Latest / The NanceCast / The Godfather of AI says "It's going to kill us all!!!"
Transcript
- 0:01to the bottom of the clock.
- 0:02And, uh, look, look, see.
- 0:04Yeah,
- 0:04yeah,
- 0:05yeah,
- 0:05SPEAKER_03: yeah,
- 0:05SPEAKER_03: yeah,
- 0:05SPEAKER_03: yeah,
- 0:05yeah,
- 0:05SPEAKER_03: yeah,
- 0:06yeah,
- 0:06yeah,
- 0:06SPEAKER_03: yeah,
- 0:06yeah,
- 0:07yeah,
- 0:07yeah,
- 0:07yeah,
- 0:07yeah,
- 0:07yeah,
- 0:07yeah,
- 0:08yeah,
- 0:08yeah,
- 0:08yeah,
- 0:08yeah,
- 0:08yeah,
- 0:09yeah,
- 0:09yeah,
- 0:09yeah,
- 0:09yeah,
- 0:10SPEAKER_03: yeah,
- 0:10SPEAKER_03: yeah,
- 0:10SPEAKER_03: yeah,
- 0:10SPEAKER_03: yeah,
- 0:10SPEAKER_03: yeah,
- 0:11SPEAKER_03: yeah,
- 0:11SPEAKER_03: yeah,
- 0:11SPEAKER_03: yeah,
- 0:11yeah,
- 0:12SPEAKER_03: yeah,
- 0:13SPEAKER_03: yeah,
- 0:13SPEAKER_03: yeah,
- 0:13SPEAKER_03: yeah,
- 0:14SPEAKER_03: yeah,
- 0:14SPEAKER_03: yeah,
- 0:14SPEAKER_03: yeah,
- 0:14SPEAKER_03: yeah,
- 0:14SPEAKER_03: yeah,
- 0:14SPEAKER_03: yeah,
- 0:15SPEAKER_03: yeah,
- 0:15SPEAKER_03: yeah,
- 0:15SPEAKER_03: yeah,
- 0:15SPEAKER_03: yeah,
- 0:15SPEAKER_03: yeah,
- 0:16SPEAKER_03: yeah,
- 0:16SPEAKER_03: yeah,
- 0:16SPEAKER_03: yeah,
- 0:16SPEAKER_03: yeah,
- 0:16SPEAKER_03: yeah,
- 0:17SPEAKER_03: yeah,
- 0:17SPEAKER_03: yeah,
- 0:17SPEAKER_03: yeah,
- 0:17SPEAKER_03: yeah,
- 0:17SPEAKER_03: yeah,
- 0:17SPEAKER_03: yeah,
- 0:17SPEAKER_03: yeah,
- 0:18SPEAKER_03: yeah,
- 0:18SPEAKER_03: yeah,
- 0:18SPEAKER_03: yeah,
- 0:18yeah,
- 0:18yeah,
- 0:19SPEAKER_03: yeah,
- 0:19yeah,
- 0:19SPEAKER_03: yeah,
- 0:19SPEAKER_03: yeah,
- 0:19SPEAKER_03: yeah,
- 0:19SPEAKER_03: yeah,
- 0:19SPEAKER_03: yeah,
- 0:19SPEAKER_03: yeah,
- 0:20yeah,
- 0:20yeah,
- 0:20yeah,
- 0:21yeah,
- 0:21yeah,
- 0:21yeah,
- 0:21yeah,
- 0:21yeah,
- 0:21yeah,
- 0:21yeah,
- 0:22yeah,
- 0:22yeah,
- 0:22yeah,
- 0:22SPEAKER_03: yeah,
- 0:22yeah,
- 0:22yeah,
- 0:23yeah,
- 0:23yeah,
- 0:23yeah,
- 0:23yeah,
- 0:24yeah,
- 0:24yeah,
- 0:25yeah,
- 0:25yeah,
- 0:26SPEAKER_03: yeah,
- 0:26SPEAKER_03: yeah,
- 0:26SPEAKER_03: yeah,
- 0:26yeah
- 0:27SPEAKER_03: Yeah, yeah, yeah, yeah, yeah.
- 0:31Uh, I was really good at football, uh, and I was running back.
- 0:35I'm a player in football.
- 0:37Uh,
- 0:37sometimes I was a big,
- 0:40SPEAKER_03: big,
- 0:41SPEAKER_03: big,
- 0:41SPEAKER_03: big,
- 0:41SPEAKER_03: big,
- 0:42SPEAKER_03: big,
- 0:42SPEAKER_03: big,
- 0:42SPEAKER_03: big,
- 0:42SPEAKER_03: big,
- 0:42SPEAKER_03: big,
- 0:42SPEAKER_03: big,
- 0:43SPEAKER_03: big,
- 0:43SPEAKER_03: big,
- 0:43SPEAKER_03: big,
- 0:44big,
- 0:45big,
- 0:45SPEAKER_03: big,
- 0:46SPEAKER_03: big,
- 0:46SPEAKER_03: big,
- 0:46SPEAKER_03: big,
- 0:47SPEAKER_03: big,
- 0:47SPEAKER_03: big,
- 0:47SPEAKER_03: big,
- 0:47big,
- 0:47big,
- 0:48big,
- 0:48big,
- 0:48big,
- 0:49big,
- 0:49big,
- 0:49big,
- 0:50big,
- 0:50big,
- 0:50big,
- 0:51big,
- 0:51big.
- 0:54SPEAKER_03: What?
- 0:54SPEAKER_03: Crack?
- 0:56SPEAKER_03: How are you?
- 1:44SPEAKER_01: If you don't stop and look around once in a while, you could miss it.
- 1:47SPEAKER_01: If that you don't have to laugh at these, oh, it's horrible.
- 1:49SPEAKER_01: That's it.
- 1:49In other words,
- 1:50it's not enough to kill people,
- 1:51but it's going to be a hit.
- 1:53SPEAKER_03: Anyway, I'm going to play with that question about the free life.
- 2:18Thank you.
- 2:36SPEAKER_02: But no, we're not going to do that.
- 2:42SPEAKER_02: I actually fucked up and gave a presentation.
- 3:05SPEAKER_00: with this voice on one time for an event that was live streamed.
- 3:12And I had accidentally left this on and gave an entire
- 3:17presentation and took
- 3:18SPEAKER_02: questions and answers.
- 3:21There was a live audience and then there was a live streamed
- 3:24audience and the live
- 3:25SPEAKER_02: in person audience had no idea.
- 3:27SPEAKER_02: But the live stream went out and it was this the entire time.
- 3:33SPEAKER_00: And yeah, that was pretty funny.
- 3:35SPEAKER_00: What if I recorded an entire podcast episode with that on?
- 3:41SPEAKER_00: I think it would be entertaining.
- 3:44SPEAKER_00: I think probably everybody would be over it.
- 3:48SPEAKER_00: Probably everybody would be like, yeah, we're not listening to this.
- 3:53SPEAKER_00: And I don't blame you.
- 3:55SPEAKER_00: But anyway, what's up, everybody?
- 3:57SPEAKER_00: Welcome, welcome back to another Solo Dolo Brolo episode.
- 4:03of the nance cast i'm happy to be here i'm happy to see you all
- 4:09here i hope you're
- 4:10all doing well um and yeah what's uh what's up i don't know man
- 4:17not much not much
- 4:18is up over here uh i mean there's a lot up over here a lot
- 4:23going on in my neck of
- 4:24SPEAKER_00: the woods but also more the same really ultimately um
- 4:32But yeah,
- 4:33anyway,
- 4:34here's some things I wanted to talk about in this Solo Dolo
- 4:37Brolo episode.
- 4:39Jeffrey Hinton,
- 4:43the godfather of AI,
- 4:47he's been doing an awful lot of media spots recently.
- 4:51And it seems to be the case that a lot of people just love
- 4:55Jeffrey Hinton.
- 4:56SPEAKER_00: A lot of people are just kind of like enamored by, enamored with.
- 5:02SPEAKER_00: Yeah.
- 5:03SPEAKER_00: A lot of people are just like, oh, Jeffrey Hinton is so awesome.
- 5:06So I've done a few episodes about artificial intelligence and
- 5:13specifically like
- 5:14SPEAKER_00: machine learning and back propagation.
- 5:16SPEAKER_00: I've talked about Jeffrey Hinton before.
- 5:20And I've even talked about backpropagation,
- 5:23his major contribution to artificial intelligence.
- 5:27SPEAKER_00: Talked about it before,
- 5:30SPEAKER_00: but I want to talk about it again because Jeffrey Hinton is,
- 5:34SPEAKER_00: like I said,
- 5:34SPEAKER_00: doing a lot of media spots recently,
- 5:37SPEAKER_00: and he seems to have captivated a fair amount of people.
- 5:42SPEAKER_00: And I'm not really sure what it is he's talking about.
- 5:45SPEAKER_00: I've seen, you know, some thumbnails.
- 5:48SPEAKER_00: I've seen some headlines.
- 5:50You know,
- 5:50Godfather of AI says AI will destroy us all,
- 5:53question mark,
- 5:55and all this stuff.
- 5:59SPEAKER_00: But from what I gather,
- 6:01SPEAKER_00: having not engaged with it,
- 6:04SPEAKER_00: and not to say I dislike Jeffrey Hinton,
- 6:06SPEAKER_00: I respect Jeffrey Hinton.
- 6:08SPEAKER_00: I've read his work.
- 6:09SPEAKER_00: I've studied, you know, things he's done.
- 6:13SPEAKER_00: Yeah.
- 6:15SPEAKER_00: It's just this kind of current round of sensational takes.
- 6:22SPEAKER_00: I'm not really interested in the sensationalism.
- 6:26SPEAKER_00: You have people who are either all hype all the way to the moon.
- 6:30SPEAKER_00: AI is going to save us.
- 6:32SPEAKER_00: Or you have people who are all hype in the other direction.
- 6:35SPEAKER_00: Full on doom.
- 6:37SPEAKER_00: AI is going to destroy the world.
- 6:39SPEAKER_00: I think I sit in the Doomer camp.
- 6:41SPEAKER_00: However, it's more nuanced than that.
- 6:44SPEAKER_00: It's more complex than that, as all things are properly.
- 6:51SPEAKER_00: But yeah, it's not all that interesting.
- 6:54So I'm not exactly sure about what all is being said,
- 6:59but I can imagine...
- 7:01SPEAKER_00: You know, they're calling up Jeffrey Hinton.
- 7:03I don't know,
- 7:03you know,
- 7:04New York Times or CNN or whatever,
- 7:07all these podcasters or whatever.
- 7:08SPEAKER_00: And they're like, Jeffrey Hinton, can you talk to us about AI?
- 7:13And or maybe he's calling up all these outlets and saying,
- 7:16I want to talk to you about AI.
- 7:19SPEAKER_00: And I can imagine it's basically just like,
- 7:21SPEAKER_00: you know,
- 7:21SPEAKER_00: we're going to we're going to destroy ourselves.
- 7:24SPEAKER_00: We're going to outmode.
- 7:26SPEAKER_00: We're going to obsolete ourselves or whatever if we don't do something.
- 7:30SPEAKER_00: So let's do something.
- 7:33SPEAKER_00: I think that's probably the gist of it.
- 7:36SPEAKER_00: So he's, you know, cashing in the alarmism card, I guess.
- 7:41SPEAKER_00: And I'm not really interested to hear what he has to say, oddly.
- 7:47SPEAKER_00: And I've had this conversation with a few people.
- 7:51they're like,
- 7:51oh,
- 7:51the godfather of AI is out here talking about how AI is going
- 7:55to destroy us.
- 7:57SPEAKER_00: And that's just not interesting to me.
- 7:59SPEAKER_00: I understand why it's interesting to people.
- 8:03SPEAKER_00: I definitely do.
- 8:04But Jeffrey Hinton is known as the godfather of AI because it
- 8:09was his team that
- 8:09SPEAKER_00: developed back propagation in the 1980s.
- 8:15SPEAKER_00: The way large language models work...
- 8:19SPEAKER_00: is they are trained using backpropagation.
- 8:23SPEAKER_00: So they,
- 8:24SPEAKER_00: you know,
- 8:25SPEAKER_00: deep learning,
- 8:26SPEAKER_00: machine learning,
- 8:28SPEAKER_00: gradient descent,
- 8:31SPEAKER_00: you know,
- 8:31SPEAKER_00: adversarial networks.
- 8:32SPEAKER_00: There's all these, you know, words.
- 8:34SPEAKER_00: Some of them are buzzwords.
- 8:35SPEAKER_00: Some of them are not buzzwords.
- 8:38But machine learning,
- 8:40I guess,
- 8:41is the broad category that we can apply to.
- 8:45SPEAKER_00: to these things, these LLMs and other algorithms and other machines.
- 8:51But backpropagation is a method of allowing structures to
- 9:00emerge that can then be put to use.
- 9:08SPEAKER_00: In the,
- 9:09SPEAKER_00: I don't know,
- 9:11SPEAKER_00: in the 1950s,
- 9:13SPEAKER_00: 1960s AI heyday,
- 9:15SPEAKER_00: when guys were doing what they call good old-fashioned AI,
- 9:19SPEAKER_00: go-fi,
- 9:21SPEAKER_00: there emerged multiple schools of thought on intelligence.
- 9:28SPEAKER_00: What is intelligence?
- 9:30SPEAKER_00: How does intelligence work?
- 9:31SPEAKER_00: And how can we create an artificial intelligence system?
- 9:37uh,
- 9:37artificial intelligence was,
- 9:40uh,
- 9:41not a term up until,
- 9:42you know,
- 9:42the 1950s.
- 9:44SPEAKER_00: Um,
- 9:46SPEAKER_00: and after the 1950s,
- 9:47it kind of was the official term for what people had already
- 9:52been doing for
- 9:53SPEAKER_00: decades.
- 9:54Um,
- 9:55again,
- 9:56I think modern computation really begins with Charles Babbage
- 10:00and Ada Lovelace.
- 10:02SPEAKER_00: Um,
- 10:04SPEAKER_00: You could,
- 10:04SPEAKER_00: you know,
- 10:04Claude Shannon definitely comes into the mix there with kind of
- 10:08information theory.
- 10:10SPEAKER_00: And then from there, there's like a rich legacy of thought.
- 10:16But I really think like the key figures at the birth of modern
- 10:20computation,
- 10:22SPEAKER_00: modern AI,
- 10:26these guys are Charles Babbage and Claude Shannon and then Miss
- 10:30Ada Lovelace.
- 10:31SPEAKER_00: Yeah.
- 10:33SPEAKER_00: But I also think it goes all, you know, it goes back.
- 10:35SPEAKER_00: It goes back far, far, far.
- 10:36SPEAKER_00: It goes deep, deep, deep into history.
- 10:38SPEAKER_00: You know,
- 10:38SPEAKER_00: you have the story of the Golem,
- 10:42SPEAKER_00: which if you don't know that,
- 10:45SPEAKER_00: just Google the Golem,
- 10:48SPEAKER_00: G-O-L-E-M.
- 10:50SPEAKER_00: It's a Jewish folklore tale.
- 10:57SPEAKER_00: I think particularly like European Jew.
- 10:59SPEAKER_00: Yeah.
- 11:03uh european jewish i don't i don't even know if there's like a
- 11:06differentiation
- 11:07between i don't know i think there is but i'm not 100 sure i
- 11:12almost said it's like
- 11:12the bolsheviks and the mensheviks uh but no there's the
- 11:15ashkenazi and then there's
- 11:16the there's another kind of there's another category i think it
- 11:22starts with m uh
- 11:24that's why i thought mensheviks but anyway it doesn't matter
- 11:28look up the golem
- 11:31SPEAKER_00: Or go back to, you know, ancient Greece.
- 11:35SPEAKER_00: You have what was it?
- 11:36Archimedes was out here trying to create robots and trying to
- 11:39instrumentalize tasks
- 11:41SPEAKER_00: and thinking about artificial intelligence.
- 11:46You also have,
- 11:46I think Descartes is another thinker of artificial
- 11:51intelligence.
- 11:52SPEAKER_00: You know, if you actually read, it's really interesting.
- 11:55SPEAKER_00: If you read the discourse on the method,
- 11:57SPEAKER_00: he talks about artificial intelligence and he doesn't call it artificial
- 12:01SPEAKER_00: intelligence,
- 12:02SPEAKER_00: but,
- 12:02SPEAKER_00: you know,
- 12:02SPEAKER_00: he talks about finding ways to automate activity,
- 12:06SPEAKER_00: including cognition.
- 12:09SPEAKER_00: But again, he doesn't particularly specifically say cognition either.
- 12:13SPEAKER_00: But anyway, artificial intelligence.
- 12:16SPEAKER_00: Kind of a buzzword.
- 12:17SPEAKER_00: It's kind of always been with us.
- 12:22SPEAKER_00: But like the modern version of it goes back to Babbage, Lovelace.
- 12:30SPEAKER_00: Shannon.
- 12:31SPEAKER_00: And then from there, you know, you had the 40s and the 50s.
- 12:36SPEAKER_00: You know, there was a lot of development with computation.
- 12:41SPEAKER_00: And then the 50s, you know, artificial intelligence, AI was kind of born.
- 12:49SPEAKER_00: And then from that, you know, we live in the world that we live in today.
- 12:52SPEAKER_00: And there are some people who are AI maximalists.
- 12:55SPEAKER_00: There are some people who are total AI poo-pooers.
- 13:00SPEAKER_00: And there's everyone in between.
- 13:02SPEAKER_00: But Jeffrey Hinton.
- 13:05SPEAKER_00: Oh, yeah.
- 13:06SPEAKER_00: The schools of thought.
- 13:07SPEAKER_00: So there were kind of differing schools of thought.
- 13:09SPEAKER_00: Some people thought that in order to have an artificial intelligence,
- 13:17SPEAKER_00: you would have to encode meaning,
- 13:20SPEAKER_00: significance,
- 13:21SPEAKER_00: logic.
- 13:24SPEAKER_00: So you would have to, like, by hand, write things.
- 13:31everything you wanted the computer to be able to do everything
- 13:34you wanted the
- 13:34algorithm to be able to do you had to encode it manually um and
- 13:41then there was a
- 13:42school of people who thought that uh structures could emerge um
- 13:51kind of between
- 13:54SPEAKER_00: between the symbolic representations themselves.
- 13:59SPEAKER_00: So almost like structuralism from linguistics.
- 14:06It's almost like this Caesarian computer where it's all nothing
- 14:12but an interrelated
- 14:14SPEAKER_00: web of kind of symbols.
- 14:18SPEAKER_00: So there was one school of thought
- 14:21SPEAKER_00: who believed meaning had to be external.
- 14:25It had to come from somewhere and it had to be kind of imparted
- 14:28into the machine.
- 14:31And then there was this other school of thought who believed
- 14:35meaning could emerge.
- 14:36SPEAKER_00: Significance could emerge.
- 14:37SPEAKER_00: Logic could emerge.
- 14:40SPEAKER_00: And it's really interesting.
- 14:41Early artificial intelligence,
- 14:43they had things that were called expert systems.
- 14:46SPEAKER_00: And there's actually like the early cyberpunk to say early cyberpunk.
- 14:54SPEAKER_00: I mean,
- 14:54SPEAKER_00: there's some people who think there was only ever,
- 14:57SPEAKER_00: you know,
- 14:58SPEAKER_00: one short wave of cyberpunk and everything after is derivative.
- 15:04SPEAKER_00: And sure.
- 15:05SPEAKER_00: I mean,
- 15:06you want to talk about like the new wave of British heavy metal
- 15:09and how like heavy
- 15:10SPEAKER_00: metal there was only ever,
- 15:12SPEAKER_00: you know,
- 15:12SPEAKER_00: Deep Purple was the only heavy metal band.
- 15:16Everything before it was proto metal and everything after it
- 15:19was just derivative.
- 15:21SPEAKER_00: You know, there's no such thing as heavy metal other than Deep Purple.
- 15:25You might have people who want to say there's no such thing as
- 15:28cyberpunk other than
- 15:31SPEAKER_00: like Pat Cadigan and Runner.
- 15:37SPEAKER_00: And there's like a handful of this like first wave of cyberpunk writers.
- 15:43SPEAKER_00: um, stand on Zanzibar and sheep look up.
- 15:45SPEAKER_00: And, uh, it was a cat again wrote sinners, I believe, but yeah, whatever.
- 15:50But like in,
- 15:51uh,
- 15:53in the cyberpunk esque cyberpunk ish fiction of the eighties.
- 15:58SPEAKER_00: And it even kind of leaked out,
- 16:00SPEAKER_00: you know,
- 16:00SPEAKER_00: into the nineties you would have like William Gibson,
- 16:04SPEAKER_00: uh,
- 16:04SPEAKER_00: has talked about expert systems.
- 16:06Uh,
- 16:06but yeah,
- 16:07it,
- 16:07in a lot of the sci-fi of the eighties and early nineties,
- 16:10uh,
- 16:12SPEAKER_00: you'll have writers talking about expert systems.
- 16:18SPEAKER_00: And I think even in...
- 16:21SPEAKER_00: Shadowrun, which is a TTRPG.
- 16:24They also have a few video games and I'm sure they have like
- 16:28some novelizations and
- 16:30SPEAKER_00: stuff like that.
- 16:31SPEAKER_00: Shadowrun,
- 16:32SPEAKER_00: I think you can actually,
- 16:33SPEAKER_00: when you're hacking,
- 16:35SPEAKER_00: you know,
- 16:35you jack into the computer and then you're like in cyberspace
- 16:38and then there's like
- 16:40a virtual representation of you and you can move through
- 16:43virtual space and you can
- 16:45SPEAKER_00: get in virtual combat.
- 16:46SPEAKER_00: You actually have to like combat expert systems and
- 16:51SPEAKER_00: But an expert system,
- 16:53they were so-called expert systems because if you wanted a
- 16:55piece of accounting
- 16:56SPEAKER_00: software,
- 16:59SPEAKER_00: you would get people who knew all the ins and outs of accounting.
- 17:03SPEAKER_00: And they would write all the laws,
- 17:05SPEAKER_00: all the rules,
- 17:06SPEAKER_00: all the norms,
- 17:07SPEAKER_00: all the ins and outs and ups and downs and lefts and rights.
- 17:11They would dot all the T's and cross all the I's when it came
- 17:14to accounting.
- 17:14SPEAKER_00: And then you would have an accounting expert system.
- 17:18SPEAKER_00: So you could automate tasks and you could have it do things for you.
- 17:22SPEAKER_00: And you could even have it work as an agent, in quotes.
- 17:27SPEAKER_00: We're starting to see agents kind of be released to the public today.
- 17:31SPEAKER_00: And people had that vision in the 1980s.
- 17:35SPEAKER_00: Where they had agents,
- 17:36SPEAKER_00: they had virtual agents to accomplish tasks for them to do things,
- 17:40SPEAKER_00: to automate things.
- 17:42SPEAKER_00: And they even achieved it in a very small,
- 17:45SPEAKER_00: very scaled down at a very tiny scale with expert systems.
- 17:53SPEAKER_00: But they were very time consuming and they were very resource heavy.
- 17:58And they never really caught on because they were expensive and
- 18:00they weren't that
- 18:01SPEAKER_00: versatile because they were good at one thing and one thing only.
- 18:06SPEAKER_00: So now my phone,
- 18:07like I can pick my phone up and probably without it even being
- 18:12connected to the
- 18:13SPEAKER_00: Internet.
- 18:15SPEAKER_00: I could have it do all kinds of, you know, useful things for me.
- 18:22SPEAKER_00: calculations and conversions and,
- 18:25SPEAKER_00: uh,
- 18:25moving media around and moving files around and doing all types
- 18:30of stuff and
- 18:30SPEAKER_00: manipulating files in certain kinds of ways and stuff like that.
- 18:33SPEAKER_00: Like you can, you can do an awful lot with one tiny piece of technology.
- 18:39SPEAKER_00: Um,
- 18:40SPEAKER_00: back in the day,
- 18:41SPEAKER_00: if you paid for an expert system that was a accounting agent or whatever,
- 18:47SPEAKER_00: it was only good at accounting and,
- 18:49SPEAKER_00: If you wanted one that was also good at,
- 18:52SPEAKER_00: I don't know,
- 18:52SPEAKER_00: being a manager of your contacts,
- 18:55like your email and phone numbers and physical addresses and
- 18:59stuff like this,
- 19:00like some kind of digital Rolodex system or something with an
- 19:04index that you could
- 19:04SPEAKER_00: search by name or by state or whatever,
- 19:08SPEAKER_00: anything like that,
- 19:08SPEAKER_00: it would be an entirely different expert system.
- 19:10SPEAKER_00: And so you would have to purchase...
- 19:14SPEAKER_00: many,
- 19:14SPEAKER_00: many systems,
- 19:15SPEAKER_00: and it just all became really unwieldy,
- 19:18SPEAKER_00: really bulky,
- 19:19SPEAKER_00: and really inefficient.
- 19:21SPEAKER_00: Because the way these things were coded,
- 19:26SPEAKER_00: they were coded with all the meaning being handwritten into them.
- 19:39SPEAKER_00: Kind of, I don't know, like the metaphor of people think
- 19:42SPEAKER_00: Oh,
- 19:43SPEAKER_00: well,
- 19:43SPEAKER_00: God has to exist because morality,
- 19:46SPEAKER_00: like people,
- 19:47SPEAKER_00: people think like,
- 19:48SPEAKER_00: Oh,
- 19:48SPEAKER_00: since morality is a thing that has to come from somewhere.
- 19:52Now I disagree with that,
- 19:54but I,
- 19:55it's also not like you don't have to disagree with that.
- 19:58SPEAKER_00: Okay.
- 19:58SPEAKER_00: Yeah, sure.
- 19:59SPEAKER_00: Fine.
- 19:59SPEAKER_00: Whatever we can, we can talk about that.
- 20:01That's not what I'm here to talk about,
- 20:02but it's that kind of relationship.
- 20:04SPEAKER_00: People who think me, uh, morality must come from God and
- 20:09SPEAKER_00: It's that same type of thing.
- 20:10Significance or meaning or logic must come from outside the
- 20:16computer system itself.
- 20:18SPEAKER_00: So that was one school of thought.
- 20:21SPEAKER_00: And even the predominant school of thought,
- 20:23SPEAKER_00: like there were corporations marketing these expert systems.
- 20:27And that was kind of like the predominant belief of computer
- 20:30intelligence,
- 20:31SPEAKER_00: machine intelligence at the time was logic had to come from somewhere.
- 20:36SPEAKER_00: But Jeffrey Hinton was like, I don't know about all that, man.
- 20:41SPEAKER_00: I think maybe we can allow computers to do this kind of Caesarian thing.
- 20:48I think maybe we can allow computers to do what they do and
- 20:54structures will emerge.
- 20:56SPEAKER_00: Some type of logic will emerge.
- 20:59SPEAKER_00: And that is how he kind of created back propagation.
- 21:03SPEAKER_00: So he started...
- 21:06SPEAKER_00: with machine vision.
- 21:07He started trying to figure out how to get a computer to
- 21:14recognize an image and
- 21:16SPEAKER_00: reliably spit back out,
- 21:19SPEAKER_00: this is an image of a bird.
- 21:21SPEAKER_00: This is an image of a cat.
- 21:22SPEAKER_00: This is an image of a man.
- 21:23SPEAKER_00: This is an image of a dog.
- 21:27SPEAKER_00: And rather than finding someone
- 21:32SPEAKER_00: to describe a dog or a cat or a bird in such a way that,
- 21:35SPEAKER_00: that you could imagine every single rule that would mean this is,
- 21:44SPEAKER_00: or is not a dog.
- 21:47SPEAKER_00: They set up the system in a way to take an image,
- 21:52SPEAKER_00: you know,
- 21:52SPEAKER_00: a,
- 21:53SPEAKER_00: a sensor,
- 21:54SPEAKER_00: a sensor receives light,
- 21:58SPEAKER_00: the light,
- 21:58SPEAKER_00: uh,
- 21:59SPEAKER_00: in the sensor is transformed into,
- 22:02SPEAKER_00: uh,
- 22:04SPEAKER_00: data,
- 22:04SPEAKER_00: you know,
- 22:05SPEAKER_00: with,
- 22:05with numerical values that then go on and they can be computed
- 22:08and they can be
- 22:09SPEAKER_00: worked with and this and that.
- 22:12SPEAKER_00: Um, um,
- 22:12So it's not like there's an eyeball inside the computer that's
- 22:15looking at the image
- 22:17and that eyeball is connected to a brain and then the brain is
- 22:20running this logic.
- 22:22SPEAKER_00: But, you know, that's a metaphor.
- 22:24SPEAKER_00: What's really going on is they have an image sensor.
- 22:27SPEAKER_00: The image sensor is collecting light.
- 22:29SPEAKER_00: The individual pixels are registering either,
- 22:31SPEAKER_00: you know,
- 22:32SPEAKER_00: light,
- 22:34SPEAKER_00: X percent,
- 22:35SPEAKER_00: dark,
- 22:36SPEAKER_00: you know,
- 22:36SPEAKER_00: zero percent.
- 22:38SPEAKER_00: We'll say 0% is black, 100% is white.
- 22:41SPEAKER_00: I don't know if they had color sensors initially.
- 22:44SPEAKER_00: Probably not.
- 22:45SPEAKER_00: They probably started with just monochrome, black and white sensors.
- 22:50SPEAKER_00: But yeah, it was like, is it on or is it off?
- 22:55SPEAKER_00: Kind of basically.
- 22:58SPEAKER_00: With some gradation.
- 23:00SPEAKER_00: It's not a simple on-off.
- 23:01SPEAKER_00: It's not a simple binary.
- 23:03SPEAKER_00: But what they did is they stacked binary qualifiers on.
- 23:06SPEAKER_00: These things are called perceptrons, and they're artificial neurons.
- 23:10SPEAKER_00: They're little machines, or they could be in software.
- 23:14SPEAKER_00: They could be algorithms,
- 23:15SPEAKER_00: but they're little machines that are binary classifiers,
- 23:19SPEAKER_00: and they operate in either yes or no.
- 23:22SPEAKER_00: Is the light on?
- 23:23SPEAKER_00: Yes.
- 23:23SPEAKER_00: Is the light off?
- 23:25SPEAKER_00: No.
- 23:28SPEAKER_00: Is it on?
- 23:28SPEAKER_00: Yes or no.
- 23:30And they would stack them in this kind of way so that the
- 23:33output goes into another
- 23:37SPEAKER_00: layer of perceptrons or artificial neurons or whatever.
- 23:40SPEAKER_00: And then you get like compound structure.
- 23:42SPEAKER_00: And then at the end of it, you have a list of like, this is a dog.
- 23:47SPEAKER_00: This is a cat.
- 23:48SPEAKER_00: This is a bird.
- 23:48SPEAKER_00: This is a man.
- 23:50SPEAKER_00: So initially...
- 23:52SPEAKER_00: Rather than writing in all the rules of dogness,
- 23:55SPEAKER_00: which we couldn't even do,
- 23:56SPEAKER_00: we would always fail to do that.
- 23:59SPEAKER_00: Or birdness, like we would always fail to write the list of rules.
- 24:02SPEAKER_00: What makes something a bird?
- 24:04SPEAKER_00: What makes something a dog?
- 24:08SPEAKER_00: In general, we would fail to do that.
- 24:10But then you have to abstract one level further because all you
- 24:14have access to is a
- 24:15SPEAKER_00: two-dimensional image.
- 24:17And so thinking of all the rules of every possible
- 24:20two-dimensional image of a dog
- 24:23SPEAKER_00: or a cat or a bird or a man,
- 24:25SPEAKER_00: that would be an impossible task.
- 24:27SPEAKER_00: Jeffrey Hinton was like,
- 24:28SPEAKER_00: yeah,
- 24:28SPEAKER_00: I think maybe let's just try to see if something can emerge,
- 24:32SPEAKER_00: if a structure can emerge that can classify,
- 24:35SPEAKER_00: qualify and classify on its own,
- 24:38SPEAKER_00: that can,
- 24:38SPEAKER_00: in quotes,
- 24:39SPEAKER_00: learn to do this type of classification.
- 24:43SPEAKER_00: So initially...
- 24:44They took a bunch of photos of cats and birds and dogs and men
- 24:49and they trained the
- 24:51machine by having it take a picture and it would go through its
- 24:56little compound
- 25:01SPEAKER_00: architecture and then they would manually select this is cat,
- 25:05SPEAKER_00: this is dog,
- 25:06SPEAKER_00: this is man,
- 25:07SPEAKER_00: this is bird or whatever.
- 25:10SPEAKER_00: And over time, patterns, structures emerged.
- 25:16SPEAKER_00: You have these binary qualifiers,
- 25:19SPEAKER_00: these on-off,
- 25:20SPEAKER_00: these yes-no conditions,
- 25:21SPEAKER_00: and they're stacked in compounding ways.
- 25:25SPEAKER_00: And then they start running it without being told what it is.
- 25:30SPEAKER_00: And rather than telling it,
- 25:32SPEAKER_00: is this a cat or this is a cat,
- 25:33SPEAKER_00: this is a dog,
- 25:34SPEAKER_00: this is a bird,
- 25:35SPEAKER_00: they say,
- 25:36SPEAKER_00: is this a cat,
- 25:36SPEAKER_00: is this a dog,
- 25:37SPEAKER_00: is this a bird?
- 25:40SPEAKER_00: when the machine got it right,
- 25:42SPEAKER_00: you know,
- 25:42SPEAKER_00: they kept that state and reinforced that state when they got it wrong,
- 25:47SPEAKER_00: they,
- 25:47SPEAKER_00: you know,
- 25:47SPEAKER_00: wipe that state and do it again and again and again and again and again.
- 25:51So it was this way to train a machine,
- 25:55to train algorithms,
- 25:56to create emergent structures.
- 26:01SPEAKER_00: Um,
- 26:02SPEAKER_00: So the machine can look at a photo and recognize it is a cat,
- 26:06SPEAKER_00: it is a bird,
- 26:07SPEAKER_00: it is a dog,
- 26:07it is a man without someone having to sit down and write down
- 26:12the unending list of
- 26:14SPEAKER_00: qualities,
- 26:14of observable qualities that make something a man versus a dog
- 26:19or a man versus a
- 26:20woman or a dog versus a cat or a bird versus a bat or any of
- 26:24these things.
- 26:26SPEAKER_00: So that is back propagation.
- 26:28SPEAKER_00: Jeffrey Hinton and his team created that.
- 26:31SPEAKER_00: And that was revolutionary.
- 26:33SPEAKER_00: And that was still very resource hungry, resource intensive.
- 26:36SPEAKER_00: That's why it took three decades to
- 26:40SPEAKER_00: Um, for it to, to really become main mainstreamed.
- 26:47Uh,
- 26:47although it was being commercially implemented,
- 26:50uh,
- 26:51it was the military was using it.
- 26:54Digital photography,
- 26:55um,
- 26:56like you had Nikon or Nikon,
- 26:59uh,
- 26:59Canon,
- 27:00Sony,
- 27:00uh,
- 27:01SPEAKER_00: Like they have autofocus in their cameras.
- 27:05And that's all thanks to like computer vision programs and
- 27:08stuff like that.
- 27:08SPEAKER_00: Really light algorithms.
- 27:11SPEAKER_00: You had surveillance, you know, states, surveillance, all that stuff.
- 27:15SPEAKER_00: So it was being in use.
- 27:17SPEAKER_00: It was just never mainstreamed as like a mainstream consumer product.
- 27:26SPEAKER_00: Until, golly, was it 20...
- 27:32SPEAKER_00: 13 or 16, I can't even remember now.
- 27:35There was a guy who worked at Google who kind of figured out a
- 27:40way to do the same
- 27:42thing that Hinton did with computer vision to do that with
- 27:47natural language or to
- 27:48SPEAKER_00: do it with spoken or rather written language.
- 27:51SPEAKER_00: Yeah.
- 27:53SPEAKER_00: Uh,
- 27:53SPEAKER_00: but they basically did the same thing where with computer vision,
- 27:58SPEAKER_00: there's,
- 27:59SPEAKER_00: you know,
- 27:59SPEAKER_00: numeric value from the image on,
- 28:01SPEAKER_00: on the sensor,
- 28:02like the actual thing that's happening in front of the sensor
- 28:06is converted into
- 28:07SPEAKER_00: data,
- 28:08SPEAKER_00: into,
- 28:08SPEAKER_00: you know,
- 28:10SPEAKER_00: numeric values that the computer ingests and then does its little thing.
- 28:16SPEAKER_00: Um,
- 28:17SPEAKER_00: I think it might have been 2013, but I'm not sure.
- 28:20And I can't remember the guy's name,
- 28:21and I can't remember the name of the project.
- 28:22SPEAKER_00: I should have looked all this up before I started talking.
- 28:25But he basically created the architecture to do this with
- 28:30natural language.
- 28:32SPEAKER_00: And basically,
- 28:36SPEAKER_00: instead of taking an image and allowing structures to emerge,
- 28:40SPEAKER_00: you take text and allow a structure to emerge.
- 28:43SPEAKER_00: But it's the same thing.
- 28:45And you're always getting that human reinforcement on the back
- 28:47end for the pre-training process.
- 28:51SPEAKER_00: To say like, yes, this is significant.
- 28:53SPEAKER_00: Yes, this is meaningful.
- 28:54SPEAKER_00: Yes, this is coherent and legible and all this stuff.
- 28:58SPEAKER_00: The language or the text,
- 29:01SPEAKER_00: you know,
- 29:02SPEAKER_00: the text that's being analyzed is broken down into numeric values.
- 29:07SPEAKER_00: And then it uses gradient descent, which is pretty interesting.
- 29:11I did an episode kind of talking about that,
- 29:13but also kind of failed to explain it.
- 29:17SPEAKER_00: But it's all like numerical values and how they relate to one another.
- 29:20SPEAKER_00: Again, it's very Caesarian.
- 29:23SPEAKER_00: I doubt any... Many.
- 29:25I'm sure some of these guys have,
- 29:27but I doubt many of these guys have read Caesare.
- 29:30SPEAKER_00: But it's all like this interrelated web.
- 29:33SPEAKER_00: There is no outside.
- 29:35SPEAKER_00: And there's nothing, you know, real or external or whatever.
- 29:42SPEAKER_00: It's all there.
- 29:44SPEAKER_00: It's all a self-contained...
- 29:46system or whatever,
- 29:48and it's a system,
- 29:49it's a web,
- 29:50it's a system of connections.
- 29:52SPEAKER_00: But that's going on with large language models,
- 29:54SPEAKER_00: basically the same thing,
- 29:57SPEAKER_00: backpropagation,
- 29:58SPEAKER_00: these structures emerge over iterative,
- 30:06SPEAKER_00: over iterations of classifying,
- 30:09SPEAKER_00: like
- 30:10SPEAKER_00: Classify this image.
- 30:11SPEAKER_00: Is it a dog?
- 30:12SPEAKER_00: Is it a cat?
- 30:12SPEAKER_00: If it, yes.
- 30:13SPEAKER_00: Okay.
- 30:14SPEAKER_00: Then that is a point for you.
- 30:15SPEAKER_00: Now we're going to run it again.
- 30:16SPEAKER_00: See if you get it right.
- 30:17If you continue to get it right,
- 30:18we keep reinforcing whatever structure is already there.
- 30:24SPEAKER_00: You know,
- 30:24SPEAKER_00: we save that structure in the state and we discard all the ones that,
- 30:29SPEAKER_00: that returned a null or a negative response.
- 30:37SPEAKER_00: They do the same thing with language.
- 30:39The reason people talk about AI as a black box or people refer
- 30:44to it as like,
- 30:45SPEAKER_00: oh,
- 30:45SPEAKER_00: this is a mystery or whatever,
- 30:48SPEAKER_00: is because there are now billions of these binary,
- 30:53SPEAKER_00: of these nodes that are binary positions,
- 30:58SPEAKER_00: probably even dozens of billions at this point in these models.
- 31:02SPEAKER_00: So to even take one instance...
- 31:05and try to track that instance would take so much time and so
- 31:08many resources,
- 31:09SPEAKER_00: it would be pointless to do.
- 31:11SPEAKER_00: Now,
- 31:11SPEAKER_00: back when Jeffrey Hinton was doing this,
- 31:13SPEAKER_00: they had on the order of maybe dozens,
- 31:15SPEAKER_00: maybe even up to like 100 or so,
- 31:17SPEAKER_00: maybe.
- 31:18SPEAKER_00: And even then,
- 31:21it was a time-consuming project to actually map out one
- 31:26instance where the machine
- 31:29SPEAKER_00: spits out a positive result.
- 31:30SPEAKER_00: But it could have been done.
- 31:32SPEAKER_00: You know, but now on the order of billions or dozens of billions.
- 31:40SPEAKER_00: Yeah, no one's going to take all the time.
- 31:42SPEAKER_00: No one has all the time to to trace.
- 31:46SPEAKER_00: this back to,
- 31:48to propagate back along the back propagation and kind of find
- 31:53out why it's doing
- 31:54SPEAKER_00: what it's doing.
- 31:56SPEAKER_00: Uh, so that's why they call it a black box.
- 31:58SPEAKER_00: It's, it's not a mystery.
- 31:59SPEAKER_00: We know what's going on.
- 32:01SPEAKER_00: Uh,
- 32:01SPEAKER_00: it's just too time consuming,
- 32:03too resource intensive and not rewarding enough for anybody to
- 32:08actually do it,
- 32:09SPEAKER_00: to go on and map out the topology of this compound complex,
- 32:15SPEAKER_00: uh,
- 32:16SPEAKER_00: of binary qualifiers.
- 32:19SPEAKER_00: But that's why people are so amazed by it because they're like,
- 32:22SPEAKER_00: yeah,
- 32:22SPEAKER_00: we just type in text and it spits out text.
- 32:27SPEAKER_00: Well, it's been trained to do that.
- 32:28SPEAKER_00: It was pre-trained by being told what meaning is.
- 32:35SPEAKER_00: And then it was trained with human feedback and it was nudged and,
- 32:38SPEAKER_00: you know,
- 32:40SPEAKER_00: adjusting the weights and stuff like this.
- 32:42SPEAKER_00: And you can kind of steer it in certain directions.
- 32:45SPEAKER_00: And there's always human feedback.
- 32:47SPEAKER_00: And at the end of the day,
- 32:49SPEAKER_00: there's always a human verifying the result of the AI and saying,
- 32:56SPEAKER_00: yes,
- 32:56SPEAKER_00: this is in fact what we intended to.
- 33:00SPEAKER_00: to be produced.
- 33:01SPEAKER_00: And so there's this element of randomness in quotes.
- 33:07SPEAKER_00: Stochastic is a word I've used a lot.
- 33:10SPEAKER_00: It's a word I like a lot,
- 33:11but I also think it's kind of a boring word to be using because
- 33:16it got big a few
- 33:17SPEAKER_00: years ago.
- 33:18SPEAKER_00: people all of a sudden started talking about stochastic parrots.
- 33:25SPEAKER_00: And I don't remember where it came from,
- 33:27SPEAKER_00: where it started,
- 33:28SPEAKER_00: or when it started,
- 33:29SPEAKER_00: but it was a handful of years ago now.
- 33:32And that was like the buzzword for a while,
- 33:35stochastic parrots,
- 33:36stochastic parrots.
- 33:39SPEAKER_00: And I don't know, people also started talking about stochastic terrorism.
- 33:45SPEAKER_00: maybe around the same time.
- 33:46SPEAKER_00: I wonder which conversation came first.
- 33:49Now that I'm trying to think back,
- 33:50I can't really remember which one came first.
- 33:54SPEAKER_00: Either way, stochastic was a buzzword for a while a few years ago.
- 33:59SPEAKER_00: It seems to have fallen off in popularity.
- 34:02SPEAKER_00: I still like it, but I also don't...
- 34:06SPEAKER_00: I don't know.
- 34:06I think it's probably one of those words that annoys people
- 34:09because it was like a
- 34:10SPEAKER_00: super duper buzzword for a while.
- 34:12But,
- 34:12you know,
- 34:12stochastic is basically just like statistically determined
- 34:15randomness.
- 34:17SPEAKER_00: I just watched a video.
- 34:20SPEAKER_00: Uh, Matt Parker, stand up maths.
- 34:23SPEAKER_00: Excellent, excellent content creator about maths.
- 34:26SPEAKER_00: Uh, Matt Parker.
- 34:28SPEAKER_00: Uh, I, I want to hang out with you.
- 34:30SPEAKER_00: I also would love to have a conversation with you.
- 34:32If anybody,
- 34:33uh,
- 34:34fucks with the Nance cast who also fucks with Matt Parker.
- 34:37SPEAKER_00: Uh, shout out to him.
- 34:40SPEAKER_00: I said, what's up?
- 34:40I would love to talk to him,
- 34:42but I recently watched the video he did about,
- 34:44um,
- 34:46uh,
- 34:47SPEAKER_00: creating black and white images that don't look like shit.
- 34:52And he was talking about digital noise and he did different
- 34:56types of digital noise.
- 34:58SPEAKER_00: He did pure noise, just pure randomness.
- 35:02SPEAKER_00: And of course, that's going to look like shit because it's purely random.
- 35:05SPEAKER_00: But even then, it's not like purely random, right?
- 35:07SPEAKER_00: Like we can still talk about pseudo randomness.
- 35:10SPEAKER_00: But anyway,
- 35:11SPEAKER_00: Random noise versus Bayer interpolation,
- 35:16SPEAKER_00: which I think the two by two looked better than the four by four,
- 35:24SPEAKER_00: if you watched that video.
- 35:26SPEAKER_00: But no, I also understand why the 4x4 looks better.
- 35:30SPEAKER_00: I just prefer the aesthetic of the 2x2.
- 35:32SPEAKER_00: But anyway,
- 35:33SPEAKER_00: and then he created his own interpolation,
- 35:36SPEAKER_00: his own type of noise called the Parker noise,
- 35:39SPEAKER_00: I think,
- 35:40SPEAKER_00: Parker Square.
- 35:42SPEAKER_00: But he also talked about blue noise.
- 35:45SPEAKER_00: And blue noise is pseudo-random properly in that it's random,
- 35:52SPEAKER_00: but it's also statistically determined.
- 35:55And I think that's probably the best way that you can talk
- 35:58about stochastic
- 35:59SPEAKER_00: anything is that it's functionally random up to a point.
- 36:03SPEAKER_00: It's also statistically determined.
- 36:05SPEAKER_00: Yeah.
- 36:07SPEAKER_00: Think about the word stoichiometry.
- 36:10SPEAKER_00: If you have any kids that are high school aged,
- 36:12SPEAKER_00: which probably not many people do,
- 36:14but maybe your kid is taking a chemistry class and maybe
- 36:19they're learning about
- 36:20SPEAKER_00: stoichiometry.
- 36:22SPEAKER_00: Stoichiometry is kind of related to the word stochastic.
- 36:28SPEAKER_00: In that it also has to do with a type of randomness, but also limits.
- 36:34Like stoichiometry is mixing chemicals and trying to understand
- 36:42the results of what
- 36:45SPEAKER_00: your mix will be.
- 36:48SPEAKER_00: So there's an element of randomness-ish, but there's also limits to that.
- 36:54SPEAKER_00: So it's statistically determined.
- 36:55SPEAKER_00: Anyway...
- 36:57SPEAKER_00: God,
- 36:57what the hell was I talking about with why did I even bring up
- 37:01stochastic and then
- 37:01SPEAKER_00: go on that long tangent?
- 37:03SPEAKER_00: Anyway,
- 37:05we take a thing and then we feed it into a machine and the
- 37:15machine chops it up into
- 37:17SPEAKER_00: little molecular bits and observes it.
- 37:23SPEAKER_00: And then we create
- 37:26SPEAKER_00: a pathway in the machine to spit out something.
- 37:33So it takes an input,
- 37:34chops it up into little molecular bits,
- 37:37and it outputs something.
- 37:39SPEAKER_00: That's what all these machines do.
- 37:43SPEAKER_00: And we reinforce the results.
- 37:48SPEAKER_00: So it's going to give a seemingly random output.
- 37:54SPEAKER_00: And when it gives us a desired output,
- 37:58SPEAKER_00: we reinforce whatever state the computer is in at that moment,
- 38:03SPEAKER_00: we reinforce it.
- 38:03SPEAKER_00: So we'll save it.
- 38:06SPEAKER_00: We'll reinforce it and we'll run it through.
- 38:07We'll run more through that same state and do the same thing
- 38:11over and over and
- 38:12SPEAKER_00: over,
- 38:12iteration upon iteration upon iteration until we have a
- 38:16structure that produces a
- 38:18SPEAKER_00: desired result.
- 38:19SPEAKER_00: The desired result necessarily either resembles or refers to the input.
- 38:24SPEAKER_00: based on what we're doing.
- 38:26SPEAKER_00: But I mean,
- 38:26I'm sure you could imagine a situation in which it has no
- 38:31direct reference or
- 38:33SPEAKER_00: resemblance to,
- 38:34SPEAKER_00: but that wouldn't be useful again,
- 38:35SPEAKER_00: would it?
- 38:37SPEAKER_00: But anyway,
- 38:38SPEAKER_00: and all the things that are meaningless,
- 38:40SPEAKER_00: all the things that we don't want,
- 38:43SPEAKER_00: we delete them.
- 38:44SPEAKER_00: We say, you know, wipe that state.
- 38:46SPEAKER_00: Let's start again.
- 38:47SPEAKER_00: Let's do it over and over again.
- 38:48SPEAKER_00: And we just do this, you know, hundreds, thousands of,
- 38:52SPEAKER_00: thousands of thousands,
- 38:55SPEAKER_00: dozens of thousands,
- 38:57SPEAKER_00: millions,
- 38:58SPEAKER_00: dozens of millions,
- 39:00SPEAKER_00: billions even of times over and over and over again.
- 39:03SPEAKER_00: So there's these emergent structures.
- 39:07SPEAKER_00: Because again, we leave it up to statistical determination.
- 39:10SPEAKER_00: We're like, we're going to go hands off.
- 39:11SPEAKER_00: We're going to roll the dice and let the dice fall where they may.
- 39:16SPEAKER_00: But when we get the desired dice roll,
- 39:20SPEAKER_00: we run with those conditions and we do it over and over and over again.
- 39:24SPEAKER_00: And you can even have,
- 39:26SPEAKER_00: you know,
- 39:26manifold of these running in parallel and then you can fold
- 39:29them back in onto each
- 39:30SPEAKER_00: other.
- 39:31SPEAKER_00: It all becomes very complex.
- 39:32There's a lot of compounding,
- 39:34but at the end of the day,
- 39:36we understand what's happening.
- 39:38We also understand that it's far too resource intensive and far
- 39:42too useless to,
- 39:46to trace the,
- 39:48SPEAKER_00: every single state or even one state at this point in time,
- 39:53SPEAKER_00: it's,
- 39:53SPEAKER_00: it would be pointless to trace it all out and,
- 39:56SPEAKER_00: and propagate ourselves along the back propagation algorithm.
- 40:01SPEAKER_00: But it is there.
- 40:01SPEAKER_00: We do know that it's there.
- 40:05SPEAKER_00: So that's the way machine learning works.
- 40:06SPEAKER_00: Deep, deep learning, uh,
- 40:10SPEAKER_00: back propagation, gradient descent, like all this stuff.
- 40:15SPEAKER_00: There is a human involved from the get-go.
- 40:18SPEAKER_00: There's a human validator at the end of it all.
- 40:21SPEAKER_00: They're doing what they have been built to do.
- 40:24SPEAKER_00: They're doing what we expect them to do.
- 40:26SPEAKER_00: So when they do things like hallucinate,
- 40:28it's just because there's these weird emergent structures and
- 40:32we haven't gone into
- 40:33SPEAKER_00: every single thing
- 40:36SPEAKER_00: to verify, is this correct or is this incorrect?
- 40:38SPEAKER_00: So there will be fuzziness.
- 40:43SPEAKER_00: And that's okay, right?
- 40:44SPEAKER_00: I mean, there's fuzziness with humans.
- 40:46I just talked about how it would be impossible to write a list
- 40:49of rules that
- 40:50SPEAKER_00: describes what is a man,
- 40:52SPEAKER_00: what is a dog.
- 40:53SPEAKER_00: That would be an impossible thing to do.
- 40:57SPEAKER_00: So of course,
- 40:58it's going to be impossible to write code that does that and to
- 41:03expect code to do
- 41:04SPEAKER_00: the same thing
- 41:06However,
- 41:07it's also silly to think it's doing the exact same thing that
- 41:14humans are doing.
- 41:15SPEAKER_00: However, I think it is doing something that humans do.
- 41:21SPEAKER_00: I just think humans do more.
- 41:22SPEAKER_00: I think humans do something different.
- 41:24SPEAKER_00: I think that's not all there is to this story.
- 41:27SPEAKER_00: I think there's more to this story.
- 41:29When it comes to human intelligence,
- 41:31once again,
- 41:32you know,
- 41:32what is intelligence?
- 41:34SPEAKER_00: What is intelligence?
- 41:36SPEAKER_00: What is intelligence?
- 41:38SPEAKER_00: I think we do something more and something different.
- 41:45SPEAKER_00: However, I also, I agree.
- 41:48SPEAKER_00: I think this is...
- 41:50SPEAKER_00: a great insight into the way that,
- 41:53SPEAKER_00: uh,
- 41:53SPEAKER_00: that cognition works and maybe even how cognition developed.
- 41:57Um,
- 41:59you know,
- 41:59going with this definition of intelligence,
- 42:03uh,
- 42:04ability to achieve goals.
- 42:09SPEAKER_00: I guess that's the most reductive version, the ability to achieve a goal.
- 42:15SPEAKER_00: Um,
- 42:17SPEAKER_00: or proficiency at playing games.
- 42:19I would argue that proficiency at playing games and the ability
- 42:22to achieve a goal
- 42:23SPEAKER_00: is very similar,
- 42:25SPEAKER_00: very functionally similar.
- 42:27Maybe not exactly similar because maybe a game presupposes
- 42:31something that a goal
- 42:32SPEAKER_00: doesn't presuppose and vice versa.
- 42:36SPEAKER_00: But in general,
- 42:37the most widely agreed upon definition of intelligence is the
- 42:41ability to achieve a
- 42:44SPEAKER_00: goal and
- 42:47then you'll have people,
- 42:48you know,
- 42:49the ability to achieve compound goals.
- 42:53Then you'll have,
- 42:53you know,
- 42:54and this and that you can go on and on and on.
- 42:55But I think the most reductive definition of intelligence that
- 42:59would be widely
- 43:00SPEAKER_00: accepted is the ability to achieve a goal.
- 43:03SPEAKER_00: And I think,
- 43:03SPEAKER_00: yeah,
- 43:04SPEAKER_00: the way that this kind of emergent structure operates,
- 43:10that probably does tell us something about how biology
- 43:14discovered intelligence and
- 43:16SPEAKER_00: before computers discovered intelligence.
- 43:22SPEAKER_00: Now, are computers other than biology?
- 43:28SPEAKER_00: Are computers just one aspect of human activity?
- 43:33SPEAKER_00: And humans are just one very tiny subset of biology.
- 43:39SPEAKER_00: So you could say there's nothing new under the sun.
- 43:42SPEAKER_00: There's nothing new here.
- 43:43SPEAKER_00: Nothing has been
- 43:45SPEAKER_00: Nothing has happened.
- 43:46Biology discovered intelligence and computers are just part of
- 43:51what humans do and
- 43:52SPEAKER_00: humans are just part of biology.
- 43:53SPEAKER_00: And so therefore it's nothing new.
- 43:55SPEAKER_00: I don't know.
- 43:58SPEAKER_00: Maybe, maybe not.
- 43:59I do think it's probably significant that we have machines that
- 44:06are external to our biology and
- 44:11And that are able to complete tasks that,
- 44:16you know,
- 44:17heretofore only we could complete.
- 44:20SPEAKER_00: I do think that's interesting.
- 44:26SPEAKER_00: However,
- 44:28there's something else going on in the human that I call
- 44:32subjectivity that many,
- 44:33SPEAKER_00: many,
- 44:33SPEAKER_00: many,
- 44:33SPEAKER_00: many,
- 44:34SPEAKER_00: many people...
- 44:35SPEAKER_00: call subjectivity, but also some people don't call it subjectivity.
- 44:38SPEAKER_00: And that's fine too.
- 44:41But there's also this element of negativity when we talk about
- 44:45what humans are,
- 44:47what humans do.
- 44:49SPEAKER_00: Humans are not identical with themselves.
- 44:53Humans are not identical with their biological reality,
- 44:57with their material identity.
- 45:00SPEAKER_00: We have a
- 45:04SPEAKER_00: So like that, we're already split.
- 45:07SPEAKER_00: We're already denatured.
- 45:12SPEAKER_00: Because my symbolic identity is different from my material identity.
- 45:17SPEAKER_00: The self.
- 45:19SPEAKER_00: Now, I don't know.
- 45:20SPEAKER_00: Do dogs think of themselves as selves?
- 45:23SPEAKER_00: Maybe there's some dim selfness with dogs.
- 45:26SPEAKER_00: Maybe dolphins and octopi, octopodes, octopuses.
- 45:31SPEAKER_00: Maybe these things have their own selves or whatever.
- 45:35SPEAKER_00: Maybe monkeys have selves.
- 45:36SPEAKER_00: I think monkeys probably do have selves because they get sad.
- 45:41Um,
- 45:41that whole,
- 45:42there's something going on called punch the monkey right now.
- 45:46SPEAKER_00: And,
- 45:46SPEAKER_00: uh,
- 45:47he's a monkey and people are sad about him because he's sad and
- 45:52he's sad because
- 45:53SPEAKER_00: people won't like hug him or some other monkeys won't hug him.
- 45:57SPEAKER_00: So yeah, I think monkeys probably have selfs.
- 46:00SPEAKER_00: probably dim, maybe proto-selves.
- 46:03SPEAKER_00: So it's not to say that selves are unique to humans,
- 46:08SPEAKER_00: but I do think that humans take the self,
- 46:11SPEAKER_00: the symbolic identity,
- 46:13SPEAKER_00: and run with it in a way that other species just aren't able to do.
- 46:19SPEAKER_00: And maybe it's because we have more processing power.
- 46:21Maybe it's because through language,
- 46:24we've found a way to externalize cognition.
- 46:27SPEAKER_00: I almost said computation.
- 46:30And I don't know,
- 46:32man,
- 46:32these are questions that no one has an answer for.
- 46:34SPEAKER_00: Don't expect me to have an answer for a lot of people have thoughts on.
- 46:39SPEAKER_00: But yeah, so that the self is already distinct.
- 46:45SPEAKER_00: From the material identity,
- 46:47SPEAKER_00: the material identity would is just,
- 46:49SPEAKER_00: well,
- 46:49SPEAKER_00: I'm,
- 46:49SPEAKER_00: you know,
- 46:50trillions of cells locked together that that kind of create a
- 46:54meta system.
- 46:56SPEAKER_00: You can if you zoom in on my eyeball,
- 46:58SPEAKER_00: you'll see individual living cells and you'll be like,
- 47:00SPEAKER_00: that's that's an organism because it's a cell.
- 47:04SPEAKER_00: And it's life and blah, blah, blah.
- 47:06SPEAKER_00: And what is life in the first place?
- 47:08SPEAKER_00: Life is another one of these things.
- 47:09SPEAKER_00: There's no perfect definition of life.
- 47:12SPEAKER_00: There's no universally agreed upon definition of life.
- 47:19Some people have said,
- 47:20you know,
- 47:22life is anything that is able to fear death is alive.
- 47:26SPEAKER_00: I don't know about that.
- 47:29SPEAKER_00: There's a lot of people,
- 47:30SPEAKER_00: I think,
- 47:31SPEAKER_00: again,
- 47:31SPEAKER_00: the most widely agreed upon,
- 47:33SPEAKER_00: though imperfect and not universally accepted,
- 47:36SPEAKER_00: definition of life is the ability to multiply or reproduce,
- 47:41SPEAKER_00: the ability to move around or motility,
- 47:44SPEAKER_00: and metabolism,
- 47:46SPEAKER_00: the ability to eat and excrete,
- 47:49SPEAKER_00: or the ability to convert resources into waste.
- 47:53SPEAKER_00: Okay.
- 47:54SPEAKER_00: Motility, multiplication, metabolism, the three Ms.
- 48:00SPEAKER_00: That's the most widely agreed upon definition of life.
- 48:03SPEAKER_00: And even that is imperfect.
- 48:04SPEAKER_00: Even that fails and it's not universally accepted by biologists.
- 48:12SPEAKER_00: But yeah,
- 48:14so we have a symbolic identity that is separate from our
- 48:18material identity,
- 48:20SPEAKER_00: our biological identity.
- 48:21SPEAKER_00: So we're already not self-identical.
- 48:24SPEAKER_00: whereas computers don't have that.
- 48:29SPEAKER_00: Now,
- 48:29SPEAKER_00: we also further go on with the human subject,
- 48:35SPEAKER_00: and we recognize that we have a social existence.
- 48:39SPEAKER_00: We recognize that I am living,
- 48:42SPEAKER_00: I am a motile,
- 48:45SPEAKER_00: metabolizing,
- 48:46SPEAKER_00: multiplying metasystem in my own right,
- 48:51SPEAKER_00: but in the eyes of the other,
- 48:54SPEAKER_00: I have a social life.
- 48:56I represent that self or that identity in the self and the
- 49:01identity are not
- 49:02SPEAKER_00: identical things either.
- 49:03SPEAKER_00: Um,
- 49:04SPEAKER_00: but I like,
- 49:06SPEAKER_00: I occupy space,
- 49:09SPEAKER_00: social space in the,
- 49:11SPEAKER_00: in the world of the other,
- 49:12SPEAKER_00: just as the other occupies social space in my world.
- 49:17SPEAKER_00: Um, um,
- 49:18SPEAKER_00: So again,
- 49:19SPEAKER_00: we have this disconnection from identity,
- 49:24SPEAKER_00: material identity,
- 49:26SPEAKER_00: symbolic identity,
- 49:28SPEAKER_00: social identity.
- 49:30SPEAKER_00: So there's already these kind of moments of severance.
- 49:34SPEAKER_00: There's already these moments where we are denatured,
- 49:38SPEAKER_00: we are de-worlded,
- 49:39SPEAKER_00: or we're reborn or whatever.
- 49:43In the eyes of ourselves,
- 49:44in the eyes of the other and so on and so forth.
- 49:51And not only do we recognize kind of social relations and we
- 49:55recognize there's more
- 49:57SPEAKER_00: to life than just the metabolic,
- 50:00SPEAKER_00: motile,
- 50:01SPEAKER_00: multiplicative stuff that's happening at the level of this meta system.
- 50:07SPEAKER_00: There's a social life.
- 50:09SPEAKER_00: And I am a person to this other person.
- 50:12SPEAKER_00: I represent something to this other person.
- 50:15So when we talk about like meaning and life and significance
- 50:18and all this stuff,
- 50:19SPEAKER_00: it's really going on at that level,
- 50:21at that level that's already two steps removed from what's
- 50:26happening at a material
- 50:28SPEAKER_00: level.
- 50:29And we take that and we run with it and we do more interesting
- 50:32things with it,
- 50:33right?
- 50:33SPEAKER_00: We have this inbuilt kind of
- 50:37SPEAKER_00: motor of undermining where we constantly undermine ourselves.
- 50:42SPEAKER_00: Uh,
- 50:42SPEAKER_00: and we cast off the symbolic identity,
- 50:45SPEAKER_00: um,
- 50:47SPEAKER_00: because we're kind of like,
- 50:49SPEAKER_00: there's this tension,
- 50:50SPEAKER_00: there's this contradiction,
- 50:51SPEAKER_00: um,
- 50:52There's what I see myself as and there's what the other sees me
- 50:56as that is contradictory.
- 50:59SPEAKER_00: And so that kind of provides a motor for things to happen.
- 51:04And we go on and we do things and we're always constantly going
- 51:07through life asking,
- 51:09who am I?
- 51:10SPEAKER_00: What am I?
- 51:12SPEAKER_00: Why?
- 51:13SPEAKER_00: And so on and so forth.
- 51:17So,
- 51:18yeah,
- 51:20the way that these computer guys have worked with intelligence,
- 51:26it's interesting.
- 51:26And I do think it does tell us something about the way
- 51:30intelligence evolved,
- 51:32SPEAKER_00: the way biology discovered intelligence,
- 51:35SPEAKER_00: emergent structure.
- 51:40The thing is with biological intelligence,
- 51:42as far as we know,
- 51:44there was no...
- 51:47SPEAKER_00: human feedback layer.
- 51:49SPEAKER_00: There was no external validator.
- 51:54SPEAKER_00: There was no God.
- 51:56SPEAKER_00: Now, maybe you do believe that that's the case and fair enough.
- 51:58But for those of you who don't,
- 52:00for those of us who don't,
- 52:03there was no external logic.
- 52:05There was no external rule set that was validating intelligence
- 52:10as it was discovering itself.
- 52:12SPEAKER_00: So it took an awful lot longer
- 52:15SPEAKER_00: Um, and, uh, and the fuzziness is a necessary part of it.
- 52:21SPEAKER_00: Um, but the fuzziness, we always want to disavow the fuzziness.
- 52:24SPEAKER_00: You know, we're always looking for ways to disavow the fuzziness.
- 52:27And,
- 52:28uh,
- 52:28actually in,
- 52:28in the most recent Nancecast episode I did with Sean,
- 52:31um,
- 52:33SPEAKER_00: Hail the shining future.
- 52:34SPEAKER_00: We get into that a little bit on why we need definite quantities,
- 52:39SPEAKER_00: why we need kind of definite categories for all things.
- 52:43Because the world we live in today is a capitalist world and
- 52:46capital demands definite articles.
- 52:52SPEAKER_00: There...
- 52:54SPEAKER_00: There's no room for nuance.
- 52:56SPEAKER_00: There's no room for complexity.
- 52:58SPEAKER_00: There's no room for gray areas when everything has to be a commodity.
- 53:03So what capitalism did is it came into this world and it
- 53:08eradicated everything and
- 53:11SPEAKER_00: it replaced everything.
- 53:15with exchange exchange value and not everything i'm being
- 53:18hyperbolic and it didn't
- 53:20actually eradicate anything that's that's therein lies the rub
- 53:24my friends it didn't
- 53:25actually eradicate anything however it also very much did
- 53:30eradicate everything and
- 53:31that's the contradiction we're trying to figure out that's the
- 53:35contradiction
- 53:36SPEAKER_00: contradiction we're tarrying with here in modernity um
- 53:41SPEAKER_00: But capitalism came in and eradicated the things as they were.
- 53:49SPEAKER_00: And it eradicated the utility and the facticity of all things.
- 53:53SPEAKER_00: And it replaced that all with exchange value or proposed exchange value.
- 53:58SPEAKER_00: And that has gone on to be further modulated.
- 54:03SPEAKER_00: But the way the world works,
- 54:06SPEAKER_00: the logic of the world,
- 54:07SPEAKER_00: everything has to represent exchange value in some way.
- 54:10Everything has to be able to be entered into the cosmic
- 54:16spreadsheet of exchange value.
- 54:18And you just can't do that when things are too fuzzy,
- 54:21when things are too unknown.
- 54:23SPEAKER_00: Everything has to be definite.
- 54:25SPEAKER_00: Um, that's why we're, you know, living in the library of Babel.
- 54:29SPEAKER_00: That's why we're living in this world of infinite categorization,
- 54:32SPEAKER_00: classification,
- 54:34SPEAKER_00: and definition.
- 54:35SPEAKER_00: Uh, and it happens at the level of, of consumer products.
- 54:41SPEAKER_00: Oh, my Google phone is so much better than your Apple phone.
- 54:45SPEAKER_00: Uh, it also happens at the level of, of like sociality.
- 54:48SPEAKER_00: Oh, I'm a punk rocker.
- 54:49SPEAKER_00: I'm a skateboarder.
- 54:50SPEAKER_00: You're a jock.
- 54:51SPEAKER_00: You're a basketball player.
- 54:53SPEAKER_00: I don't like you.
- 54:54SPEAKER_00: Um,
- 54:55SPEAKER_00: It happens at the level of religion.
- 54:58SPEAKER_00: I'm a Christian.
- 54:59SPEAKER_00: I'm an atheist.
- 55:00SPEAKER_00: I'm a Jew.
- 55:01SPEAKER_00: I'm a Muslim.
- 55:02SPEAKER_00: I'm a this.
- 55:03SPEAKER_00: I'm a that.
- 55:03SPEAKER_00: It happens at the level of state and nation.
- 55:06SPEAKER_00: I'm an American.
- 55:07SPEAKER_00: I'm a Mexican.
- 55:08SPEAKER_00: I'm a Canadian.
- 55:09SPEAKER_00: I'm a Dane.
- 55:12SPEAKER_00: I'm a Swede.
- 55:12SPEAKER_00: I'm a this.
- 55:13SPEAKER_00: I'm a that.
- 55:13SPEAKER_00: It happens in all things.
- 55:15SPEAKER_00: Capital eradicates nuance,
- 55:17SPEAKER_00: eradicates complexity,
- 55:18SPEAKER_00: and replaces it with the Excel spreadsheet at the end of time.
- 55:24SPEAKER_00: Um, and we're constantly trying to disavow this fuzziness.
- 55:27SPEAKER_00: That's why like AI hallucinations,
- 55:30SPEAKER_00: I think freak us out so much because it presents,
- 55:34SPEAKER_00: it kind of spits it right back into our face.
- 55:35SPEAKER_00: It spits the fuzziness right back into our face.
- 55:37SPEAKER_00: And I think that frightens a lot of people.
- 55:40SPEAKER_00: Um, but it's also a wonderful thing, right?
- 55:45Because at the level of us,
- 55:46uh,
- 55:47at the level of us,
- 55:49at the level of,
- 55:49of people,
- 55:50um,
- 55:51SPEAKER_00: That opens up space.
- 55:52That opens up freedom for us to kind of figure things out and
- 55:56make significance,
- 55:57SPEAKER_00: generate new significance.
- 55:59SPEAKER_00: Poetics, right?
- 56:02SPEAKER_00: And capital abhors poetry.
- 56:09SPEAKER_00: Or poetics, if you will.
- 56:14SPEAKER_00: But yeah, I do think the computer guys are onto something.
- 56:20SPEAKER_00: Um, however, that is not all that humans do.
- 56:27SPEAKER_00: That's, that's not what intelligence is.
- 56:29SPEAKER_00: That's not what we do.
- 56:30SPEAKER_00: That's not what we are.
- 56:31SPEAKER_00: We're so much more than that.
- 56:33SPEAKER_00: And we're also so much less than that.
- 56:36SPEAKER_00: Right.
- 56:36Cause it all does come down to negation,
- 56:40contradiction,
- 56:42impossibility,
- 56:44um,
- 56:46SPEAKER_00: And that's not what capital wants.
- 56:48And that's not what the people who are developing these
- 56:52algorithms and these
- 56:53SPEAKER_00: complex machines,
- 56:55SPEAKER_00: they don't want that.
- 56:56SPEAKER_00: They don't want lack.
- 56:58SPEAKER_00: They don't want negation.
- 56:59SPEAKER_00: They don't want contradiction.
- 57:00They want utility or simulated utility so that these things can
- 57:05represent higher
- 57:06SPEAKER_00: and higher values of exchange value.
- 57:09Because everything is reduced down to economic activity here in
- 57:11the world we live in today.
- 57:13SPEAKER_00: So...
- 57:15SPEAKER_00: that's why I kind of don't care about what Jeffrey Hinton has to say.
- 57:21SPEAKER_00: Um, although I, I, I do care about what Jeffrey Hinton has to say.
- 57:27SPEAKER_00: And like I said, I respect the shit out of Jeffrey Hinton.
- 57:30SPEAKER_00: He is, his work was revolutionary, uh, and his team as well.
- 57:33Uh,
- 57:34he instructed,
- 57:36uh,
- 57:36a group of people who went on to all be very important people
- 57:40in,
- 57:41uh,
- 57:42SPEAKER_00: kind of computer tech machine learning world today.
- 57:47SPEAKER_00: So I don't want to come off flippant or disrespectful or whatever,
- 57:50SPEAKER_00: but I just wanted to talk about this.
- 57:53SPEAKER_00: You know,
- 57:54SPEAKER_00: I saw Jeffrey Hinton is doing a lot of media spots,
- 57:57SPEAKER_00: had some conversation about it a little bit.
- 57:59And it just kind of struck me like we're not going to find the
- 58:06answers from people who only see
- 58:12SPEAKER_00: this one side of it, the kind of instrumental functional side of it.
- 58:18SPEAKER_00: I agree.
- 58:19Like I've said before,
- 58:20like sure,
- 58:20call it intelligence,
- 58:21then we need something else.
- 58:22SPEAKER_00: That's why I started searching for different terms,
- 58:26SPEAKER_00: better terms,
- 58:27SPEAKER_00: you know,
- 58:27SPEAKER_00: intelligence as opposed to men intelligence or,
- 58:31SPEAKER_00: you know,
- 58:31SPEAKER_00: subjectivity and all of these words fail and they're bound to fail.
- 58:37SPEAKER_00: And that's the point we have to allow them to fail.
- 58:42SPEAKER_00: But yeah,
- 58:42SPEAKER_00: just only being concerned with the intelligence side of it and not really
- 58:47SPEAKER_00: understanding that the human is already this fractured thing.
- 58:56SPEAKER_00: We are not identical with ourselves.
- 58:59And it's because that split,
- 59:02that negation,
- 59:04that incongruity,
- 59:05that discontinuity,
- 59:07whatever...
- 59:09SPEAKER_00: That we do anything that we recognize as human in the first place.
- 59:13SPEAKER_00: If all we did was was chase, you know, utility.
- 59:19SPEAKER_00: then we wouldn't have poetry.
- 59:22SPEAKER_00: We wouldn't have music.
- 59:23SPEAKER_00: We wouldn't have art.
- 59:24SPEAKER_00: We wouldn't have, you know, love.
- 59:26SPEAKER_00: We wouldn't have all these things that we claim are important to us.
- 59:31SPEAKER_00: We wouldn't even have, you know, the notion of God.
- 59:33SPEAKER_00: You know,
- 59:34a lot of these people who are very functional and who are
- 59:37concerned with just the
- 59:39SPEAKER_00: functional side,
- 59:40SPEAKER_00: just intelligence,
- 59:41SPEAKER_00: bare intelligence,
- 59:44SPEAKER_00: you know,
- 59:44SPEAKER_00: a lot of them also at least make movements.
- 59:47SPEAKER_00: toward God or sometimes point to God and this and that.
- 59:50And it's like,
- 59:51well,
- 59:51yeah,
- 59:51but God is this,
- 59:53God is the God of the gaps,
- 59:54right?
- 59:55SPEAKER_00: God is this thing born of negativity.
- 59:59SPEAKER_00: Uh,
- 59:59SPEAKER_00: so that just strikes me,
- 1:00:01SPEAKER_00: uh,
- 1:00:01as a bit amusing that these people who have already totalized
- 1:00:06the world and
- 1:00:07SPEAKER_00: instrumentalize the world,
- 1:00:09SPEAKER_00: um,
- 1:00:10SPEAKER_00: also imagine that there's an outside,
- 1:00:12SPEAKER_00: um,
- 1:00:13SPEAKER_00: um,
- 1:00:15SPEAKER_00: I don't know, man.
- 1:00:16SPEAKER_00: It's like we're living in this emergent structure,
- 1:00:20SPEAKER_00: but you can't take up the emergent quality of it all.
- 1:00:26SPEAKER_00: You're like, no, but it still has to have had to come from somewhere.
- 1:00:28SPEAKER_00: It's like, okay, man, sure.
- 1:00:31SPEAKER_00: I don't know.
- 1:00:31SPEAKER_00: And I don't have answers to those questions either.
- 1:00:33SPEAKER_00: And they're not that simple.
- 1:00:35SPEAKER_00: They're not that reductionistic.
- 1:00:37SPEAKER_00: I'm being hyperbolic.
- 1:00:38SPEAKER_00: I'm being reductionistic.
- 1:00:40SPEAKER_00: because it is what it is,
- 1:00:41because I don't have the time or the resources to deal with all
- 1:00:44those questions.
- 1:00:46SPEAKER_00: So I'm not trying to be flippant or disrespectful to anybody.
- 1:00:48SPEAKER_00: I'm really not.
- 1:00:50SPEAKER_00: It just, I don't know.
- 1:00:51SPEAKER_00: It struck me.
- 1:00:51SPEAKER_00: Like I said, Jeffrey Hinton is out there doing a bunch of media spots.
- 1:00:56SPEAKER_00: I think it's important that we concern ourselves with these notions,
- 1:01:00SPEAKER_00: intelligence,
- 1:01:01SPEAKER_00: subjectivity,
- 1:01:02SPEAKER_00: life,
- 1:01:03SPEAKER_00: meaning,
- 1:01:04SPEAKER_00: love,
- 1:01:04SPEAKER_00: joy,
- 1:01:05SPEAKER_00: happiness,
- 1:01:06SPEAKER_00: completion,
- 1:01:06SPEAKER_00: blah,
- 1:01:07SPEAKER_00: blah,
- 1:01:07SPEAKER_00: blah,
- 1:01:07SPEAKER_00: all these things.
- 1:01:10But I don't think the answer is going to come from the guy who
- 1:01:15just makes tools.
- 1:01:19SPEAKER_00: You know,
- 1:01:19you have to find someone who is aware of the negation that
- 1:01:27these tools themselves
- 1:01:29SPEAKER_00: present or deal with or whatever.
- 1:01:33SPEAKER_00: And yeah, but yeah, like I said, I'd love to talk to Matt Parker.
- 1:01:36SPEAKER_00: I would also love to talk to Jeffrey Hinton.
- 1:01:38SPEAKER_00: Shout out Jeffrey Hinton.
- 1:01:39Somebody email him and tell him to hang out with me and Mikey
- 1:01:43for an afternoon.
- 1:01:44SPEAKER_00: That'd be a wonderful afternoon.
- 1:01:46SPEAKER_00: And any of these guys as well.
- 1:01:47SPEAKER_00: And I do intend to reach out to some of these people.
- 1:01:51SPEAKER_00: Not, you know, not Jeffrey Hinton.
- 1:01:53SPEAKER_00: But,
- 1:01:53SPEAKER_00: you know,
- 1:01:54I do want to reach out to more people in the kind of computer
- 1:01:57tech world or
- 1:01:58SPEAKER_00: whatever.
- 1:01:58SPEAKER_00: Because I think we need more crossover.
- 1:02:02from whatever the fuck it is I talk about and think about in
- 1:02:07the computer tech world.
- 1:02:09Because I think the real problem is not that we're going to
- 1:02:17create Skynet.
- 1:02:18SPEAKER_00: Maybe we'll create Skynet.
- 1:02:20SPEAKER_00: I doubt it.
- 1:02:21I think it's far more likely that we'll just deplete all our
- 1:02:24resources before
- 1:02:25SPEAKER_00: Skynet is ever realized.
- 1:02:28SPEAKER_00: And it's not that we're going to...
- 1:02:31automate everything and put everybody out of work and that'll
- 1:02:33be the end of,
- 1:02:34SPEAKER_00: of the human race or humanity as such or whatever.
- 1:02:38I do know that the massive automation and the massive upheaval
- 1:02:44that's going to
- 1:02:45SPEAKER_00: happen economically will be huge and important and catastrophic even,
- 1:02:51SPEAKER_00: but that even that won't be the end.
- 1:02:53The real problem with AI,
- 1:02:55the real problem standing,
- 1:02:57um,
- 1:02:58you know,
- 1:02:58across,
- 1:02:59um,
- 1:03:00Barring,
- 1:03:01standing across the highway,
- 1:03:02barring us from the future,
- 1:03:04I think is objectification.
- 1:03:07SPEAKER_00: Ways in which we are coming to imitate objects ourselves.
- 1:03:11SPEAKER_00: Ways in which we are coming to imitate the machine.
- 1:03:14And that,
- 1:03:15you know,
- 1:03:15read the piece Mikey and I wrote,
- 1:03:18Enjoy Your Surveillance.
- 1:03:20SPEAKER_00: Uh,
- 1:03:21SPEAKER_00: I,
- 1:03:21SPEAKER_00: I think that the,
- 1:03:22SPEAKER_00: the super egoic element of capital and,
- 1:03:24SPEAKER_00: and all this that's going on,
- 1:03:26SPEAKER_00: I think that's contributing to an extremely deleterious process.
- 1:03:30SPEAKER_00: And I think that will be the end of humanity.
- 1:03:32SPEAKER_00: If anything is the end of humanity.
- 1:03:35SPEAKER_00: I don't think it'll be,
- 1:03:37SPEAKER_00: you know,
- 1:03:37SPEAKER_00: killer killer robots from outer space,
- 1:03:40SPEAKER_00: uh,
- 1:03:40or killer robots from labs in Northern California or China or
- 1:03:47any of these places.
- 1:03:49SPEAKER_00: Um, um,
- 1:03:50SPEAKER_00: And I don't think it'll be a complete economic destruction of us.
- 1:03:58Although I do think it will be,
- 1:03:59it is going to be,
- 1:04:00it is going to be catastrophic.
- 1:04:02SPEAKER_00: Not might be,
- 1:04:03SPEAKER_00: not could maybe,
- 1:04:06I think it is going to be catastrophic for us and we must find
- 1:04:10some way to deal
- 1:04:10SPEAKER_00: with it.
- 1:04:11SPEAKER_00: But even that won't be it.
- 1:04:13SPEAKER_00: The real problem is just like PKD tried to warn us.
- 1:04:20SPEAKER_00: And do androids dream more human than human?
- 1:04:25SPEAKER_00: That's what he was tapped into.
- 1:04:26SPEAKER_00: Like we are becoming objects.
- 1:04:32SPEAKER_00: That's the real problem.
- 1:04:33SPEAKER_00: That will be the real end of humanity as such.
- 1:04:37SPEAKER_00: It's not that we're,
- 1:04:39SPEAKER_00: you know,
- 1:04:39creating Skynet or the Decepticons are going to come destroy us
- 1:04:43and enslave us.
- 1:04:45SPEAKER_00: I mean, these things all could happen very much so.
- 1:04:48SPEAKER_00: But even then, I think, you know,
- 1:04:50SPEAKER_00: Humanity would find a way to carry on just like Jeff Goldblum.
- 1:04:55SPEAKER_00: Life finds a way.
- 1:04:58SPEAKER_00: I think humanity will find a way.
- 1:05:00But no,
- 1:05:01I think the real danger in all this stuff is this superegoic
- 1:05:05objectification.
- 1:05:07SPEAKER_00: And for more on that, read the piece Mikey and I wrote.
- 1:05:11SPEAKER_00: It's currently up on his medium, The Dangerous Maybe.
- 1:05:13SPEAKER_00: It's currently up on my substack, 3 Billion Nances.
- 1:05:16SPEAKER_00: Enjoy your surveillance.
- 1:05:18SPEAKER_00: We're working on a few more pieces in this series talking about superego,
- 1:05:24SPEAKER_00: capital,
- 1:05:25SPEAKER_00: AI,
- 1:05:25SPEAKER_00: tech,
- 1:05:26SPEAKER_00: objectification,
- 1:05:27SPEAKER_00: all this stuff that's going on.
- 1:05:31SPEAKER_00: So yeah, go there for all of that.
- 1:05:33Stay tuned to the NanceCast,
- 1:05:343 Billion Nances on YouTube,
- 1:05:37Substack,
- 1:05:38and on Patreon.
- 1:05:40SPEAKER_00: The Dangerous Maybe on YouTube, Substack, and on Patreon.
- 1:05:45SPEAKER_00: And all our homies, all the fellers and the fellarettes and the fells.
- 1:05:55SPEAKER_00: Yeah.
- 1:05:55SPEAKER_00: Anyway, thanks for coming to my Solo Dolobrolo episode.
- 1:06:04SPEAKER_00: I'm going to go ahead and boogie woogie oogie.
- 1:06:06I got some stuff to do,
- 1:06:08but yeah,
- 1:06:08I'm going to try to get this out right now.
- 1:06:11SPEAKER_00: I hope you like it.
- 1:06:12SPEAKER_00: I hope you come back for more and we'll catch you on the next one.
- 1:06:53SPEAKER_01: It's pretty bad.
- 1:06:53SPEAKER_01: You don't stop and look around once in a while.
- 1:06:56SPEAKER_01: You could miss it.
- 1:06:56SPEAKER_01: It's that you don't have to laugh at these.
- 1:06:58SPEAKER_01: Oh, it's horrible.
- 1:06:59SPEAKER_01: That's it.
- 1:06:59SPEAKER_03: In other words, it's not enough to keep going.
- 1:07:01SPEAKER_03: I'm going to play with that question.
- 1:07:04SPEAKER_03: That's a great laugh.
- 1:07:27Thank you.