https://youtubetranscript.com/?v=33mlMJAB-AY
So we’ll start with this first proposition. To see the world, we must, must prioritize our perceptions. So John, I’ll ask you about that first because that’s a particular, I believe a particular focus of yours. I don’t think that’s an exaggeration. Yeah, it’s not. That’s the core of my work. And so the main way I would respond to this is I would say, I think the work that’s coming out from artificial intelligence and the work that’s coming out from attention lines up with this very well. I don’t have any significant disagreement with that proposition. And the must part of it as well? So the must, I took, well, let me tell you how I took the must. And I took it as what’s called constitutive necessity. I took it to be, if you’re going to be a cognitive agent, then you must do this. I didn’t take it to be a metaphysical necessity. I took it to be that kind of constituent necessity. I think it’s useful to start with what you describe as constituent necessities before you move into the realm of metaphysical necessity. I think that’s a good way to argue. You should, right? And so I think, and I’m not gonna recapitulate all these arguments, but a lot of work, I think, zeros in on the idea that the core of what makes us intelligent and the thing that we’re finding difficult to give to machines to make them artificially general intelligence is a process I call relevance realization, which is exactly, I think, lines up with this very well. The amount of information available to you in the world is astronomically vast. All the things you could pay attention to. The amount of information in your long-term memory, especially if you think of all the ways it could be combined is also astronomically vast. The number of options of potential lines of behavior. I could move this finger or this finger. I could move them. I could lift, like, and the ways I could move around. That’s combinatorial explosion. All of it, all of it. And then, right, and then you can also consider all of the options of different potential worlds you might want to consider trying to produce or moving into, right? And so the point is, in many different dimensions, we face combinatorial explosion. And what you can’t do, and this is where it lines up with the most, because we’re finite beings, with finite resources and finite time, is you can’t check all of that information. So you can’t go and say, no, that’s not relevant. That’s not relevant. That memory’s not relevant. That will take, like, the rest of the history of the universe. Right, right. So we don’t know how we do it, in fact, because of that in part. Well, I mean, I think there’s getting some clues towards it, but we can talk about that later. Okay. Right? So the must and the prioritization on the perception side, you’re fine with. It has to be, it has to be. But here’s the tricky thing, which is, the fact that we can’t check it means, and this sounds almost like a Zen Cohen, is the prioritization is odd when you say it sort of like prima facie. Yeah. Because it means we intelligently ignore most of the information. So the prioritization, the one I wanna put. Yeah, that’s a good code of silly. So you’re saying that you don’t wanna misinterpret the necessity for prioritization as something like the necessity or our ability to make a numbered list of that number of possibilities that lay out in front of us, because that’s actually impossible. Right. So that isn’t how we do it. However we do it, isn’t that? Exactly. So if you’re okay with that reading, and it sounds like you are, prioritization doesn’t mean what we normally mean by prioritization, where we set things out, explicit and focal, and then choose between. Right, right. It’s implicit. It’s implicit and it’s self-organizing, and our ability to think. And it’s unconscious. Yes, emerges out of it. We can influence it top down, but because it is an absolute requirement for our cognition, I would argue that our ability to do anything that we do consciously is ultimately dependent on it, and presupposes it. Okay, fine. So that’s good. Jonathan, do you want to add to that? The only thing that I would add is, you have to phrase it in a certain way. There’s no, you have to have a sense, but there’s a sense in with perception. When we say we must prioritize our perceptions, I think the best way to understand it is that perception is already prioritization. In order to perceive, there has to be a hierarchy in itself. Perception is in and of itself an act of implicit prioritization. And to use the word implicit would be a good idea. So to avoid the idea that we are consciously doing it, but that in order to even perceive the world, there already has to be a given hierarchy that is making you able to focus on anything, because or else we would be lost in a wave, a sea of infinite details. Okay, so I think that’s a good quote, still. And so we could also make a little technical case here, quickly. So part of the problem that Jon referred to is that in some sense, it’s the problem of the finite confronting the infinite. And so we could make a neurological argument for that. So for example, when you move your eyes around, or when they move around as a consequence of being directed by unconscious structures of prioritization, because that happens all the time, you move your eyes around because you want to direct the high resolution part of your visual system to whatever you’re attending to. That’s the fovea. And the fovea is a very small part of your retina. And it’s a very high resolution part. So each cell in the fovea is connected at the level of the primary visual cortex to 10,000 cells. And then each of those have 10,000 connections. And so if your whole vision was foveal in its resolution, you’d have to have a skull like an alien to contain that much brain. And so, and that’s a real indication of that finitude, is that you do have limited cognitive resources and limited means practically and physically limited, but it also means metabolically limited. The cost of running your brain is already extremely high. And so you’re gonna shepherd your available attentional resources because they are finite and they’re finite in no small part because they are technically metabolically costly. That all seems okay. So I would add one thing to that, which is I would put an emphasis on how this process has to be self-organizing because we wanna avoid a perennial problem, which you and I both know shows up in psychology, which is to posit the internal homunculus that actually doesn’t explain the problem, but just shifts it. The central executive is an example of this, et cetera. So we don’t wanna say that there’s someone that’s doing the prioritization because that someone- Is just as mysterious. Right, and is facing the very problem that we’re trying to explain. So the process has to be dynamically self-organized. Well, one of the ways I’ve realized how that problem works is in an attempt to solve the mind-body problem because you can’t solve the mind-body problem, but you can say, let’s say you wanna explore an idea and you decide to do that by writing an essay. So then you sit down in front of the computer, which is not an idea. It’s actually that you’re sitting and then you move your fingers on the keyboard. And so there’s a hierarchy of transformation from mind, which might be the abstract intent, to body. And so the spirit hits the body in the finger movements. And then the spirit disappears in some sense under the finger movements because you can move your fingers voluntarily, but you have no idea what muscles you’re moving to do that. And you can’t control the cells or anything like that. Oh, I did that with my students in the lecture this morning. I was talking about this very fact that I said, put up your finger, okay? Bend your finger. What do you do to bend your finger? Right. Exactly. So interesting. It’s so interesting that you have that level of consciousness at that level of detail, which is pretty detailed, but no more than that. Yes, yes. Yeah, so that’s a mystery, man. That localization of consciousness between the body and part of the spirit, there’s like a, what would you call it? There’s a bandwidth of resolution for consciousness. And why that band? See, the social psychologists who studied language sort of caught on to this because one of the things they realized was that short words, first of all, short words tend to be old words. So, because as language develops, words that are used a lot get more efficient, but the short words also map extraordinarily well onto the self-evident level of perception. And so, for example, a short word is cat because a cat presents itself, for some reason, to our perception. The species cat doesn’t, and the fur of the cat in some sense doesn’t. It’s the cat. Yes. And you can see that primary object level recognition, I think is basic level. Basic level. That’s right. A little rush. Yes, yes, yes, yes. And so you see that with babies because they get doggy pretty damn fast. And that’s because the language maps onto the primary domain of perception. And so, I think that’s the thing that we need to do to get the best out of that basic level perception quite nicely. And that is associated with something like the natural bandwidth of consciousness. Yeah. I would say that that lines up with, if Rasha’s explanation is, you’re getting the best trade-off between differences between category and similarities within categories. Right. And then the question is what does best trade-off mean? Exactly. And that’s a nuance I’d want to put onto the prioritization because the prioritization sounds very, but sort of like an imposition. Whereas I think what we’re talking about is something more like what Marlo Ponte talked about when he talked about optimal gripping. Right. You bet, you bet. What’s the correct distance to look at this? Well, it depends. Because if I zoom in, I lose the gestalt. If I zoom out, I lose the detail. It depends on what you want to do. Exactly. Yes, well, that’s it. That’s why I’m kind of attracted to pragmatism. It’s like, well, to some degree, our theories of truth need to be embedded in the practicalities of action. And so is that a grippable object that I can drink from? Well, I want my perception to match that problem. Yeah, but it doesn’t, I think that if you understand that the prioritization, the thing that you have heaven and earth, I’m gonna use, sorry, I’m gonna use mythical categories, but so you have- That’s why I’m here. You have heaven and earth and that the way in which heaven meets earth is a mutual relationship, right? We always see it as a relationship of lovers, you could say. But it’s not, the prioritization isn’t just about an imposition from above, but it’s about the manner in which that which is above, let’s say the hierarchy is able to encounter the potential in which it’s- We were talking about that last night. So Jonathan made this funny joke last night. We were talking about Sam Harris and Sam Harris has this line of argumentation where, and he used this on me where I interpreted a biblical story and then he interpreted a recipe. Yes. And he said, well, look at all the interpretations. And that is a problem of- Semiotic drift, yes. Well, it’s also a problem of this horizon of infinite possibility. There are multiple interpretive schemes. So Jonathan said he’d like to do a video where he shows that a recipe is actually necessarily embedded inside a mythological framework. And we started to talk about that because imagine, well, the recipe implies that you need to make an edible meal, that you want to make an edible meal, that you’re going to serve it to family and friends, that that’s part of a kind of communion, that you think that’s a good thing that’s worth spending time on, that serves your family and friends, that’s maybe nested in something like an ethic of service to the community. Like there’s a whole network of purpose. I would add more to that. There’s all kinds of implicit assumptions that I can capture in a sequence of propositions, procedural skills that are not completely capturable in words, and that those procedures and skills can also map on to the particular virtues and skills that people are bringing to it. Most things can’t be solved by a recipe. Right. And yet, so a recipe is a significant cognitive cultural achievement, and we don’t recognize, and we tend to overgeneralize the things we think for which we can provide recipes. This is one of the- Right, that’s an algorithm issue. Yes, exactly, exactly. And so there’s lots. Yeah. But even in the recipe itself, you will notice that the way in which we name things and the way in which we order things will be related to a normal prioritization, hierarchy prioritization. But if you’re making chicken, you’ll have the chicken, and then you’ll have the spices, and you’ll understand that these elements that I’m adding are spices, and that they’re, let’s say, something like a marginalia that I’m adding to the central meal. It’s actually the very, it’s like it’s the pattern of a church, actually, where you have a movement towards the central identity that we understand, and then we have the way in which it’s complemented to other things. And so even the actual recipe itself is like a little microcosm. And also the judgment you use is like, well, how much spice? Well, the answer is, well, what function is the spice going to serve? And you say, well, I wanna add a little zest and interest to my cooking. And so then you have a philosophy of zest and interest that’s associated with that, because just predictable chicken isn’t good enough. And maybe you wanna put a little more spice on, because you wanna, what would you say? You want to challenge your guests a little bit in an interesting way, and you’re thinking this all through. And- For the same reason you’d wear funny socks or a tie that has just a little bit too much on it, you know? Well, it’s the same thing. I mean, that’s actually a feature of general problem solving. Like you, when people are solving a problem, especially if they might get the wrong frame, moderately distracting you from the central concern is an optimal way to do it. Yeah, exactly, you need to do that. So what I’m hearing both of you say is the prioritization is really a multi-dimensional optimal gripping. That’s right, that’s right. We’re concentrating on. Okay, well then we can also expand on that to some degree, because multi-dimensional and optimal brings a lot of other concerns into it. So imagine that one of the principles, and Kant moved towards this with his theory of universal ethic in some sense, although I think, you know, I hesitate to criticize Kant, but I think that there’s a deeper explanation for what he observed is, well, how should I treat you? Well, that’s a complex question, but one of the constraints is, well, what if we meet a hundred times? So we’re gonna establish an actual relationship. So however I conduct myself in the present moment has to be in accordance with a value hierarchy that takes into account the desirability of our mutual interrelationship into the future. And that produces a very serious series of, I would say, often intrinsic constraints. So I can’t be too insulting, I can’t be too unwelcoming. I have to offer you something approximating a true reciprocity for the thing not to degenerate. And so, and all of that, and I would say that also governs how you cook for someone if you actually want to make friends. So it’s like treat other people as you would like them to treat you. And it’s pretty funny that that’s the intrinsic ethic and the recipe. And so that’s such a funny argument.