Lionel Page

Yeah, yeah. And you may wonder, why? Why are we designing like that? Why can't we have something like happiness, which is something like a mountain, and you start from the bottom, and as you're more and more successful in life, you get more and more happiness. Why are we not designing that?

And the quick answer is that designing a system which instead of measuring big difference like that, only focus on measuring variations related to expectations, it's a more efficient system to treat information and to, uh, use whatever, um, cognitive capacity you have in your brain to produce a signal, which is going to help you. So it's a bit abstract, but I can say that, um,

And the quick answer is that designing a system which instead of measuring big difference like that, only focus on measuring variations related to expectations, it's a more efficient system to treat information and to, uh, use whatever, um, cognitive capacity you have in your brain to produce a signal, which is going to help you. So it's a bit abstract, but I can say that, um,

And the quick answer is that designing a system which instead of measuring big difference like that, only focus on measuring variations related to expectations, it's a more efficient system to treat information and to, uh, use whatever, um, cognitive capacity you have in your brain to produce a signal, which is going to help you. So it's a bit abstract, but I can say that, um,

Something which we have learned in the last 30 years is having a very interesting convergence between AI research and reinforcement learning and cognitive neuroscience. And what some cognitive neuroscience found out is that the brain looks like when the brain rewards you as a difference relative to your expectations, it pretty much looks like it's implementing.

Something which we have learned in the last 30 years is having a very interesting convergence between AI research and reinforcement learning and cognitive neuroscience. And what some cognitive neuroscience found out is that the brain looks like when the brain rewards you as a difference relative to your expectations, it pretty much looks like it's implementing.

Something which we have learned in the last 30 years is having a very interesting convergence between AI research and reinforcement learning and cognitive neuroscience. And what some cognitive neuroscience found out is that the brain looks like when the brain rewards you as a difference relative to your expectations, it pretty much looks like it's implementing.

optimal algorithms used in machine learning. So you'd have people working in artificial intelligence, trying to program how a program is going to learn the right thing to do. And the best one simple thing to... for this program to learn is to say, well, form expectations about what different actions are going to lead to, and then try out. And when you try the action, you just compare.

optimal algorithms used in machine learning. So you'd have people working in artificial intelligence, trying to program how a program is going to learn the right thing to do. And the best one simple thing to... for this program to learn is to say, well, form expectations about what different actions are going to lead to, and then try out. And when you try the action, you just compare.

optimal algorithms used in machine learning. So you'd have people working in artificial intelligence, trying to program how a program is going to learn the right thing to do. And the best one simple thing to... for this program to learn is to say, well, form expectations about what different actions are going to lead to, and then try out. And when you try the action, you just compare.

Is this action, is the outcome better than expected or worse than expected? And then you adjust your expectation. And if you try a lot, eventually you are going to learn to do the right thing. It's pretty much exactly what we do. And it's an efficient way of processing information.

Is this action, is the outcome better than expected or worse than expected? And then you adjust your expectation. And if you try a lot, eventually you are going to learn to do the right thing. It's pretty much exactly what we do. And it's an efficient way of processing information.

Is this action, is the outcome better than expected or worse than expected? And then you adjust your expectation. And if you try a lot, eventually you are going to learn to do the right thing. It's pretty much exactly what we do. And it's an efficient way of processing information.

it would be much more difficult for your brain to have a very complete map about, you know, happiness from zero to the top. It's better to have a kind of a local stuff guiding you locally. Incremental. Exactly. Incremental. Yeah.

it would be much more difficult for your brain to have a very complete map about, you know, happiness from zero to the top. It's better to have a kind of a local stuff guiding you locally. Incremental. Exactly. Incremental. Yeah.

it would be much more difficult for your brain to have a very complete map about, you know, happiness from zero to the top. It's better to have a kind of a local stuff guiding you locally. Incremental. Exactly. Incremental. Yeah.

Well, it's exactly what I was saying before, is that it's more efficient. I think a very good comparison is our visual system does exactly the same thing. So your visual system doesn't kind of recall the objective luminosity in a room the objective luminosity. Actually, it's not measuring it.

Well, it's exactly what I was saying before, is that it's more efficient. I think a very good comparison is our visual system does exactly the same thing. So your visual system doesn't kind of recall the objective luminosity in a room the objective luminosity. Actually, it's not measuring it.

Well, it's exactly what I was saying before, is that it's more efficient. I think a very good comparison is our visual system does exactly the same thing. So your visual system doesn't kind of recall the objective luminosity in a room the objective luminosity. Actually, it's not measuring it.

From the time where the light hits your retina, what's recorded is actually a divergence relative to expectations. And what you see is that, you know, if you were to turn off the light somewhere, so now you see things, you turn off the light, everything is bleak. So you can't see anymore. But if you wait a bit, your eye is going to adapt, you're going to start seeing shades, et cetera.