Maybe Consciousness Is a Crutch AI Won't Need

追加: 2026-05-23

[00:00:00]Okay, well, let's talk about Penrose AI and consciousness. Okay, fine. I just saw Roger in um I like to call him Sir Roger, although sometimes the British uh tease me for that cuz it's so formal.

[00:00:11]It's an Apparently, when someone is knighted, you don't call them Sir Penrose, you call them Sir Roger.

[00:00:16]Uh [laughter] And I think it's quite endearing to call him Sir Roger. Um but I saw him recently. We had a conversation together at Oxford just in October.

[00:00:26]And he remains one of the most interesting people I've ever known.

[00:00:30]Great. Okay, I'll place a link to that on screen and in the description as well. So, he also ties uncomputability to consciousness and says that AI is not conscious, computers may never be conscious. Now, you also take the view that AI, current AI, is not conscious. I imagine for a different reason. So, I want to spell out where do you agree with Ro- Roger and where do you diverge with Sir Roger? Yes.

[00:00:53]>> [laughter] >> Well, uh I'm not sure of all of his dimensions in that. Last time I spoke to him, we were actually talking about black holes, in fairness. But um but I I think from from my point of view, AI isn't yet conscious.

[00:01:07]I am not a person who thinks uh consciousness is substrate dependent.

[00:01:12]I don't see a barrier to consciousness in a different material. That strikes me as really odd. I can imagine it being a way too hard for us to simply algorithmically program. I can imagine that. Um I can imagine that this whole all that's going on with these biological systems 2 billion years to evolve and are just very complex and so we're kind of stuck with biology because nature did this for us through evolution.

[00:01:39]Um but I I can't find a logical reason to preclude the possibility that it's conceivable that there could be um consciousness in a different substrate.

[00:01:52]So, I'm not against AI consciousness as a possibility. I just think we're I just believe we're so far off even though I am stunned by LLM's. They're incredible, man. And I you know, I'm not a big uh person who using AI for um emotional or intellectual support, but man, what a search engine. Right. I mean, what a search engine. It is unbelievable. This thing does talk to me. It won't stop using scare quotes. I can't get it to stop doing that.

[00:02:23]>> [laughter] >> What do you mean?

[00:02:25]So, anytime uh an LLM wants to say something metaphorical, you'll find a lot of them will put quotes around the word.

[00:02:34]And I and I wonder you know, and I keep saying, "You don't have to put I understand what a metaphor is. You don't You don't have to keep putting a quote around this word." Like, the universe chooses like if it wants to say, "The universe chooses a state." It will put quotes around chooses. Like, I find I find that I can't disabuse it from this habit.

[00:02:52]>> [laughter] >> Do you have custom instructions on them?

[00:02:55]Um I'm not that deep into it. I think I have I think I have tried that cuz this is such a pet peeve of mine. I see. But anyway, anyway, I I'm not so deep in my I I'm just a newbie in some sense to this whole thing.

[00:03:08]>> My pet peeve is it's not X, it's Y.

[00:03:11]>> [laughter] >> That you you know where you know where it's it's not that the universe cycles, it's that the universe does this and that and it's just it's so frustrating and then the M dashes. Right, the M dashes. Um I probably have a bit of a bad habit with M dashes myself.

[00:03:25]Yeah, but that's because LLM's are trained on you. You trained the LLM's.

[00:03:29]Right. Right.

[00:03:31]I mean, there's a lot of things to complain about the LLM's and what it's doing to society, you know, and how people are are less and less original every day and more and more average, but um but putting and how can we innovate if we're simply copying ourselves frozen at a particular moment in the history of humanity, you know, all of this stuff.

[00:03:47]But that's not the big questions. The really interesting questions are these ones about consciousness. And so, I I guess what I I guess what I feel about it is if I think about human beings as being a series of chemicals interacting according largely electromagnetism.

[00:04:05]Largely, right? It's usually large molecules forming DNA RNA and proteins and all of this stuff happening just on the basis of electromagnetic interactions.

[00:04:18]Really basic laws of physics.

[00:04:20]At what stage does that become human consciousness? Of course, one of the most tremendously thorny and evasive, elusive questions that humanity's ever asked. So, I don't see why we're expecting the LLMs or or any AI to for us to be able to solve that problem before we can solve this problem.

[00:04:44]Like we don't understand how this happened. I still cannot in a laboratory go from molecules interacting electromagnetically to comprehending why there's an interiority and a sense and an imagination that hallucinates the color red.

[00:05:03]That absolutely evades us. So, as a physicist, where do you stand on the hard problem?

[00:05:10]So, uh David Chalmers is a good friend.

[00:05:14]Um my It's funny cuz David Chalmers basically has become superb at dismantling any attempts to solve the hard problem. That's his job. He is this gatekeeper and no matter what solution you come at him with, he's going to dismantle your attempt to solve the hard problem. Sometimes uh I have other friends in neuroscience who think that it's just a total red herring, that it's just really not the question we should be asking and it's just silly. Sometimes I liken it to it's the it's the how question. So, we can say where, what, who, when, why. We can answer all of these questions about consciousness somewhat. You know, it's evolution, where it happened, and where it happens, where the um biological correlates are, and all of these things are answerable. Nobody thinks that those are the hard problem.

[00:06:06]The hard problem is how. Like, how does a photon hitting my eye and triggering some electromagnetic interaction along some fibers result in me imagining, hallucinating the color red? The I just nobody knows that. And nobody's come in my opinion anywhere close. And when people try, they're really usually answering a why question. I have you know, human beings evolved consciousness. Why? Well, because our compute is low. We don't have infinite compute. We struggle to understand all of this data. So, we evolved a mechanism to approximate really quickly. I can out of the corner of my eye, I know it's a baby.

[00:06:51]Out of the corner of my eye, I know it's a tree. Or, you know, or a car speeding at me.

[00:06:56]All of that it consciousness is a really good tool with low compute. I I can just this approximate hallucination I have allows me to survive and very quickly with very approximate um understanding. And and so, I understand a lot of the whys. You can talk about whether emotions are important or the unconscious. All this is why stuff. It's all interesting. But, none of it really explains how.

[00:07:22]How I have an interiority and how I have this hallucination of the world. Um so, I and I think it's a fascinating and wonderful thing. It's basically we're all obsessed with our hallucinations of the world. It's everything to us, right?

[00:07:34]But, um but I would say uh that I sometimes wonder a stronger thing, which is if a machine has tremendous compute, like practically infinite compared to us, >> [gasps] >> will it need to evolve consciousness? Is there any reason why it would have to?

[00:07:56]Or can it just continue to do perfectly well by simulate by pretending by simulating consciousness the way an LLM does, a large language model. And and that I don't know. Maybe maybe it just won't maybe consciousness is a crutch AI won't need.

[00:08:12]And so it will never develop.

[00:08:14]You know I have a bone to pick with you.

[00:08:16]Oh, please.

[00:08:17]>> [laughter] >> I can hang up, you know, and take my toys and go.

[00:08:22]I subscribe to The Economist. Their science and their AI coverage is among the best I've found anywhere. And I say that as someone who reads plenty of it.

[00:08:31]I'll give you some examples. They just ran an analysis on how attitudes towards science are changing in American politics and what this means for research and funding in scientific institutions moving forward. This sort of high-quality reporting is fantastic.

[00:08:47]They even covered how dark energy may be weakening over time. Now, if that holds up, it completely changes our understanding of the universe's fate. If you watch this channel, those are exactly the kinds of questions that we explore every week. I subscribe to The Economist because their science and their AI reporting regularly surprises me with how deep it goes. And they're also, of course, known for global affairs, both political and economic reporting. They are top tier and interestingly and flatteringly, TOE is one of the only podcasts that The Economist partners with. So, as a listener, you get an exclusive 35% off.

[00:09:24]That's not a deal that they have just anywhere. Head to economist.com/toe to subscribe. That's economist.com/toe for 35% off. I recently signed this book deal with Penguin. Penguin is fabulous.

[00:09:42]Penguin is wonderful.

[00:09:44]>> So, in it I made this analogy to averted vision.

[00:09:48]And I haven't heard anyone make that analogy before. And also, I have the characters of Gödel and Turing as well.

[00:09:53]And then I just And then I see, "Oh gosh, Jen 11 just beat me to it."

[00:09:58]I gave a lecture a couple days ago at the Mind at Large conference, which is a conference on panpsychism. And I was saying that, "Look, some phenomena may be such that when you try to look at it directly, it actually disappears." So, in physics, there's something called a non-Newtonian fluid, which it looks fluid, but you touch it and then it becomes solid. But then the opposite is thixotropic, where it looks solid, but you try to touch it and it becomes liquidy. It escapes your grasp. And I was saying potentially the self is like that. Potentially consciousness is like that. Potentially free will is like that. And what happens is we take this fuzzy intuitive concept and we make a formalization of it. We look at the formalization and say, "It's incoherent." We then conclude it it does not exist. And I think that's akin to looking at something directly, not seeing it, feeling it in the sense with by not looking at it, but then you try to make it explicit, then it disappears.

[00:10:49]And then I saw that you also made a similar analogy. And I was upset with you. I'm so sorry.

[00:10:55]>> [laughter] >> Yeah, that's in a Madman Dreams of Turing Machines.

[00:10:59]Um where truth is just like that. You can see it, but only out of the corner of your eye.

[00:11:05]Um and it disappears when you try to look at it directly. I think it's a beautiful analogy for for what happens What I was trying to do in a Madman Dreams of Turing Machines is admit that you could not get to the truth of the stories of these lives in an axiomatic way.

[00:11:24]Simply by as though rolling from theorem to theorem, listing biographical facts.

[00:11:30]I couldn't write, in other words, nonfiction and get any closer to the truth about them in the same way that Gödel's true statement can't be arrived at through theorems. And I felt that that underscored sort of this idea of well, sometimes our human narrative, which destroys things in the process of telling the story, is as close to the truth as you're going to get. You can only get to it out of the corner of your eye, you know.

[00:12:01]Yeah. So, well, now I'm looking forward to your book. Is it fiction fictionalized? No, no, no. I wanted to bring people to the frontier of physics because people keep thinking the frontier of physics is string theory, it's ADSCFT. I'm thinking that's 20 years, 30 years old. Like, what's the frontier? Yeah.

[00:12:16]>> As I was going through this, I was thinking, okay, because I'm not a physicalist, but I'm not an idealist, either. Uh-huh. I'm not a monist in in the sense of of any traditional sense of the word, but I'm philosophically homeless. I don't know what I am. I just am unconvinced of all the the mainline positions.

[00:12:32]>> [gasps and laughter] >> Well, there's always there's always more to be done.

[00:12:38]>> [gasps] >> I mean, it's almost this idea that we we should we should be reapproaching theories and ideas and saying them different ways and each person having a unique approach. You know, this idea that oh, the novel's already been written, there's no new novel. You know, of course, that's absurd. It's almost like saying, well, let's not make more human beings because a human being has already experienced the world. So, nobody else should experience the world.

[00:12:59]And um and of course, that's not what we actually believe. We actually believe that each experience of the world is unique and brings something and and experiences something unique.