Will AI Ever Be Conscious? (YouTube Video Transcript)

(00:00:00) Your YouTube transcript will appear here (00:00:00) I'm in therapy for [laughter] you. (00:00:02) >> And and it's interesting because it (00:00:04) raises and this is not my field of (00:00:05) expertise, but it raises questions about (00:00:07) what life is, (00:00:08) >> right? (00:00:09) >> Because you could say that life is just (00:00:11) it's about information. It's really (00:00:13) computing is what life is on some level. (00:00:17) >> So it's not really what you're saying (00:00:18) there really is biology, the the nature, (00:00:21) the physicality that we think of as (00:00:24) life. We think of biological systems (00:00:26) with DNA and all those things, (00:00:27) >> right? But you can you can argue that (00:00:30) that's not the really interesting bit. (00:00:32) That that's just the way that it's (00:00:34) realized. (00:00:36) It's an expression of the true thing (00:00:38) which is the computing the computing (00:00:40) which if that is the case then we have (00:00:42) stumbled into the creation of life that (00:00:44) will replace us which is if we ever get (00:00:48) to artificial general intelligence and (00:00:51) what you're saying is an emerging (00:00:52) property of computing which is also an (00:00:56) expression of life then it's only a (00:00:59) matter of time before that particular (00:01:01) computing becomes a life form which of (00:01:04) course will outthink us, outlive us, out (00:01:07) everything us. (00:01:08) >> Terminator. [laughter] (00:01:10) >> Oh yeah. And and this is, you know, (00:01:12) again, (00:01:14) >> smile while you're agreeing with him on (00:01:15) that. (00:01:17) >> Sad face for once on your that mug. (00:01:20) >> But one of the um things I've been (00:01:22) involved in, we have um I'm involved at (00:01:24) a a research institute called the (00:01:25) Francis Crick Institute in London, which (00:01:27) is a biosciences. It's it's a wonderful (00:01:29) place. It's it's a temple to curiosity. (00:01:32) I love the place. There's a a great (00:01:34) Nobel Prize winner called the Paul Nurse (00:01:36) who's a good friend of mine who won the (00:01:38) Nobel Prize for cancer research actually (00:01:40) by looking at what yeast cells. So it's (00:01:42) a it's a remarkable sort of fundamental (00:01:44) study of life but he really pioneered (00:01:47) the building of this institute or (00:01:48) inspired it in his image which is about (00:01:51) >> co-discovered the DNA (00:01:53) >> double healing (00:01:54) >> that's why it's called the (00:01:55) >> Institute um but we did some podcasts (00:01:57) called a question of science actually (00:01:58) which are around um and we just did them (00:02:00) at the Cric institute and with panels of (00:02:03) experts and so I just it was wonderful (00:02:05) for me because I just chaired it and (00:02:07) asked the questions and it were mainly (00:02:09) audience questions actually but one of (00:02:11) them was on AI And there was an (00:02:12) interesting split in the panel between (00:02:15) um the neuroscientists (00:02:17) and the and the computer scientists (00:02:19) >> really. (00:02:20) >> So so the neuroscientists (00:02:22) really felt that for example large (00:02:24) language models which is what we have at (00:02:26) the moment right (00:02:27) >> were just symbol shuffling things and (00:02:31) they and the brain is fundamentally (00:02:33) different to that. So we are not large (00:02:35) language models. (00:02:36) >> I kind of feel that way about them as (00:02:38) well. I kind of feel that way too. It's (00:02:39) just rearranging statistical (00:02:41) juapositions of words, (00:02:43) >> right? (00:02:43) >> And (00:02:44) >> it's seeing all the probabilities. (00:02:46) >> I don't feel like it understands (00:02:48) anything. (00:02:49) >> Yeah. (00:02:49) >> When I interact with a large language (00:02:51) model, it's like there's this vacuous (00:02:53) eyes staring back at me and there's no (00:02:55) soul behind it. (00:02:57) >> Yeah. Well, the the argument one of the (00:02:59) panelists gave was that imagine that (00:03:01) imagine that you're immortal. The time (00:03:03) doesn't matter to you, but we like we (00:03:04) could be in this room if we were (00:03:06) immortal and someone could start putting (00:03:08) little symbols in under the door and if (00:03:10) we put the right symbol out, we'd get (00:03:11) some food, right? So, we'd soon learn (00:03:14) what the right symbol was. And then they (00:03:16) put two through the door and we do the (00:03:17) same thing and then three. And (00:03:18) ultimately, if we had a huge amount of (00:03:20) time, kind of a near infinite amount of (00:03:22) time, we'd end up having a conversation, (00:03:25) right? (00:03:25) >> And we'd do it right. But at no point (00:03:27) would we have any clue what was going (00:03:29) on. But we'd not have any understanding (00:03:31) at all of what we were doing. (00:03:33) >> It's it's it's a transactional exchange (00:03:36) of simple information that itself (00:03:39) >> Yeah. (00:03:40) >> is not anything more than just (00:03:42) >> there's no understanding. (00:03:43) >> There's no understanding. (00:03:44) >> That's that's an one of the points of (00:03:47) view that were expressed. But (00:03:49) >> was that was that the neuroscientist? (00:03:50) >> That was a neuroscientist who said that. (00:03:51) I think it goes back to there's a (00:03:52) philosopher called Cell. I think there's (00:03:54) a an argument he made a long time ago (00:03:57) about symbol shuffling. Cell's argument. (00:03:59) So it's similar to that but one of the (00:04:01) computer scientists said no that that (00:04:04) irrespective of what you think about (00:04:06) that that's what we are. So we don't (00:04:08) know what we are we don't know what (00:04:10) consciousness is. So it could be that (00:04:12) that's all we're doing. We we're really (00:04:13) and it's true I suppose at the cellular (00:04:15) level at the level of a neuron. (00:04:17) >> Wow. (00:04:17) >> There's no understanding. (00:04:19) >> I don't I don't want to think I don't (00:04:21) want to believe that (00:04:22) >> now that you mention it. (00:04:23) >> Yeah. There are acoustic (00:04:26) stimuli coming from your mouth, entering (00:04:29) my ear, hitting my brain, and now I (00:04:32) process that and some other response (00:04:34) comes out. (00:04:36) >> And maybe I'm not conscious of anything. (00:04:38) >> No, you you're just a [laughter] like (00:04:41) information processing and response (00:04:44) machine. (00:04:45) >> Yeah, it's very possible. And I think (00:04:46) that this debate is quite live actually (00:04:49) amongst people among many people who all (00:04:52) know what they're talking about and and (00:04:54) there are different views which just (00:04:56) shows you it's complex a complex (00:04:57) emerging phenomena (00:04:59) >> that's makes sense and that is why a lot (00:05:01) of like and these aren't like (00:05:03) neuroscientists computer scientists but (00:05:05) there's many in the AI world who feel (00:05:08) like given enough time you just train (00:05:10) the AI on everything if you have enough (00:05:13) time and enough computing power, they (00:05:16) will definitely be truly thinking. (00:05:20) They're like thinking the way we (00:05:21) consider thinking. Uh in (00:05:23) >> especially when you think of thinking in (00:05:25) that way, right? And it reminds me of a (00:05:27) New Yorker comic. I think it was there (00:05:29) were two dolphins swimming right in in (00:05:32) this water park and there humans up (00:05:34) walking on the on the walkway and one (00:05:36) dolphin says to the other. They open (00:05:38) their mouths and noises go between them, (00:05:41) but it's not clear they're actually (00:05:42) communicating. [laughter] (00:05:44) >> Yes. Exactly. Right. Right. (00:05:47) >> Yes. [laughter] (00:05:49) >> So, I get that there's emergence in (00:05:51) these complex systems, but what is this (00:05:54) talk I hear of emergence from the (00:05:58) standard model of particle physics? (00:06:00) What's going on there? I thought that's (00:06:01) a pretty straightforward grid of what (00:06:04) exists and what should exist or how they (00:06:06) interact (00:06:07) >> if I understand the question right. So (00:06:09) there are things there are quite basic (00:06:11) things about particles that are (00:06:14) difficult to derive from the standard (00:06:15) model. So the standard model is you know (00:06:17) that here is the the quarks and the so (00:06:21) up quark down quark electron electron (00:06:23) >> it's an inventory. (00:06:24) >> So so we have 12 matter particle the (00:06:26) higs bzon and then three forces that it (00:06:29) describes. (00:06:30) >> It's an inventory. (00:06:31) >> Yeah. Well, and then it and it tells us (00:06:33) about the interactions, but it's got so (00:06:35) how particles interact with each other (00:06:37) and through which forces do they (00:06:39) interact. [music] (00:06:46) [music] (00:06:51) [music] (00:06:56) >> [music] (00:07:02) [music]