(calming instrumental music) – Well I think actually there
are some basic principles to how human thinking works. My last book, “How to Create a Mind”, which came out in 2012, which resulted in me
being recruited to Google, I practiced those ideas, talked about how human neocortex works, it’s outgrowth of a thesis I have had actually for 50 years because I wrote a paper when I was 14 or 15 in 1962, and won a science contest to
Westinghouse Talent Search. Now the intel inside this
talent got to be present content and I described human
thinking as consisting of modules of neurons, each module can recognize a pattern, and the basis of human thinking
is pattern recognition. And those patterns are
actually sequential. And they’re in one direction. And I gave a lot of evidence for this. For example, try to recite the alphabet. Now, most of you can do that fast. Okay well, recite it backwards. You probably can’t do that, unless you learned that as a new sequence. It’s a pretty trivial transformation, and yet we can’t do it. So we have different hints as
to how the human brain works. In recent years, it’s been an explosion of neuroscience evidence. For example, the European brain
reverse engineering project has identified modules of
about 100 neurons each, and since the neocortex
has 30 billion neurons, that means 300 million modules, and they all are pretty much the same. They have the same structure, the same organization within them. And there’s no plasticity, no change, within that module for your entire life. Despite the idea that your brain is constantly rewiring itself. There is plasticity, constant
rewiring between the modules, and each module is recognizing a pattern. We can see the axons coming
in from other modules that are feeding the sequential input that represents the pattern,
that this module will learn. So it’s a hierarchy of patterns. This pattern is based on a hierarchy of patterns and the modules below it. And each one of those has input from modules below it, and it’s
a very elaborate hierarchy. And biology, biological evolution evolved this hierarchical structure in the brain, so that it can understand and learn the hierarchical structure of the world, because the world is
organized hierarchically. The neocortex emerged 200
million years ago in mammals; only mammals have a neocortex. And it was a thin structure, the neocortex means new rind, and there was about. In the first mammals, which were rodents, it was about the size of a postage stamp, and just as thin as a postage stamp, and it wrapped around
the walnut-sized brains of these early mammals, but it was capable of
a new type of thinking. You could invent new behaviors, non-mammalian animals, like reptiles, that didn’t have a
neocortex, couldn’t do that. They have fixed behaviors. Didn’t help them that much actually because the environment
changed very slowly and could take 50 thousand years for there to be an environmental
change that would require a new behavior, and over
the 13 thousand years, these non-mammalian animals could evolve using normal Darwinian
evolution, a new fixed behavior. But then something happened
65 million years ago. It was a sudden catastrophic
change to the environment; we call it the crustacean
extinction event, and that’s when mammals
overtook their ecological niche. That’s when the neocortex
actually showed its capability. And then biological
evolution then grew it. Mammals now, instead of
being just little rodents, got bigger, their brains got
bigger, at an even faster pace, taking up a larger fraction of their body. And the neocortex got
bigger even faster than that and developed these curvatures and folds. If you look at a primate brain, it’s got these characteristic curvatures, so it now takes up 80% of the brain. Then something else happened,
two million years ago. If your remember, two million years ago, we were walking around; we
didn’t have these big foreheads. So humanoids came along
with a big forehead. And that houses the frontal cortex. And up until recently, it was said “Well, the frontal cortex does “such qualitatively different things, “it must be organized differently. “It must have a different
method, a different algorithm.” I make the case, and I
think the neuroscientists coming around to this view, it really was just an additional
quantity of neocortex. Well, so what did we do with
that additional quantity? Well, we were already
doing a very good job of being primates, so we put it at the top of the neocortical hierarchy. So this hierarchy that I
mentioned now got bigger. As you go up the hierarchy, things get more general, more
intelligent, more abstract. The very bottom, I can tell
that that’s a straight line. At the top, I can tell that’s funny, that’s ironic, she’s pretty. So that additional hierarchy that we got two million years ago was the enabling factor for us to invent language, and art, and science, and music. Every human culture we
every discovered has music. No primate or any other animal has music. That came from this additional neocortex. And I make the case in my
book, “How to Create a Mind”, what the algorithm is,
of each of these modules. They all have the same algorithm. So a lot of people like to say, “Oh, the brain is so complex, “it’s the most complex
thing in the universe”; that may be true, but it has
a regular repeating structure. Each of these 300 million
modules is basically the same. Now they self-organize
into these hierarchies, and each module discovers a pattern, and learns it, remembers
it, and can recognize it, even in a different context,
so it’s very good at metaphor. And I describe my thesis on how this works as we continue doing more
brain reverse engineering, we will refine that model, but I’ve been working with this model, and we find that it can
in fact master things like language, not yet at human levels, but doing still some impressive things. We look beyond, for example,
for things like jeopardy, which itself was pretty sophisticated. So there is kind of a master algorithm, at least I have a proposal for one. These deep neural nets, which there’s tremendous excitement about, which is a little bit different
from the model I have, but they have done remarkable things. I mean, they won the Go Championship, and they can recognize
images as I mentioned, better than humans, and can drive cars. And that’s actually pretty simple. You can read about deep neural nets, the algorithm is again,
a repeating structure that’s not that complicated. So the mathematics of thinking, I think is being understood, but I would not claim that
we understand it fully. But we’re getting more and more hints as we learn more and more
about the human brain.