moving through the world. And the 0:09:48.880 --> 0:09:51.520 term for that is a centory motor system. And the 0:09:51.559 --> 0:09:54.560 brain can't understand its inputs unless it knows how it's 0:09:54.559 --> 0:09:57.280 moving through the world. So we learn by exploring, by 0:09:57.679 --> 0:10:01.880 moving different places, picking things up, touching, so on, and 0:10:01.920 --> 0:10:05.120 that's all Animals that move in the world learn this way. 0:10:05.320 --> 0:10:07.760 So this idea that the brain is the central motor 0:10:07.800 --> 0:10:10.320 system has been known back in the late eighteen hundreds, 0:10:10.480 --> 0:10:14.400 but it's pretty much ignored by everybody. But it leads 0:10:14.440 --> 0:10:18.080 to a very fundamental different way of how we acquire 0:10:18.120 --> 0:10:21.520 knowledge and how knowledge is represented in the brain. Whereas 0:10:22.080 --> 0:10:25.000 today's AI is most of it's built on well deep 0:10:25.080 --> 0:10:28.280 learning of transformer technologies, which are essentially we feed data 0:10:28.320 --> 0:10:31.360 to it. We don't it doesn't explore it, and we 0:10:31.400 --> 0:10:34.280 feed to large language models. We just feed a language. 0:10:34.400 --> 0:10:37.280 So there's no inherent knowledge about what these words mean, 0:10:37.400 --> 0:10:40.720 only what these words mean in the context of other words. Right, 0:10:40.760 --> 0:10:43.320 But you and I can pick up a cat and 0:10:43.400 --> 0:10:45.640 touch it and feel it and know this warmth, and 0:10:45.640 --> 0:10:48.360 we understand how its body's moved because no one has 0:10:48.440 --> 0:10:52.240 to tell us that. We just experience it directly. So 0:10:52.880 --> 0:10:56.880 this is a huge gap between brains. Pretty much all 0:10:56.920 --> 0:11:00.080 brains work by century motor learning and almost all oh 0:11:00.559 --> 0:11:03.920 it doesn't. And you can just peel the layers apart 0:11:03.960 --> 0:11:05.920 and see what the differences are, and it makes a 0:11:06.040 --> 0:11:08.679 huge difference. So I'm not a fan. I'm a fan 0:11:08.760 --> 0:11:10.920 of AI today, but I don't think it's the future 0:11:10.960 --> 0:11:12.960 of AI. I don't think it's going to get you 0:11:13.080 --> 0:11:15.480 to what people really want or truly intelligent machines. 0:11:16.840 --> 0:11:19.000 Okay, terrific, And we'll dive into that more in a 0:11:19.040 --> 0:11:19.480 little bit. 0:11:19.559 --> 0:11:19.760 Now. 0:11:19.960 --> 0:11:23.559 When we look at let's say the human brain, there's 0:11:24.000 --> 0:11:26.760 lots of areas that we can point to. There's the 0:11:26.920 --> 0:11:31.199 cortext and wrinkly outer bit, there's all these subcortical areas. 0:11:31.440 --> 0:11:35.640 When you think about intelligence and the stuff that we're 0:11:35.640 --> 0:11:37.280 going to talk about today, what is the part that 0:11:37.320 --> 0:11:38.040 you concentrate on. 0:11:38.160 --> 0:11:41.319 Well, we concentrate first and foremost in the New York cortext, 0:11:41.360 --> 0:11:43.560 which is about seventy five percent of the volume of 0:11:43.600 --> 0:11:45.120 your brain. I mean it's what you see, as you said, 0:11:45.360 --> 0:11:46.760 if you can take a scope, and that's what you 0:11:46.800 --> 0:11:50.240 see in the New York cortext. And so it's a pretty 0:11:50.280 --> 0:11:53.600 dominant part of what we think of intelligence. You can't 0:11:53.640 --> 0:11:56.120 consider it completely on its own. I mean, it's connected 0:11:56.160 --> 0:11:59.000 to all these other things. And so we also study 0:11:59.080 --> 0:12:02.160 those other things and in service to the new or cortex. 0:12:02.320 --> 0:12:05.040 So we study the thalmbists, and we study the cerebellopment, 0:12:05.080 --> 0:12:07.320 We study the basic just because you have to know 0:12:07.360 --> 0:12:12.640 how the cortex workship these other things. But our primarily 0:12:12.679 --> 0:12:15.400 our goal and many neurosign this goal is to understand 0:12:15.440 --> 0:12:18.360 the New York cortex, because that's what mammals have. We've 0:12:18.360 --> 0:12:21.280 got a big one. You know, everything we think, most 0:12:21.320 --> 0:12:23.560 of what we think about being intelligent, about our ability 0:12:23.600 --> 0:12:25.960 to understand the world and generate language and see and 0:12:26.040 --> 0:12:28.080 hear and and so on, is the New York cortext 0:12:28.160 --> 0:12:31.640 not one hundred percent, but most of it. Unfortunately. Also, 0:12:31.800 --> 0:12:33.520 not only is it the biggest structure, but it's a 0:12:33.679 --> 0:12:36.800 very very regular structure. So you can look at this thing, 0:12:36.800 --> 0:12:38.760 the New or Courts is like a sheet of cells. 0:12:38.880 --> 0:12:41.600 It's you know, it's like a size of a large 0:12:41.600 --> 0:12:44.439 dinner napkin and only a few millimeters thick, and it 0:12:44.480 --> 0:12:48.520 gets wrinkly because it's stuck it in your head that 0:12:48.960 --> 0:12:53.239 everywhere you look on it, it looks remarkably complicated and remarkably 0:12:53.280 --> 0:12:56.240 the same. So the areas are doing vision, look like 0:12:56.520 --> 0:12:59.880 the areas are doing languages, look like the areas are 0:12:59.880 --> 0:13:02.520 doing touch. Look the areas that are doing everything. Really 0:13:02.960 --> 0:13:05.839 and so there's been long speculated that there's sort of 0:13:05.880 --> 0:13:10.480 a common algorithmic principle that's applying to everything everything we do, 0:13:10.520 --> 0:13:12.679 all of our sensory inputs, all of our thinking, all 0:13:12.720 --> 0:13:16.160 the language, it's hard to believe, but the evidence is overwhelming, 0:13:16.800 --> 0:13:19.560 and so our research has really been to understand what 0:13:19.720 --> 0:13:23.679 is that that algorithm, corical algorithm, often referred to as 0:13:23.679 --> 0:13:27.640 a cortical column. You know, this repeated structure that seems 0:13:27.640 --> 0:13:29.760 to underline vision and hearing and touch and thought and 0:13:29.800 --> 0:13:33.760 everything we do. And that's that's just just an appealing 0:13:33.800 --> 0:13:35.920 thing to try to understand. And we've cracked it. We've 0:13:35.960 --> 0:13:38.160 actually we actually cracked it. We understand what's going on. 0:13:38.280 --> 0:13:38.880 That's awesome. 0:13:38.920 --> 0:13:41.640 Okay, so a couple of things, right, So the way 0:13:41.679 --> 0:13:43.959 I sometimes phrase this to people is that if I 0:13:44.000 --> 0:13:46.720 had a magical microscope and could show you a part 0:13:46.720 --> 0:13:48.360 of the brain and you can see all the activity 0:13:48.480 --> 0:13:51.680 running around in the cortext there, could you tell me 0:13:51.800 --> 0:13:54.599 is that visual cortex are auditory or somatosensory? And the 0:13:54.640 --> 0:13:56.400 answer is you couldn't tell me, and I couldn't tell you. 0:13:56.400 --> 0:13:59.120 Guys, Right, it all looks the same, and there's a 0:13:59.280 --> 0:14:01.240 and there's as you know, oh, there's these experiments people 0:14:01.240 --> 0:14:02.840 have done where well, first of all, if you have 0:14:02.920 --> 0:14:05.319 trauma to one part of the cortext other parts will 0:14:05.320 --> 0:14:08.800 pick up the same function. You can also people re 0:14:08.960 --> 0:14:11.280 routed sensory and puts the different parts of the cortext 0:14:11.559 --> 0:14:13.280 in animals and they seem the work. 0:14:13.559 --> 0:14:17.480 So, for example, you have visual information instead of going 0:14:17.480 --> 0:14:19.920 to the back of the visual cortex, that gets rerouted 0:14:19.960 --> 0:14:24.520 to the auditory cortex, and that auditory cortex becomes visual cortext. 0:14:24.640 --> 0:14:30.080 Right, It's incredibly powerful and flexible system. And mammals, you know, 0:14:30.160 --> 0:14:32.800 all we have we have a set of sensors, quite 0:14:32.800 --> 0:14:34.440 a few action more than most people think because the 0:14:34.480 --> 0:14:37.800 skin there's a lot of different sensors. But other animals 0:14:37.880 --> 0:14:40.680 have different sensors and they have cortext too, And so 0:14:41.920 --> 0:14:45.760 there's seems to be this universal algorithm that can be applied. 0:14:46.440 --> 0:14:50.000 And now we know it's a century motor algorithm. It 0:14:50.000 --> 0:14:52.840 can be applied, and and we've spent decades trying to 0:14:52.840 --> 0:14:55.480 figure this out and we've cracked it. Oh that's awesome. 0:14:55.760 --> 0:14:57.520 Just before you tell us about that, So tell us 0:14:57.520 --> 0:14:58.880 what a cortical column is. 0:14:59.000 --> 0:15:01.240 Okay, So imagine we talked about. The near cortex is 0:15:01.240 --> 0:15:04.800 a sheet of cells like three milimeters stick. A cortical 0:15:04.920 --> 0:15:07.600 column is a little section of that going through the 0:15:07.640 --> 0:15:11.480 three free milimeters. It's convariating with from a third of 0:15:11.480 --> 0:15:14.040 a millimeters to a millimeters in diameter. It's it's it's 0:15:14.040 --> 0:15:15.920 a it's not something you would see. It's not like 0:15:15.960 --> 0:15:18.520 it's sitting there to be plucked out. But we know 0:15:18.560 --> 0:15:21.920 they exist, and so within that, let's let's say it's 0:15:21.960 --> 0:15:25.600 a three millimeters tall and a half millimeters wide cylinder 0:15:25.640 --> 0:15:30.320 that goes across the cortex that contains all the neural 0:15:30.400 --> 0:15:33.560 machinery that you would see anywhere in the cortex and 0:15:33.560 --> 0:15:36.880 in each cortical column, because they look like a little 0:15:36.880 --> 0:15:38.520 grain of rice in some sense. Why you can imagine 0:15:38.560 --> 0:15:40.440 what's a little brains of ice stacked next to each other. 0:15:41.160 --> 0:15:44.640 Each cornico column gets input from some well in parts 0:15:44.640 --> 0:15:47.320 of the bend they get from some pats of sensory input, 0:15:47.400 --> 0:15:49.520 so from pats to the retina, from pats to the 0:15:49.560 --> 0:15:53.600 cochlea of patrio skin. Other parts get information from parts 0:15:53.640 --> 0:15:56.560 of the near cortex. So the cortex is connected to cortex, 0:15:57.200 --> 0:15:58.960 but each one is looking at a small If you 0:15:58.960 --> 0:16:01.640 think about the primary entry regions of the cortex, which 0:16:01.640 --> 0:16:06.240 are quite large, they're getting input from a small sensory area, right, 0:16:06.320 --> 0:16:09.400 And so people used to think that, well, if this 0:16:09.560 --> 0:16:11.480 bil colm is only getting input from a small part 0:16:11.480 --> 0:16:13.840 of the rent. Now, it can't really doing very much right, 0:16:13.880 --> 0:16:15.920 It can't be very smart. All you could do is 0:16:15.960 --> 0:16:18.880 process a little piece of information there, and therefore maybe 0:16:18.920 --> 0:16:20.760 it's going to detect an edge or something like that, 0:16:20.800 --> 0:16:22.560 and there's a lot of evidence for that. But we 0:16:22.600 --> 0:16:27.560 now know what happens is that the cortical comms they 0:16:27.600 --> 0:16:30.800 get input from over time, from different parts of the world. 0:16:30.880 --> 0:16:33.160 So the eyes are moving like three times a second, 0:16:33.360 --> 0:16:35.320 and so that cortical comms may be looking at three 0:16:35.360 --> 0:16:39.640 different things every second, and it can integrate how the 0:16:39.720 --> 0:16:42.240 sensor is moving, how your eyes are moving, with what 0:16:42.320 --> 0:16:45.400 it's sensing to build models that are much larger than 0:16:45.400 --> 0:16:47.840 it can sense. And the same way that you could 0:16:47.840 --> 0:16:51.280 take your finger in a dark room and say, okay, David, 0:16:51.320 --> 0:16:53.280 I want you to learn this new object. Let's call it. 0:16:53.320 --> 0:16:55.880 You know, a coffee cup. You never touch it, and 0:16:55.880 --> 0:16:57.640 so you could do is you touch the coffee cup 0:16:57.680 --> 0:16:59.600 and you move your finger along and around, and as 0:16:59.640 --> 0:17:01.840 you do, you build a three dimensional model of the cup. 0:17:01.920 --> 0:17:04.639 Even though you're only getting input from one fingertip, the 0:17:04.720 --> 0:17:06.640 eyes are doing the same thing. It's surprising you don't 0:17:06.680 --> 0:17:10.200 realize this so every quarter do Colm, when we understand 0:17:10.240 --> 0:17:13.880 now is doing this sort of processing movement, information and 0:17:13.960 --> 0:17:17.200 sense for information, building what we call structure or three 0:17:17.280 --> 0:17:19.400 D models of things in the world. So it's quite 0:17:19.440 --> 0:17:22.560 different than even those neurosciences think about it, and there's 0:17:22.600 --> 0:17:24.080 a lot of reasons we can talk about how it 0:17:24.160 --> 0:17:25.840 was missed for all these years. 0:17:25.960 --> 0:17:28.200 So in the court, you have a century six layers 0:17:28.240 --> 0:17:32.679 of cells, and a column is all six layers, is 0:17:32.720 --> 0:17:36.639 all six layers. It's going up and down. It's like 0:17:36.880 --> 0:17:38.440 think of it like layers of a cake. And the 0:17:38.520 --> 0:17:41.480 column is you're taking a straw and shoving it through 0:17:41.520 --> 0:17:43.000 the top, and so you've got. 0:17:42.760 --> 0:17:46.320 This, Okay, got a straw cake. 0:17:46.560 --> 0:17:49.280 Okay, great, And so the idea is if you're looking 0:17:49.280 --> 0:17:54.120 at some column in you know, in primary visual cortex. Yeah, 0:17:54.240 --> 0:17:57.600 your point, Jeff was that, you know, it's it's like 0:17:57.680 --> 0:17:59.960 looking at the world through a straw. 0:18:00.119 --> 0:18:02.439 It only sees a little tiny piece of the world. 0:18:02.480 --> 0:18:05.520 But because the eyes are moving arout, because you're exploring 0:18:05.560 --> 0:18:09.560 the world, this is actually getting lots of parts of information. 0:18:09.600 --> 0:18:11.920 It's exploring the world in the same way that your 0:18:11.920 --> 0:18:13.000 finger typically. 0:18:12.640 --> 0:18:15.160 Right, and it has to integrate information over time, that's 0:18:15.160 --> 0:18:17.040 the key, right, And you can literally do this. You 0:18:17.040 --> 0:18:19.720 can look at the world through a straw, right, and 0:18:19.720 --> 0:18:21.600 and you can say, oh, what am I looking at? Well, 0:18:21.600 --> 0:18:24.320 you can't tell them. You start moving the straw and 0:18:24.359 --> 0:18:26.600 then you can start and you can also learn objects 0:18:26.640 --> 0:18:29.200 that way. So literally you can learn by looking through 0:18:29.200 --> 0:18:31.640 a straw, which is what sort of what one column 0:18:31.680 --> 0:18:32.920 is doing? Got it? 0:18:33.040 --> 0:18:37.119 And in your model there are thousands of such columns 0:18:37.320 --> 0:18:42.359 and each one of these is learning a model of 0:18:42.400 --> 0:18:44.800 the world as it's going. So tell us about right, right. 0:18:44.920 --> 0:18:48.720 So I think this idea that there's all these columns 0:18:48.800 --> 0:18:50.960 is not a new idea and that they have this 0:18:51.040 --> 0:18:54.320 fundamental argithm. But what we were I think the first 0:18:54.320 --> 0:18:56.439 people to kind of figure out what it is and 0:18:56.480 --> 0:18:59.119 what it's doing. So the trick of this thing is 0:18:59.280 --> 0:19:01.680 it's trick it here. You know, when you look out 0:19:01.680 --> 0:19:04.439 at the world, you have a sense anybody you have 0:19:04.480 --> 0:19:06.200 a sense where things are. I have a sense where 0:19:06.240 --> 0:19:07.760 you are relative to me. I have a sense where 0:19:07.800 --> 0:19:09.240 this microphone is relative to me. I know where my 0:19:09.240 --> 0:19:12.399 hand is relative to this cop I Now there turns 0:19:12.400 --> 0:19:14.880 out that you have any kind of sense of location 0:19:15.040 --> 0:19:17.840 in space, you have to have neurons representing it. There's 0:19:17.880 --> 0:19:19.920 nothing goes on in the brain if there aren't neurons 0:19:19.960 --> 0:19:22.960 firing doing it. Turns out most of the machinery in 0:19:23.000 --> 0:19:25.600 the New York cortex is keeping track of where things 0:19:25.640 --> 0:19:28.720 are relative to other things. So those six layers, all 0:19:28.720 --> 0:19:32.640 those cells, at least half of that circuitry is tracking 0:19:33.160 --> 0:19:35.679 where the sensory input is coming from in the world. 0:19:36.000 --> 0:19:38.120 So if I move my finger over this coffee cup, 0:19:38.760 --> 0:19:41.560 the part that's getting information from the sensory like I'm 0:19:41.600 --> 0:19:44.399 sensing an edge, for your example, as I move my finger, 0:19:44.560 --> 0:19:47.000 it has to keep track of where my finger is, 0:19:47.040 --> 0:19:50.080 a location of it and its orientation relative this cup. 0:19:50.160 --> 0:19:53.719 Is quite complicated, but that's what it has to do 0:19:53.800 --> 0:19:55.639 to build the models. And now we know how it 0:19:55.720 --> 0:19:59.000 does it. There's all this evidence for it. So the 0:19:59.080 --> 0:20:01.280 brain is just trying to track of where all of 0:20:01.280 --> 0:20:03.479 its inputs are in the world, all relative other things. 0:20:03.520 --> 0:20:05.800 Then it builds up these three dimensional models of the world. 0:20:06.000 --> 0:20:08.120 So tell us about how it does that then, right, 0:20:08.200 --> 0:20:10.520 so you can think about when you're in high school, 0:20:10.520 --> 0:20:13.360 you learned about Cartesian coordinates, x, y, and z coordinates, right, 0:20:13.800 --> 0:20:15.359 and so if I wanted to say where is something? 0:20:15.400 --> 0:20:17.399 Where are your relative to me? I might say, okay, 0:20:17.440 --> 0:20:19.359 your nose the origin, and I could say it's some 0:20:19.680 --> 0:20:21.879 distance from here, and you know X, Y and Z. 0:20:22.520 --> 0:20:26.080 Well you have to have something like that. But brains 0:20:26.119 --> 0:20:27.840 don't do it that way. They do it another way. 0:20:27.880 --> 0:20:31.320 And this was some very clever research in the last 0:20:31.359 --> 0:20:35.280 twenty years that people discovered in the antarilo cortex and hippocampus. 0:20:35.280 --> 0:20:37.639 These cells called grid cells and play cells, which actually 0:20:38.000 --> 0:20:42.280 operate as reference frames. They are a way of neurons 0:20:42.320 --> 0:20:46.800 to represent locations and they work differently than X, Y 0:20:46.840 --> 0:20:48.880 and Z, so there's no origin. It's kind of really 0:20:48.920 --> 0:20:52.639