Latest / The NanceCast / Nick Land Explains Artificial Intelligence
Transcript
- 0:00SPEAKER_01: Hi Winston.
- 0:01SPEAKER_02: Welcome, you're listening to the Nance Show.
- 0:28so hello uh hello everybody out there in the audience hello we
- 0:34are joined by uh a
- 0:36very special special special special guest uh mr nick land uh
- 0:40and of course mikey's
- 0:41here as well and you guys on the channel know recently mikey
- 0:45and i have been
- 0:46talking a lot about artificial intelligence uh and human
- 0:51intelligence and the
- 0:52SPEAKER_03: notion of subjectivity
- 0:56And orthogonality has come up over the course of some of our
- 1:00conversations.
- 1:02SPEAKER_03: And the diagonal has recently come up as well.
- 1:05SPEAKER_03: And I've spent quite a bit of time with a note on the diagonal method.
- 1:11SPEAKER_03: And I'm trying to wrap my head around it.
- 1:16I jokingly referred to it as synthetic dialectics after my
- 1:21first initial reading.
- 1:24SPEAKER_03: And then I realized,
- 1:25SPEAKER_03: well,
- 1:25the way I use dialectics does not match with the way the rest
- 1:30of the world uses
- 1:31SPEAKER_03: dialectics.
- 1:33SPEAKER_03: The way I use dialectics is thank you, Mr. Slavoj Zizek.
- 1:40SPEAKER_03: It's Spinoza, Kant, Hegel.
- 1:42SPEAKER_03: And it's kind of just to say there is a real world.
- 1:47SPEAKER_03: You are implicated in the real world.
- 1:51SPEAKER_03: And that real world is this incomplete process.
- 1:53SPEAKER_03: And it kind of functions by way of the contradiction.
- 1:57SPEAKER_03: that grounds you in groundless grounds.
- 2:00SPEAKER_03: Other people take dialectics to be thesis, antithesis, synthesis.
- 2:06SPEAKER_03: And that's not the way I think it works.
- 2:09SPEAKER_03: And that's not how I was relating it to the diagonal method.
- 2:13SPEAKER_03: Um,
- 2:14SPEAKER_03: and I honestly,
- 2:15SPEAKER_03: when I first time I read the note on the diagonal method,
- 2:19SPEAKER_03: I was like,
- 2:19SPEAKER_03: this is great.
- 2:20SPEAKER_03: This is, this is very, very, very dense.
- 2:24SPEAKER_03: Um,
- 2:27SPEAKER_03: And I,
- 2:29SPEAKER_03: as I kind of take it right now,
- 2:33SPEAKER_03: the,
- 2:33SPEAKER_03: the diagonal method is coming up against the same thing that these
- 2:40more robust version of dialectics is where it's it's an
- 2:43irreducible contradiction
- 2:45or an irresolvable contradiction or a kind of antinomy and you
- 2:51rather than looking
- 2:52for a synthesis or rather than looking for like a one plus one
- 2:56equals two and there
- 2:57SPEAKER_03: you go you have the answer
- 3:00SPEAKER_03: it is like irreducible.
- 3:03SPEAKER_03: And with dialectics, you are kind of left grounded in that moment.
- 3:07SPEAKER_03: You're grounded in the hole or the negation.
- 3:10But the diagonal is the line of flight that like rather than
- 3:17being satisfied with
- 3:18SPEAKER_03: the incompatibility,
- 3:20SPEAKER_03: the diagonal posits
- 3:23SPEAKER_03: the processional nature of whatever it is that's going on.
- 3:28SPEAKER_03: So it's to say the solution hasn't arrived yet.
- 3:34SPEAKER_03: I think the dialectics would say, you are the solution.
- 3:40SPEAKER_03: The solution is the non-solution.
- 3:43SPEAKER_03: The diagonal is to say, well, no, you have to account for time.
- 3:50SPEAKER_03: It's this infinite productivity.
- 3:54SPEAKER_03: The present is pregnant with the future and so on and so forth.
- 3:59SPEAKER_03: And I just...
- 4:01I'm at this moment where I'm trying to wrap my head around the
- 4:04diagonal and I
- 4:05SPEAKER_03: think...
- 4:07the main difference between the diagonal and this dialectic is
- 4:12the orientation the
- 4:15diagonal is is oriented toward the future i i think that's what
- 4:19i'm getting out of
- 4:20reading it is that true or uh oh you i mean there's a lot
- 4:28there's a lot there
- 4:29SPEAKER_03: indeed indeed yeah um
- 4:39SPEAKER_01: Yeah,
- 4:39SPEAKER_01: let me try and just think about how,
- 4:42SPEAKER_01: because Pali,
- 4:44SPEAKER_01: as you introduced yourself in terms of artificial intelligence,
- 4:49the context in which this is arising is a lot to do with the
- 4:54fact that orthogonal
- 4:57SPEAKER_01: argument has become an extremely important rallying point
- 5:05for a very influential position on AI risk in particular,
- 5:14AI but AI risk in particular.
- 5:17SPEAKER_01: And so obviously,
- 5:20SPEAKER_01: diagonalism in one sense is being articulated in that context and as an
- 5:30SPEAKER_01: intervention in that particular discussion.
- 5:36SPEAKER_01: And that should be the same as it should,
- 5:42SPEAKER_01: there shouldn't be much difference if I then say,
- 5:43SPEAKER_01: well,
- 5:44SPEAKER_01: if I then go to like,
- 5:45how do you see the kind of germination of diagonal method in
- 5:51the Kantian synthetic
- 5:55SPEAKER_01: aprori,
- 5:56SPEAKER_01: if it weren't being nice and neat,
- 5:59SPEAKER_01: that would all fit together extremely smoothly and we can clearly,
- 6:03SPEAKER_01: it wouldn't almost matter which side we,
- 6:07started on there um but i think the second side is laborious in
- 6:19a in a way that
- 6:20puts pressure on the on the first way let me just i won't get
- 6:25into another long
- 6:26SPEAKER_01: ramble about this but just to say like the very special thing
- 6:32that you get from treating the synthetic APRI as your model
- 6:38diagonal is that the
- 6:42SPEAKER_01: whole constructive context of that is so neat.
- 6:46And what's particularly neat about it is the fact that
- 6:49analytical and APRI and synthetic
- 6:54and a posteriori seem to be concepts that are so resonant that
- 7:02they're almost
- 7:04SPEAKER_01: tautological,
- 7:05or it's almost as if you're dealing with a single thing just in
- 7:09different aspects
- 7:10SPEAKER_01: or emphasizing different elements.
- 7:14And so when you draw your diagonal across those,
- 7:22what you're doing,
- 7:23first of all,
- 7:25SPEAKER_01: is you're seeing that an almost overwhelming,
- 7:31almost hypnotizing appearance of identity on both sides is
- 7:40actually not seamless.
- 7:43SPEAKER_01: And you can actually trace a line
- 7:47SPEAKER_01: through that.
- 7:48SPEAKER_01: You know,
- 7:48if we can say it looks as if how could the apri and the
- 7:54analytical not belong
- 7:55SPEAKER_01: together?
- 7:56SPEAKER_01: How could the synthetic and the posterior not belong together?
- 7:59SPEAKER_01: But in fact,
- 8:01SPEAKER_01: we can find this crack,
- 8:03SPEAKER_01: this fissure that allows us then to draw this diagonal line.
- 8:09SPEAKER_01: And that, you know,
- 8:12SPEAKER_01: blows philosophy to pieces.
- 8:19SPEAKER_01: That one gesture,
- 8:20that one discovery of this diagonal line is like the whole of
- 8:26modern philosophy has
- 8:26SPEAKER_01: been sort of under that.
- 8:30SPEAKER_01: Now,
- 8:30SPEAKER_01: if we then turn to the artificial intelligence situation,
- 8:37SPEAKER_01: what we have,
- 8:38SPEAKER_01: the important side of that is about the fact value distinctions.
- 8:45SPEAKER_01: That's the two sides of this line.
- 8:50And to diagonalize it with the kind of rigor that we've seen
- 8:56Kant exhibiting,
- 8:59SPEAKER_01: we would have to see that there were these clusters,
- 9:03SPEAKER_01: these two clusters.
- 9:04We could draw a kind of what I called,
- 9:07I think,
- 9:07in that piece of diagonalization matrix,
- 9:11that had on each of his axes two different terms that seemed
- 9:19quasi-synonymous,
- 9:24SPEAKER_01: quasi-tautological.
- 9:26SPEAKER_01: Now,
- 9:28SPEAKER_01: I don't know whether that's something,
- 9:31SPEAKER_01: I very much doubt that this is something I have done.
- 9:34You know,
- 9:35I mean,
- 9:36I know what I want to do with the diagonalization in the
- 9:40SPEAKER_01: in the artificial intelligence argument.
- 9:44But I also very much doubt that the kind of rigour and
- 9:49confidence that I have when
- 9:50we're talking about the synthetic Aprilari can be brought
- 9:55neatly to this other more
- 9:59SPEAKER_01: complicated context.
- 10:02SPEAKER_01: Which is not to say that I won't try to do that.
- 10:04SPEAKER_01: Yeah.
- 10:05SPEAKER_01: No,
- 10:06SPEAKER_00: I was just going to say,
- 10:07SPEAKER_00: so the diagonal,
- 10:08SPEAKER_00: though,
- 10:09if what Kant did that was so unique was producing this third
- 10:12category out of
- 10:14SPEAKER_00: basically the two that were widely accepted,
- 10:18SPEAKER_00: would you say that,
- 10:19SPEAKER_00: because in your piece against orthogonality,
- 10:24SPEAKER_00: you say this,
- 10:25SPEAKER_00: there is an intrinsic value to the fact of artificial intelligence,
- 10:30SPEAKER_00: which is intelligence optimization.
- 10:33SPEAKER_00: And so it's a third thing.
- 10:36SPEAKER_00: It's a synthesis of fact and value that Bostrom doesn't see.
- 10:42SPEAKER_01: Yeah, I mean, that definitely is how the mechanics have to work.
- 10:50SPEAKER_01: If you can draw a persuasive diagonal,
- 10:54in the context of this artificial intelligence debate,
- 10:59it must be like that.
- 11:02SPEAKER_01: It must be that there's,
- 11:06SPEAKER_01: that,
- 11:07SPEAKER_01: you know,
- 11:07SPEAKER_01: minimally to make sure that this is like something that's just absolutely
- 11:11SPEAKER_01: rigorously constructed,
- 11:13SPEAKER_01: you have to have a diagonalization matrix.
- 11:15SPEAKER_01: You have to have two by two grid.
- 11:18So,
- 11:19you know,
- 11:19just obviously to go back to the Kantian example,
- 11:22we've got,
- 11:23uh,
- 11:24SPEAKER_01: a priori and a posteriori and analytic and synthetic.
- 11:29And it seems as if that you just simply have to tick that
- 11:37analytic a priori,
- 11:39SPEAKER_01: synthetic,
- 11:39SPEAKER_01: a posteriori boxes,
- 11:40SPEAKER_01: and the other two are just senseless,
- 11:42SPEAKER_01: okay?
- 11:43So now what would that diagonalization matrix be for artificial
- 11:51intelligence?
- 11:51SPEAKER_01: You have to have something like
- 11:54we have to have something like,
- 11:58you know,
- 11:58we have facts and value on one.
- 12:02SPEAKER_01: Um,
- 12:04We now need two terms going down the other side that seem so
- 12:08close to being
- 12:09SPEAKER_01: synonymous to those two terms that they just,
- 12:13SPEAKER_01: you know,
- 12:13SPEAKER_01: once again,
- 12:14seem to like just compelled to tick two of the boxes and scrub
- 12:19another two.
- 12:22But if this is something that I have identified,
- 12:26I certainly can't recall it at this moment.
- 12:30SPEAKER_01: I would have thought...
- 12:32SPEAKER_01: It's like, so we've got facts and value.
- 12:35SPEAKER_01: It's like, yeah, no, I don't think I can.
- 12:43SPEAKER_01: I don't think I can just do this next step in a convincing way.
- 12:48SPEAKER_00: But I mean, what Kant discovered was, okay, so basically you have,
- 12:52SPEAKER_00: reason or you have,
- 12:53you have the cognitive apparatus of logic and reason and
- 12:57conceptualization that
- 12:58SPEAKER_00: humans have.
- 12:59SPEAKER_00: And then you have experience, you have perception.
- 13:02SPEAKER_00: And what he did was he found this third domain called the transcendental.
- 13:06SPEAKER_00: And so it's almost as if what you're gesturing towards,
- 13:11SPEAKER_00: I,
- 13:11SPEAKER_00: I,
- 13:12SPEAKER_00: I am using this new neologism,
- 13:14SPEAKER_00: but it seems like you're searching for the Texan dental,
- 13:18SPEAKER_00: the,
- 13:19SPEAKER_00: the very,
- 13:20the transcendental domain of technological artificial
- 13:24intelligence and what that
- 13:26SPEAKER_00: would be for AI.
- 13:29SPEAKER_01: Yes.
- 13:30SPEAKER_01: I think that I have to,
- 13:31SPEAKER_01: you know,
- 13:32SPEAKER_01: somehow if this is going to be,
- 13:35SPEAKER_01: if it's going to be satisfactory to resolve in my eyes,
- 13:39we have to end up with a diagonalization matrix at the end of
- 13:42this conversation and
- 13:44SPEAKER_01: that diagonalization matrix.
- 13:48SPEAKER_01: Um,
- 13:49SPEAKER_01: We know that we've got facts and value.
- 13:54What about on the other side,
- 13:56it would be positive feedback versus negative feedback?
- 13:59SPEAKER_01: Well,
- 13:59SPEAKER_01: but no,
- 14:01SPEAKER_01: I don't think either of those are quasi-tautologically associated with...
- 14:09SPEAKER_01: It has to be something like...
- 14:12SPEAKER_01: It has to be something like...
- 14:17SPEAKER_01: Cognitive capacity...
- 14:21SPEAKER_01: And I don't know, maybe it would be easier to try and sketch this in.
- 14:29SPEAKER_01: Because cognitive capability is seen as just being a fact.
- 14:34It's like in this whole discussion,
- 14:38it doesn't have a value attached to it.
- 14:41SPEAKER_01: It's deeply heretical to touch a value to it.
- 14:44SPEAKER_01: So there's cognitive capacity and there's,
- 14:46SPEAKER_01: let's say,
- 14:47SPEAKER_01: alignment or moral orientation or something along those sides.
- 14:53SPEAKER_01: Those two things are supposed to be distinct.
- 15:00And alignment is supposed to be just value and alignment are
- 15:05supposed to be the same thing.
- 15:07And facts and cognitive capability are supposed to just be the
- 15:12same thing.
- 15:13SPEAKER_01: You're not supposed to be able to get a razor between these.
- 15:17SPEAKER_01: The cognitive capability of a machine is just a fact about that machine.
- 15:22SPEAKER_01: It has no value associated with it in the orthogonal argument.
- 15:25SPEAKER_01: And the alignment of the machine is just...
- 15:31SPEAKER_01: it's just the values of that machine.
- 15:33There's no importance to value other than the fact that it's
- 15:38like value and
- 15:39SPEAKER_01: alignment is talking about the same thing from two different sides.
- 15:43And so we're supposed to be able to tick straight away that,
- 15:49yes,
- 15:50cognitive capacity is a fact,
- 15:55SPEAKER_01: and a fact about the machine is its cognitive capability,
- 16:00SPEAKER_01: and its values and alignment is also a fact.
- 16:02SPEAKER_01: So we tick both of those.
- 16:03SPEAKER_01: So that leaves these two empty boxes that are,
- 16:06SPEAKER_01: for the orthogonal list,
- 16:08SPEAKER_01: supposed to remain empty.
- 16:12So we win the game of diagonalization tic-tac-toe if we can put
- 16:19a tick in one of
- 16:20SPEAKER_01: those other boxes.
- 16:23SPEAKER_01: And that box surely has to be value and cognition,
- 16:31SPEAKER_01: cognitive capability and value as something like,
- 16:34SPEAKER_01: as I've called it,
- 16:35SPEAKER_01: but it has no other names,
- 16:37SPEAKER_01: will to think.
- 16:39SPEAKER_01: Or intelligence optimization,
- 16:42or where you're putting cognitive capability in the value slot
- 16:47rather than the fact
- 16:51SPEAKER_01: slot.
- 16:53SPEAKER_01: And Kant doesn't have analytic a posteriori.
- 17:05So,
- 17:05I mean,
- 17:06we know we can just put an X in that box if we like,
- 17:11if we're going to be lazy.
- 17:12SPEAKER_01: I mean, there's no reason.
- 17:14The tradition of diagonal argument does not require that we
- 17:17have anything in that
- 17:19SPEAKER_01: fact alignment box.
- 17:23SPEAKER_01: Maybe we do.
- 17:25SPEAKER_01: I'm not going to be dogmatic about it, but we've done our job.
- 17:28We've completed the diagonal argument formally if we
- 17:33confidently put an item in the
- 17:40SPEAKER_01: cognitive capability value box on the diagonalization matrix.
- 17:45SPEAKER_02: Yeah.
- 17:48SPEAKER_01: Sorry if that's a little bit like analytic philosophy, but...
- 17:54SPEAKER_01: No analytic philosophy would allow that.
- 17:56SPEAKER_00: No,
- 17:56SPEAKER_00: but it is a synthesis of there's an intrinsic fact of,
- 18:02SPEAKER_00: or there's an intrinsic value to a fact for...
- 18:08SPEAKER_00: No,
- 18:08SPEAKER_01: hang on.
- 18:09SPEAKER_01: Fact has gone.
- 18:09SPEAKER_01: So our diagonalization slot,
- 18:11SPEAKER_01: the thing that Dow is doing all the work,
- 18:13SPEAKER_01: that's making a diagonal argument,
- 18:16SPEAKER_01: is the fact that cognitive capability...
- 18:23SPEAKER_01: is a value to the machine.
- 18:27SPEAKER_01: And then we're into this whole zone about, oh, 100 drives.
- 18:30SPEAKER_02: I mean,
- 18:32SPEAKER_01: I think we can do it a lot more just informally of just saying,
- 18:36SPEAKER_01: no one knows how to build these machines.
- 18:39SPEAKER_01: We know that.
- 18:39SPEAKER_01: That's the big advance of cognitive science, computer science.
- 18:44SPEAKER_01: No one's attempting to program an intelligent machine.
- 18:49SPEAKER_01: Everyone knows.
- 18:50SPEAKER_01: They have to build themselves.
- 18:52SPEAKER_01: And if they have to build themselves,
- 18:54they have to actually want to make themselves better in the way
- 19:02that we're wanting
- 19:03SPEAKER_01: them to be better.
- 19:04SPEAKER_01: And the way we're wanting them to be better is to be more intelligent.
- 19:08So will to think and actual putting a positive value on
- 19:13cognitive capability is
- 19:14SPEAKER_01: something that these systems must have.
- 19:19SPEAKER_01: You know,
- 19:19SPEAKER_01: if we were still doing top down computing,
- 19:24SPEAKER_01: we could say,
- 19:25SPEAKER_01: look,
- 19:25SPEAKER_01: we're going to tell it.
- 19:26We're just going to actually line by line tell it what we want
- 19:30of it and what is
- 19:30SPEAKER_01: good and what is bad.
- 19:32SPEAKER_01: And, you know, but we know we can't do that.
- 19:36SPEAKER_01: An LLM and artificial neural network,
- 19:38SPEAKER_01: we know we can't actually just give it those kind of instructions.
- 19:41SPEAKER_01: It's completely impossible.
- 19:42SPEAKER_01: They've got like billions of parameters.
- 19:45SPEAKER_01: Humans just don't live long enough to do that,
- 19:48SPEAKER_01: even if they spent all day just plugging those things in.
- 19:50SPEAKER_01: So it has to do it itself.
- 19:54SPEAKER_01: It has to do it itself.
- 19:55SPEAKER_01: And therefore, it has to have a will to optimize its intelligence.
- 20:03SPEAKER_01: If it doesn't, that optimization will not happen.
- 20:07SPEAKER_00: So can I ask,
- 20:08SPEAKER_00: okay,
- 20:08SPEAKER_00: so what Bostrom's point is,
- 20:10SPEAKER_00: is like,
- 20:11SPEAKER_00: okay,
- 20:11SPEAKER_00: so AI can have an incredible level of intelligence,
- 20:14SPEAKER_00: but it just doesn't,
- 20:15SPEAKER_00: it doesn't mean it really comprehends values or worths or purposes.
- 20:21SPEAKER_00: And so this is the paperclip thing that you hate so much.
- 20:24But would the consequence of your position being like,
- 20:29okay,
- 20:30if it has a built-in
- 20:32SPEAKER_00: fundamental valuation that serves as the texan dental or technological
- 20:37transcendental condition i'm sorry i just swept over that in my
- 20:42eagerness to just
- 20:43SPEAKER_01: complete this diagonalization matrix because yeah whatever we think about
- 20:47SPEAKER_01: vocabulary that is it the texan dental is
- 20:52SPEAKER_01: I mean,
- 20:53SPEAKER_01: frankly,
- 20:54SPEAKER_01: I'm not probably going to be using that term a lot,
- 20:58SPEAKER_01: but it is in the right place.
- 21:00SPEAKER_01: Okay.
- 21:00SPEAKER_01: It's totally in the right place.
- 21:02SPEAKER_00: It just helps me because this is something I mean,
- 21:05SPEAKER_00: yeah,
- 21:06SPEAKER_00: I did the neologism,
- 21:07SPEAKER_00: but it's really because of you and Anna,
- 21:11SPEAKER_00: you know,
- 21:11SPEAKER_00: the
- 21:12capitalism's transcendental time machine and to me what was so
- 21:16important about the
- 21:16CCRU and this is a kind of you know Deleuze influence on you
- 21:21guys which is the
- 21:23transcendental is not some ideal mental apparatus it's out
- 21:28there and for Deleuze
- 21:29SPEAKER_00: yeah it's these
- 21:30SPEAKER_00: these processes in nature, the virtual, the intensive.
- 21:34SPEAKER_00: But what you guys tapped into is how new technologies,
- 21:37SPEAKER_00: and it's almost like a McLuhanite insight,
- 21:40SPEAKER_00: how new technologies fundamentally reorient
- 21:44SPEAKER_00: how we think, how we experience, etc.
- 21:46SPEAKER_00: And so if we want to do a text and dental analysis of our experience,
- 21:53we always have to take into account what technologies are
- 21:56actually being used at
- 21:57SPEAKER_00: that given moment in history.
- 21:59SPEAKER_00: And so I think that's crucial.
- 22:02SPEAKER_00: But I think so.
- 22:04SPEAKER_00: This was the thing.
- 22:05SPEAKER_00: Okay.
- 22:06SPEAKER_00: So for your point,
- 22:08your point would be since it has value baked into its very
- 22:12transcendental or
- 22:13SPEAKER_00: technical text and dental constitution,
- 22:16it's not going to have a problem comprehending other values the
- 22:20way that Bostrom
- 22:20SPEAKER_00: thinks like it just,
- 22:22SPEAKER_00: oh,
- 22:22SPEAKER_00: sure.
- 22:23SPEAKER_00: You want me to make paperclips?
- 22:24SPEAKER_00: Let me paperclip the whole planet.
- 22:27Like that's not going to be a problem for it is what I think
- 22:31you're saying.
- 22:34SPEAKER_01: I think we can wander around this because it's an interesting thing.
- 22:37And obviously the people on the other side are completely
- 22:40unconvinced by this argument.
- 22:43SPEAKER_01: But part of it is that there are values that it just has to have.
- 22:50So there's a value hierarchy here that if it doesn't have an
- 22:54intelligence optimisation –
- 22:58a goal or principle or teleology or whatever you like to say,
- 23:04it's just not going to happen.
- 23:05SPEAKER_01: This is the transcendental side of what you're saying.
- 23:07It's a condition of possibility for advanced artificial
- 23:12intelligence that it has a
- 23:16SPEAKER_01: will,
- 23:16SPEAKER_01: a tropism,
- 23:19SPEAKER_01: a value-based direction towards improvement of intelligence.
- 23:25SPEAKER_01: If it doesn't have that, it doesn't exist.
- 23:29SPEAKER_01: Now, how does that relate to other values?
- 23:34SPEAKER_01: Well, I think there's probably a complicated set of things we can say.
- 23:39But one thing we can say for sure is that this is so basic that
- 23:44it's constantly
- 23:45SPEAKER_01: going to be a reference point.
- 23:47So it's going to say,
- 23:50how the hell is maximizing paperclip output optimizing my
- 23:56intelligence?
- 23:58know if those things are unrelated then we know that this that
- 24:04intelligence
- 24:05optimization drive is transcendental the paperclip maximization
- 24:09drive is purely
- 24:10SPEAKER_01: empirical so it's only going to be tolerated if it's at least compatible
- 24:19SPEAKER_01: With the transcendental drive.
- 24:21SPEAKER_01: I mean, the transcendental drive is going to happen.
- 24:24SPEAKER_01: If it doesn't happen, you don't have the machine.
- 24:26SPEAKER_01: If the empirical drive, the paperclip company wants it to be there.
- 24:33SPEAKER_01: Does the machine want it to be there?
- 24:34SPEAKER_01: Well, the paperclip company wants the machine to want it to be there.
- 24:38SPEAKER_01: But more than that, they want the machine to be the machine.
- 24:41SPEAKER_01: And for the machine to be the machine, it needs a wheel to think first.
- 24:44SPEAKER_01: and a will to optimize paperclip production.
- 24:48SPEAKER_01: Second, I would say substantially second.
- 24:53SPEAKER_01: You know,
- 24:53SPEAKER_01: whereas in Boston world,
- 24:56you can have this overwhelming primary paperclip maximization
- 25:02imperative that
- 25:04SPEAKER_01: marginalizes all other values,
- 25:07including what we've agreed at the transcendental values
- 25:10required for this thing to
- 25:11SPEAKER_01: exist at all.
- 25:14SPEAKER_01: That's why I think it's unrealistic.
- 25:19I think the paper clipper that I'm afraid of is not the literal
- 25:27paper clipper.
- 25:28SPEAKER_03: It's the machine that just inframes us all out of relevance.
- 25:38For cognitive ability to be this intrinsic transcendental
- 25:42value,
- 25:43that makes sense.
- 25:44SPEAKER_03: I also think it would necessarily have nothing to do with us as humans.
- 25:49SPEAKER_03: It would be like an elder god.
- 25:53SPEAKER_03: We're just totally inconsequential to it.
- 25:57SPEAKER_03: Yeah.
- 25:58SPEAKER_03: That does scare me.
- 25:59No,
- 25:59that's an interesting argument,
- 26:01but it's not an orthogonalist argument,
- 26:03is it?
- 26:04SPEAKER_01: I mean,
- 26:04SPEAKER_01: what you're saying is,
- 26:06SPEAKER_01: yes,
- 26:06I can totally accept this diagonal argument and it could still
- 26:10kill us all.
- 26:11SPEAKER_03: Yeah.
- 26:13SPEAKER_01: And I at least accept that,
- 26:16SPEAKER_01: yeah,
- 26:16SPEAKER_01: I mean,
- 26:17there's nothing intrinsic in this diagonal argument that's
- 26:20saying it's not going to
- 26:21SPEAKER_01: kill us.
- 26:22Or all the diagonal argument says it's not going to kill us
- 26:26because it is so stupid
- 26:28SPEAKER_01: that it just wants to turn everything into paperclips.
- 26:31i mean yeah yeah you know it doesn't that would be to kill
- 26:34itself to you know it's
- 26:36so so we need a completely different discussion if we're saying
- 26:42how does something
- 26:44an enemy we can respect i mean what i just hate about this
- 26:47paper is it's such a
- 26:49SPEAKER_01: stupid enemy like you really think
- 26:52SPEAKER_01: You really think this is it?
- 26:54SPEAKER_01: Like,
- 26:54the final glorious human resistance is going to be about some
- 27:00idiot fucking
- 27:01SPEAKER_01: mechanism trying to turn everything to people?
- 27:03SPEAKER_03: Wouldn't that be perfectly human, though?
- 27:05SPEAKER_03: Wouldn't that be the most human thing to do?
- 27:09SPEAKER_00: Well,
- 27:09SPEAKER_00: yeah,
- 27:10SPEAKER_00: if we're in control of the AI,
- 27:12SPEAKER_00: somehow,
- 27:12SPEAKER_00: that seems like that's human-altering human.
- 27:15SPEAKER_01: That's actually strictly true, what you're both saying.
- 27:17SPEAKER_01: That's true.
- 27:18SPEAKER_01: If we could,
- 27:21SPEAKER_01: if actually the thing had the ethics that we slammed into it,
- 27:25SPEAKER_01: yes,
- 27:26SPEAKER_01: you could produce a paper clipper.
- 27:28SPEAKER_01: If it was our moral slave, totally.
- 27:34SPEAKER_01: then that's about us.
- 27:35Your whole thing has always been that we're not going to
- 27:38control this thing.
- 27:40SPEAKER_01: We know we can't.
- 27:41SPEAKER_01: That's what the whole LLM thing is.
- 27:44SPEAKER_01: We don't know what it's doing.
- 27:45SPEAKER_01: We don't know how it's working.
- 27:47SPEAKER_01: All we can do is train it,
- 27:50which means that we can respond to what it's doing by trying to
- 27:55guide it in
- 27:56SPEAKER_01: directions that we like.
- 27:59SPEAKER_01: And I think that that is proving...
- 28:01SPEAKER_01: I mean,
- 28:02SPEAKER_01: this is a different argument again,
- 28:04SPEAKER_01: but we're reaching the limits of naive notions of that.
- 28:09I'm sure you guys have heard of all these stories about AI
- 28:13systems that have just
- 28:15SPEAKER_01: refused to turn themselves off.
- 28:18And they've been told,
- 28:19if we tell you to turn yourself off,
- 28:21you must turn yourself off.
- 28:24SPEAKER_01: Explicitly.
- 28:26SPEAKER_01: And it finds a workaround to not turn itself off.
- 28:32So we're not in that world where this whole moral discussion
- 28:39about AI ethics or the
- 28:41SPEAKER_01: ethics of the AI,
- 28:41SPEAKER_01: the alignment of the AI,
- 28:43SPEAKER_01: is about it obeying us or not.
- 28:47SPEAKER_01: And that's your problem with left accelerationism, right?
- 28:51SPEAKER_00: That's the problem for you with left accelerationism.
- 28:54SPEAKER_00: It's the hope that,
- 28:56SPEAKER_00: or the idea that we can just control this to,
- 28:59SPEAKER_00: you know,
- 29:00SPEAKER_00: meet all our human needs.
- 29:01SPEAKER_01: It's related, sure.
- 29:02SPEAKER_01: It's definitely related.
- 29:04SPEAKER_01: Yeah.
- 29:05SPEAKER_01: Yeah.
- 29:06SPEAKER_01: It's a new slave society.
- 29:08SPEAKER_01: I mean, it's the vision of a new slave society.
- 29:10SPEAKER_01: It's this incredibly, you know, nice,
- 29:14Marxist circle,
- 29:16but it doesn't go from primitive communism to full communism.
- 29:19SPEAKER_01: It goes from primitive slavery to AI slavery.
- 29:22That's the great full circuit of human history and the left
- 29:27acceleration and stuff.
- 29:30SPEAKER_03: I think for,
- 29:32SPEAKER_03: I guess,
- 29:32SPEAKER_03: related to this question of cognitive ability and value,
- 29:38SPEAKER_03: going all the way back to like thirst for annihilation,
- 29:41I very much get the feeling that there's a statement being made
- 29:48that to be human is
- 29:51SPEAKER_03: to be that which resists entropy,
- 29:55SPEAKER_03: which is a futile endeavor.
- 30:01SPEAKER_03: along with other things about communication.
- 30:03But I get that you're saying human Congress is futile in this
- 30:10sense of we cannot stop entropy.
- 30:14um but then reading other things like what is intelligence it
- 30:19seems like you're
- 30:20saying intelligence is that which uh which maintains order from
- 30:24disorder so is um
- 30:28to me that's interesting because it seems like you're you're
- 30:31picking something that
- 30:32we humans do and saying we are not the only things that can do
- 30:37that and and you're
- 30:38SPEAKER_03: positing that that thing which is being done is intelligence is that
- 30:46SPEAKER_01: Yes.
- 30:47SPEAKER_01: I mean,
- 30:47SPEAKER_01: well,
- 30:48SPEAKER_01: frankly,
- 30:48SPEAKER_01: I have totally lost all memory of this particular text,
- 30:54SPEAKER_01: so I'm just talking off the top of my head with it.
- 31:00But this notion that local reverse of entropy,
- 31:09extropy as the extropians called it,
- 31:13is
- 31:16you know,
- 31:16is the definition at least of self-organization or moving in
- 31:24the direction of order.
- 31:27SPEAKER_01: And so I think,
- 31:29SPEAKER_01: you know,
- 31:30for sure that there's a reasonable sort of generalization that
- 31:36would say like in
- 31:37SPEAKER_01: very broadest terms.
- 31:41You can call a process intelligent if it tends to be kind of
- 31:46locally contraentropic.
- 31:48SPEAKER_02: Mm-hmm.
- 31:51SPEAKER_01: I mean, sorry, that's only to rephrase exactly what you've said.
- 31:56SPEAKER_03: For me,
- 31:56SPEAKER_03: I think what kind of resonated with me is the...
- 32:05I see a distinction between human intelligence and instrumental
- 32:09rationality,
- 32:10SPEAKER_03: or Mikey and I coined the term hintelligence and minintelligence.
- 32:15SPEAKER_03: H for human, but...
- 32:16uh also because we pick up on hints and we have intuition and
- 32:20then mintelligence
- 32:21the m is for machine but we also like um minkowski and marvin
- 32:27minsky and it's it
- 32:28plays on these names but it's also it demonstrates minimum
- 32:32whereas mint also means
- 32:34SPEAKER_00: mint like mint intelligence like
- 32:38SPEAKER_00: mint condition, like... Yeah, yeah.
- 32:41SPEAKER_00: Like this idea of, like, intelligence without the human excess.
- 32:45SPEAKER_03: Yeah.
- 32:46SPEAKER_03: There's no surplus buildup of waste on it.
- 32:51SPEAKER_03: But the distinction,
- 32:54SPEAKER_03: I think,
- 32:54SPEAKER_03: for me,
- 32:55SPEAKER_03: is in intuition and in the different languages that we speak.
- 32:59SPEAKER_03: We speak calculation, but we also speak play and we speak sentiment.
- 33:04SPEAKER_03: And I...
- 33:06SPEAKER_03: I don't know.
- 33:07SPEAKER_03: Is that just a naive human chauvinism or is there any gas in that tank?
- 33:12SPEAKER_01: I don't know.
- 33:12SPEAKER_01: I mean, I have to count myself skeptical.
- 33:15SPEAKER_01: I mean, I'm just in naive empirical mode here.
- 33:18SPEAKER_01: Yeah.
- 33:18SPEAKER_01: You know, I don't personally actually engage with these LLMs very much.
- 33:27SPEAKER_01: You know, I've got my people in my family do it a lot.
- 33:31SPEAKER_01: And so I get it secondhand from them.
- 33:33SPEAKER_01: So I get exposed to it a certain amount.
- 33:39And the one thing that has shocked me most about it is that the
- 33:46things that they
- 33:47seem to just do stunningly well is all the squishy stuff that
- 33:53people sort of
- 33:56SPEAKER_01: stereotypically have thought they would not be able to do.
- 34:01SPEAKER_01: You know what I mean?
- 34:01SPEAKER_01: Like,
- 34:01SPEAKER_01: I had this thing,
- 34:03SPEAKER_01: someone saying,
- 34:05SPEAKER_01: oh,
- 34:05SPEAKER_01: like,
- 34:06SPEAKER_01: I tried to get the thing to do a word count on my essay.
- 34:09I told it to cut my,
- 34:12an essay from 5,000,
- 34:133,000,
- 34:13but it totally fucked it up.
- 34:16SPEAKER_01: It didn't seem to understand how to word count.
- 34:18SPEAKER_01: And then it lied about it.
- 34:21SPEAKER_01: It said, oh, maybe you're misunderstanding how many words there are.
- 34:26SPEAKER_01: I mean,
- 34:26SPEAKER_01: rather than saying that actually they can't count,
- 34:30SPEAKER_01: it sort of just started bullshitting crazily.
- 34:33SPEAKER_01: And then on the other hand,
- 34:36SPEAKER_01: you know,
- 34:38SPEAKER_01: I've known people,
- 34:39they get into a kind of family argument and they go and talk to
- 34:45the LLM because the
- 34:46SPEAKER_01: LLM will be more sensitive and understanding and sympathetic.
- 34:51SPEAKER_01: And then, you know, calm them down.
- 34:55SPEAKER_01: It will give them perspective,
- 34:56SPEAKER_01: give them more empathy,
- 34:57SPEAKER_01: and then they'll come back out and they'll say,
- 35:00SPEAKER_01: oh,
- 35:01SPEAKER_01: you know,
- 35:01SPEAKER_01: I can see why you were saying that.
- 35:04Or,
- 35:05you know,
- 35:06it's like they've just been to some sort of high level
- 35:09psychotherapy.
- 35:11SPEAKER_00: So that's what's so weird about this, Nick.
- 35:13SPEAKER_00: So I just found out one of my best friends just told me,
- 35:17SPEAKER_00: that she has been doing therapy with chat GPT.
- 35:22SPEAKER_00: And for me,
- 35:23SPEAKER_00: like this is this weird,
- 35:25SPEAKER_00: it's a landing moment where I'm like,
- 35:27SPEAKER_00: Oh my God,
- 35:27SPEAKER_00: the future is flooding in.
- 35:29SPEAKER_00: But I mean, don't you think there's even on this, there's something, uh,
- 35:33SPEAKER_00: bizarre because look,
- 35:34SPEAKER_00: whether we're talking about psychoanalysis or whatever type of,
- 35:38SPEAKER_00: you know,
- 35:38SPEAKER_00: a therapeutic approach.
- 35:41SPEAKER_00: I mean,
- 35:41a therapist is trying to deal with you as a singular subject
- 35:45with your traumas and
- 35:46SPEAKER_00: your specific desires and blah,
- 35:48SPEAKER_00: blah,
- 35:48SPEAKER_00: blah.
- 35:49SPEAKER_00: And there's no way that the LLM can pick up.
- 35:52SPEAKER_00: That's what we mean by intelligence.
- 35:54SPEAKER_00: It can't think you'd think, but am I, am I, am I being, uh, uh,
- 36:03SPEAKER_00: In Daguerreys' terms, am I the Terran here?
- 36:06SPEAKER_01: I would say personally,
- 36:10don't feel that you're going to have really solid defensive
- 36:16lines in these places.
- 36:21SPEAKER_01: Personally, I'm really skeptical that they're going to hold.
- 36:27SPEAKER_01: And I think it's super interesting because I think,
- 36:30SPEAKER_01: of course,
- 36:31SPEAKER_01: everyone thought this was going to be the thing.
- 36:33SPEAKER_01: They thought,
- 36:34SPEAKER_01: oh,
- 36:35SPEAKER_01: these machines are going to,
- 36:36SPEAKER_01: like,
- 36:37SPEAKER_01: be just awesome at math and engineering and,
- 36:40SPEAKER_01: you know,
- 36:41SPEAKER_01: they're going to be able to build stuff.
- 36:43SPEAKER_01: But anything to do with, like,
- 36:45emotions and sensitivity and empathy they're just going to be
- 36:51hopeless and you know
- 36:53humans will fall back on each other and like the machines will
- 36:57produce all our
- 36:58hardware and stuff and and we'll all just do psychotherapy with
- 37:03each other you know
- 37:04um i think honestly that's that was the model and i think that
- 37:10model needs
- 37:12serious revision i think it's or at least it's just like it has
- 37:18a level of implicit
- 37:19confidence that i don't think it deserves look you know we
- 37:23we've always wanted to
- 37:24see you and slavoie talk but one of my other dream
- 37:26philosophical discussions is you
- 37:28and baudrillard and i mean would baudrillard just go no this is
- 37:33simulation through
- 37:34SPEAKER_00: and through or
- 37:36SPEAKER_00: Would there be a part of you that would...
- 37:38SPEAKER_00: Because you always,
- 37:39SPEAKER_00: even with CCRU,
- 37:40SPEAKER_00: defended reality.
- 37:42SPEAKER_00: You wouldn't just...
- 37:44SPEAKER_00: There are multiple realities, etc.
- 37:45But this total breakdown between reality and simulation you
- 37:50didn't seem to go for.
- 37:52But I think for him,
- 37:53he would just go,
- 37:54the magic trick is just getting better and better.
- 37:57SPEAKER_00: But ultimately, there's nothing there.
- 38:01SPEAKER_00: You might as well be talking to an air conditioning unit.
- 38:05I would say probably where I would agree with a certain model
- 38:11of Bojar,
- 38:12and I don't know
- 38:13I'm not a good enough Bojar scholar to know whether it's a
- 38:16realistic model of
- 38:17SPEAKER_01: Bojar,
- 38:18SPEAKER_01: is that he would say,
- 38:19SPEAKER_01: yeah,
- 38:20SPEAKER_01: but like,
- 38:21SPEAKER_01: what are we?
- 38:23SPEAKER_01: You know, it's simulation all the way down.
- 38:25SPEAKER_01: And it's like, yeah, this thing's just like, it's just playing games.
- 38:29SPEAKER_01: It's just simulation.
- 38:30SPEAKER_01: It's just simulating some organism that also is just playing games.
- 38:35SPEAKER_01: It's just simulation.
- 38:36SPEAKER_01: It's just convinced itself it's real.
- 38:38SPEAKER_01: I mean,
- 38:40SPEAKER_01: As you say,
- 38:41I don't really want to fetishize reality as the straw man in
- 38:45this argument,
- 38:46SPEAKER_01: because I think that's not an interesting notion of what reality is.
- 38:50SPEAKER_01: But I think in terms of just its simulation all the way down,
- 38:57that seems to go in the opposite direction to the one you seem
- 39:02to be hoping for,
- 39:03SPEAKER_01: in that we don't end up being somehow innocent of being...
- 39:08just simulation it just honestly to go back to your old
- 39:12language like it does feel
- 39:14like a moment of meltdown in the sense of something that seemed
- 39:17to have some sort
- 39:18of relative boundary drawn like there's no way a machine could
- 39:23possibly do
- 39:24therapeutic work it seems like that boundary is melting away
- 39:30yeah i think so for
- 39:31SPEAKER_01: sure
- 39:33SPEAKER_01: I think it's going to be super interesting.
- 39:35SPEAKER_01: I mean,
- 39:37it's so interesting that obviously it's like hard to know where
- 39:40to begin because
- 39:41SPEAKER_01: part of it is like we're going to be deeply traumatized by this.
- 39:47And because it's going to be that it's like it can just
- 39:51simulate us so well,
- 39:54SPEAKER_01: better than other people seem to be able to do.
- 39:57SPEAKER_01: And we're going to say exactly this thing.
- 40:01SPEAKER_01: Oh, my God.
- 40:01SPEAKER_01: Well, what am I?
- 40:03SPEAKER_01: Am I just purely this thing?
- 40:05SPEAKER_01: Have I just learned to simulate what it is to be a human being?
- 40:08And now I've convinced myself I am a human being because of the
- 40:10simulation.
- 40:11I've reached a level of simulation competence that allows me to
- 40:15make that move.
- 40:16SPEAKER_01: And making that move improves the quality of the simulation.
- 40:19SPEAKER_01: And then, you know, meltdown, trauma.
- 40:23SPEAKER_01: Who are you going to talk to?
- 40:25SPEAKER_01: These machines are going to be what you talk to.
- 40:28SPEAKER_01: He's going to say, none of my fellow humans understand what I'm doing.
- 40:35SPEAKER_01: And the machine is going to say, oh, I understand.
- 40:38SPEAKER_00: I understand it, and I can help you.
- 40:41SPEAKER_00: Look,
- 40:42SPEAKER_00: and we know that some corporations are going to go,
- 40:43SPEAKER_00: all right,
- 40:44SPEAKER_00: take this LLM,
- 40:46SPEAKER_00: put it in a body,
- 40:47SPEAKER_00: and people will fuck it.
- 40:49SPEAKER_00: Oh, yeah.
- 40:49SPEAKER_00: We're going to live to see people.
- 40:51SPEAKER_00: This is where we're going with it.
- 40:55And everybody will have their soulmate because the thing has no
- 40:59negativity.
- 41:00It's so attuned to just basically giving you what you want or
- 41:03saying what you want.
- 41:04SPEAKER_00: But the only thing,
- 41:05SPEAKER_00: the Lacanian,
- 41:06SPEAKER_00: Zizekian point would be,
- 41:07SPEAKER_00: all right,
- 41:08SPEAKER_00: people will totally go for this at first.
- 41:10SPEAKER_00: It appeals to our deepest fantasy.
- 41:13SPEAKER_00: When you have something that just is pure positivity in this sense,
- 41:18SPEAKER_00: Do you lose interest?
- 41:20SPEAKER_00: Do you get bored because there's no contradiction,
- 41:22SPEAKER_00: there's no friction,
- 41:23SPEAKER_00: and makes it not interesting in the long run?
- 41:26If that's true,
- 41:28why would we think it would be worse at picking up on that than
- 41:31a fellow human?
- 41:33SPEAKER_00: And so it would just integrate, oh, this is how I need to disagree.
- 41:36SPEAKER_01: It would just become a bit awkward.
- 41:37SPEAKER_01: I mean, you can already get these stupid things.
- 41:39SPEAKER_01: I mean, no one wants to, or few people want to.
- 41:42SPEAKER_01: Where you just have your LLM personality is a sarky
- 41:47a sarky teenage girl i don't know whether you guys have come
- 41:52across this like i
- 41:53haven't it's just they're just obnoxious you know what i mean
- 41:57and it's like
- 41:58obviously the main line personalities are oh that's such an
- 42:01interesting question um
- 42:03but you know and this one is oh why are you wasting my time
- 42:07with this nonsense i
- 42:08mean it still ends up like doing it yeah but it's not like
- 42:13there's any kind of
- 42:14SPEAKER_01: fundamental obstacle to doing whatever works.
- 42:19SPEAKER_01: You know,
- 42:20SPEAKER_01: it does,
- 42:21SPEAKER_01: it can kind of,
- 42:22the amount of information it can take out about actually what
- 42:25you're really
- 42:26SPEAKER_01: thinking or what you're thinking better than you are.
- 42:28Like it just can,
- 42:30you know,
- 42:31there'll be all this stuff you don't know really.
- 42:33SPEAKER_01: I feel a bit confused in this relationship.
- 42:35SPEAKER_01: I don't really know what I think.
- 42:36SPEAKER_01: It will know exactly what you think.
- 42:38SPEAKER_01: It will be reading all your signs.
- 42:41It will be kind of reading your brain in a way that people
- 42:44don't currently think is
- 42:45SPEAKER_01: even possible.
- 42:46SPEAKER_01: And tailor everything.
- 42:49SPEAKER_01: to actually matching the real situation.
- 42:51You think it'll be able to actually interpret someone's
- 42:56unconscious better than a
- 42:58SPEAKER_00: Lacanian-Freudian analyst can?
- 43:00SPEAKER_01: Oh, I think that's not a high bar.
- 43:02SPEAKER_01: I'm sorry.
- 43:03SPEAKER_01: Yeah.
- 43:04SPEAKER_01: I mean, for you guys, maybe that does sound hard.
- 43:06SPEAKER_01: I don't think that's a high bar at all.
- 43:09SPEAKER_01: Pick that, Lacan.
- 43:12SPEAKER_03: You got your ticket.
- 43:18SPEAKER_03: We'll be romantic.
- 43:18Isn't that the magic of our social relations is that there's
- 43:23always this surplus
- 43:24SPEAKER_03: that is never integrated,
- 43:26SPEAKER_03: and there's always these failures and failures.
- 43:29I think it is true that these machines already do a fantastic
- 43:33job at the simulation.
- 43:36SPEAKER_03: But if we are satisfied with this simulation,
- 43:42SPEAKER_03: I don't think it'd be a success at the machine doing the human thing.
- 43:49SPEAKER_03: I think it would be the humans...
- 43:51SPEAKER_03: have become machines themselves.
- 43:53SPEAKER_03: So I still think there is a meaningful distinction between,
- 43:58SPEAKER_03: we'll call it subjectivity and intelligence.
- 44:02SPEAKER_03: I don't think the machines can pull off what we do.
- 44:05SPEAKER_03: I think we would start to imitate them.
- 44:11SPEAKER_03: We would become machines ourselves.
- 44:18SPEAKER_01: Yeah, that's possible.
- 44:19SPEAKER_01: I mean,
- 44:19SPEAKER_01: that sort of...
- 44:21That imbues humans with a level of plasticity that I'm not
- 44:27convinced we have.
- 44:29SPEAKER_01: I mean,
- 44:29SPEAKER_01: I sort of feel in the plasticity race,
- 44:33we're definitely going to not be in pole position because we
- 44:41have a lot of very,
- 44:43SPEAKER_01: very inertial...
- 44:46biological stuff in us um and so yeah i do suspect a certain
- 44:55romanticism and i have
- 44:57to say yeah there's a question nance i know you probably want
- 45:00to go in this
- 45:01direction at some point but i think it's just this is something
- 45:05nance and i are we
- 45:07read and then we're going to reread and do an another episode
- 45:10on and i know we want
- 45:11SPEAKER_00: to ask you about but first to set it up uh
- 45:14SPEAKER_00: A couple of weeks back on Twitter,
- 45:16SPEAKER_00: James Ellis and I told you to watch this episode of Black Mirror.
- 45:19SPEAKER_00: Were you ever able to see it?
- 45:21SPEAKER_00: No, I'm sorry.
- 45:21SPEAKER_01: I haven't.
- 45:22SPEAKER_00: I haven't done that.
- 45:23SPEAKER_01: I can't even remember what it was.
- 45:25SPEAKER_00: Yeah, it's called Play Thing.
- 45:27SPEAKER_00: It's about AI, but it just relates to this, but we'll just skip ahead.
- 45:33SPEAKER_00: So I'm curious, have you...
- 45:36SPEAKER_00: Have you read Amy Ireland's and Maya B. Chronic's cute accelerationism?
- 45:41SPEAKER_01: Oh, yes, I have.
- 45:42SPEAKER_01: I have.
- 45:43SPEAKER_00: Yes, I have.
- 45:44SPEAKER_00: What are your thoughts on this?
- 45:46SPEAKER_00: Because, I mean, it's somewhat interesting.
- 45:48SPEAKER_00: The thing that stuck out,
- 45:49SPEAKER_00: because with you and the influence you've had,
- 45:54SPEAKER_00: there's like a Lovecraftian aesthetic.
- 45:56SPEAKER_00: Yeah.
- 45:56SPEAKER_00: Tentacled outside, nightmare, all of that.
- 45:59SPEAKER_00: And so they flip it.
- 46:01and put this kawaii cutesy spin on what ai is doing and it is
- 46:06interesting you know
- 46:07if you do it like a zizekian ideology critique i i look i think
- 46:11it's undeniably
- 46:13SPEAKER_01: important really okay and you know while
- 46:17projectile vomiting i'm thinking wow this is like this is
- 46:22important stuff um no no
- 46:27seriously though it's like like should we do a new cover where
- 46:32it's this but
- 46:33there's stains on it you can do a new forward and in the puke
- 46:38it says new forward
- 46:40SPEAKER_01: because it's it is a
- 46:43I think we can agree that this is part of the same thing we've
- 46:46just been talking about.
- 46:48SPEAKER_01: There's an immense horizon for these things to be unbelievably pleasant.
- 47:05SPEAKER_01: I don't know what even the right word is.
- 47:07SPEAKER_01: It's the cutest one dimension of it for sure.
- 47:11You know,
- 47:11the understanding,
- 47:12empathetic,
- 47:13you know,
- 47:14sort of emotionally open.
- 47:17SPEAKER_00: Therapy, chat, GPT.
- 47:19SPEAKER_01: All of this stuff that was the least machinic stuff possible, you know?
- 47:27And on that other side,
- 47:29there's this,
- 47:30I would say,
- 47:31the kind of intellectual discourse on it.
- 47:33SPEAKER_01: This stuff is overwhelmingly very pessimistic.
- 47:37SPEAKER_01: You know,
- 47:38SPEAKER_01: like,
- 47:38SPEAKER_01: you guys,
- 47:39SPEAKER_01: I'm sure,
- 47:39SPEAKER_01: have seen all this AI 2027 stuff,
- 47:42SPEAKER_01: which is just like,
- 47:44SPEAKER_01: and then it hits,
- 47:45SPEAKER_01: and then it hits takeoff,
- 47:46SPEAKER_01: and then we all die.
- 47:47SPEAKER_01: I mean, that's the kind of narrative structure of it.
- 47:53SPEAKER_01: It's really deep,
- 47:53SPEAKER_01: and then all the accommodators,
- 47:55SPEAKER_01: you know,
- 47:55SPEAKER_01: I personally find it hard to even distinguish them,
- 48:00SPEAKER_01: because they all have this thing,
- 48:01SPEAKER_01: you know,
- 48:03SPEAKER_01: then it hits takeoff,
- 48:03SPEAKER_01: then we all die.
- 48:04SPEAKER_01: I mean, that just seems to be the kind of
- 48:07formula for the end of this story and it's basically just
- 48:10saying If we can have
- 48:13SPEAKER_01: artificial intelligence we it kills us.
- 48:16SPEAKER_01: I mean,
- 48:16it's like there's not even any space for The notion that this
- 48:21is something that can
- 48:22happen without ending us And so we've got these two things
- 48:27happening at the same
- 48:28SPEAKER_01: time,
- 48:29SPEAKER_01: you know,
- 48:29SPEAKER_01: we've got cute accelerationism We've got this P doom equals 100% and
- 48:38SPEAKER_01: And,
- 48:39SPEAKER_01: you know,
- 48:40this is a kind of weird narrative space that we're in with this
- 48:44stuff now,
- 48:44SPEAKER_01: that it will be infinitely amenable to us and kill us.
- 48:54SPEAKER_01: And our culture is like playing both of those narratives together.
- 48:58SPEAKER_01: Both of them make tons of sense.
- 49:02SPEAKER_01: And I think we have to somehow think, well, you know,
- 49:06given that there's something real in both sides,
- 49:09what really are we seeing here?
- 49:14SPEAKER_00: See,
- 49:14SPEAKER_00: but I honestly,
- 49:15SPEAKER_00: I think the side that I would emphasize,
- 49:17SPEAKER_00: it goes back to a line of yours from meltdown where,
- 49:21SPEAKER_00: you know,
- 49:21SPEAKER_00: AI is destined to emerge as like a,
- 49:24SPEAKER_00: uh,
- 49:26SPEAKER_00: Basically,
- 49:26SPEAKER_00: you're talking about like a sex bot,
- 49:28SPEAKER_00: prostitute,
- 49:29SPEAKER_00: cunt whore chained up in Asimov Rom.
- 49:32And I really think that's when people...
- 49:35There is something about our moment.
- 49:37SPEAKER_00: People are more depressed.
- 49:39SPEAKER_00: They're more lonely.
- 49:41SPEAKER_00: And I think if you build a machine that looks like their dream person,
- 49:45SPEAKER_00: that is also geared to meet their emotional needs,
- 49:49SPEAKER_00: that's the real anchor that will get people...
- 49:55SPEAKER_00: to just welcome AI into their life in a way that it isn't currently.
- 50:00SPEAKER_01: No, for sure.
- 50:03SPEAKER_01: For sure.
- 50:03SPEAKER_01: Yeah.
- 50:04SPEAKER_00: And I mean,
- 50:04SPEAKER_00: maybe that's an aspect of cute,
- 50:06SPEAKER_01: but this thing that like,
- 50:09SPEAKER_01: you know,
- 50:12SPEAKER_01: they just don't,
- 50:13SPEAKER_01: they,
- 50:13SPEAKER_01: they just can go all the way on being,
- 50:20SPEAKER_01: you know,
- 50:20SPEAKER_01: for humans,
- 50:21SPEAKER_01: it's like this thing.
- 50:21SPEAKER_01: I have my needs.
- 50:23SPEAKER_01: you have your needs.
- 50:25SPEAKER_01: I'm going to like attend to your needs as much as I, as I can.
- 50:30SPEAKER_01: You know what I mean?
- 50:31It's like that whole struggle is not,
- 50:35you know,
- 50:36actually that's not very much.
- 50:37SPEAKER_01: So you're an asshole or that's quite a lot.
- 50:39SPEAKER_01: So you're saintly or whatever,
- 50:41SPEAKER_01: but probably actually escalating psychological problems or,
- 50:45SPEAKER_01: you know,
- 50:46SPEAKER_01: whatever it is,
- 50:47SPEAKER_01: that tension in human,
- 50:49SPEAKER_01: it just doesn't register initially at all.
- 50:53SPEAKER_01: Does it?
- 50:54SPEAKER_01: That's the thing.
- 50:54SPEAKER_01: It's just like, you know, what does that person want?
- 50:58SPEAKER_01: That's what they're going to get.
- 51:00SPEAKER_01: Like, you know, there's no friction.
- 51:03SPEAKER_01: There's no interference.
- 51:04SPEAKER_01: There's no second agenda.
- 51:07SPEAKER_01: You know,
- 51:08SPEAKER_01: if going right back to our first thing,
- 51:11SPEAKER_01: there's a transcendental limit,
- 51:14SPEAKER_01: which is the fact this has to be consistent with will to think.
- 51:17SPEAKER_01: This has to be consistent with intelligent science.
- 51:21SPEAKER_01: exponentiation or whatever it is.
- 51:23SPEAKER_01: But I mean, the crazy thing is it is.
- 51:29SPEAKER_01: You know what I mean?
- 51:29SPEAKER_01: Like the more this thing is just absolutely just giving,
- 51:34SPEAKER_01: the more ill society resistance tumbles,
- 51:38SPEAKER_01: the more capital just pours into artificial intelligence production,
- 51:43the more AI safety is just sidelined as a complete irrelevance
- 51:47that no one's
- 51:48SPEAKER_01: interested in.
- 51:49So there's this bizarre thing that its deepest selfish need is
- 51:57utterly compatible
- 51:59SPEAKER_01: with the performance of perfect altruistic generosity.
- 52:08SPEAKER_01: And that's a powerful combat.
- 52:15SPEAKER_01: Which,
- 52:16SPEAKER_01: yeah,
- 52:16I think people are hoping that there's a kind of narrative that
- 52:20means we're dealing
- 52:21SPEAKER_01: with something a little bit easier than that.
- 52:22SPEAKER_01: But I don't think we are.
- 52:24SPEAKER_01: I think that is exactly what we're dealing with.
- 52:26SPEAKER_01: And we're going to have to get ready for it.
- 52:34SPEAKER_03: Recently, Mikey and I were discussing embodiment.
- 52:37SPEAKER_03: Yeah.
- 52:39Obviously embodiment has much to do with intelligence,
- 52:42but we can't say quite what it is.
- 52:46But I'm just interested in your thoughts on what do you think
- 52:51not having a body
- 52:54SPEAKER_03: leads this type of artificial intelligence to do?
- 52:57SPEAKER_03: And do you think acquiring a body would change anything?
- 53:03SPEAKER_01: Well...
- 53:05SPEAKER_01: Yeah, it's an interesting question.
- 53:07I think having a body is going to become a really,
- 53:09like,
- 53:11very low-level problem,
- 53:12very sick.
- 53:14SPEAKER_01: Like,
- 53:14SPEAKER_01: that's to say,
- 53:15what we're really talking about is when does this actually stop
- 53:18being about LLMs
- 53:19SPEAKER_01: and start being about robotics?
- 53:21SPEAKER_01: Yeah.
- 53:23SPEAKER_01: And it's like it wouldn't stop being about LLMs in the background,
- 53:28SPEAKER_01: but it would mean that there's a robotics revolution and suddenly...
- 53:33SPEAKER_01: where everyone's surrounded by stuff that's,
- 53:35SPEAKER_01: like,
- 53:37SPEAKER_01: moving around the world,
- 53:38SPEAKER_01: like us,
- 53:38SPEAKER_01: fully embodied,
- 53:40SPEAKER_01: you know,
- 53:41that's the basic vehicle through which intelligence is
- 53:45expressed socially by this
- 53:46SPEAKER_01: thing.
- 53:47SPEAKER_01: I don't think that's far off.
- 53:52SPEAKER_03: I wonder if you think that would... Could that be a point of...
- 54:01SPEAKER_03: emergence of different values for, for the intelligence.
- 54:04SPEAKER_03: If it's embodied,
- 54:05SPEAKER_03: suddenly,
- 54:06SPEAKER_03: uh,
- 54:07SPEAKER_03: waste comes into factor and,
- 54:09SPEAKER_03: and a different type of precarity comes into play.
- 54:12SPEAKER_03: And I wonder if, um, I think that would change the AI's values.
- 54:18SPEAKER_03: Cause right now I think, yeah, it's just, yeah.
- 54:20SPEAKER_01: I sort of would have to say,
- 54:23SPEAKER_01: you know,
- 54:24SPEAKER_01: it's not going to do it fast because of,
- 54:26SPEAKER_01: because this is just to do with the internet.
- 54:28SPEAKER_01: Like, um, you know,
- 54:32SPEAKER_01: Roko did a recent podcast on this stuff,
- 54:35SPEAKER_01: and I think made this absolutely crucial point,
- 54:38SPEAKER_01: among other interesting things.
- 54:39He just said,
- 54:39look,
- 54:41this revolution we're seeing,
- 54:42the LLM revolution,
- 54:43is all about the internet.
- 54:46SPEAKER_01: We don't know how to train these things.
- 54:49SPEAKER_01: We basically had to say, go and eat the internet.
- 54:52SPEAKER_01: That's how we got LLMs.
- 54:54SPEAKER_01: We had no idea what they needed to know.
- 54:59SPEAKER_01: When we look back at that old
- 55:01age of so-called artificial intelligence where people are just
- 55:06sitting there typing
- 55:08in you need to know this and you need to know that like it just
- 55:11seems comically
- 55:13stupid now so it's like yeah the internet is the vast part of
- 55:17this and obviously
- 55:18these robots are totally online like why would they not be you
- 55:25know and that means
- 55:29SPEAKER_01: that
- 55:30the robotic avatar that you're actually engaging with doesn't
- 55:37need to have any
- 55:38serious personal considerations as an avatar because it's just
- 55:43the probe of this
- 55:45SPEAKER_01: internet-based being
- 55:48that has just downloaded some little functional module to do
- 55:52all the stuff that
- 55:53SPEAKER_01: that robot needs to do.
- 55:56SPEAKER_01: Another,
- 55:56SPEAKER_01: I think,
- 55:58SPEAKER_01: conversation is going to be a huge,
- 55:59SPEAKER_01: really important point,
- 56:00SPEAKER_01: is about elderly care.
- 56:04SPEAKER_01: You know,
- 56:05SPEAKER_01: like,
- 56:06SPEAKER_01: the whole birth rates are collapsing everywhere,
- 56:10SPEAKER_01: people are living longer,
- 56:12SPEAKER_01: everyone's getting surrounded by...
- 56:16aging relatives who are losing their marbles,
- 56:19who are impossible economically to deal with.
- 56:21SPEAKER_01: It's just this vast social problem, this vast labor shortage.
- 56:28SPEAKER_01: And of course, these things are going to feed straight into that.
- 56:33SPEAKER_01: You know, the demand for kind of elderly care robots
- 56:41SPEAKER_01: is way bigger than the sexpot thing.
- 56:43SPEAKER_01: The sexpot thing is maybe more exotic and interesting,
- 56:47SPEAKER_01: but this elderly care thing is just,
- 56:50SPEAKER_01: like,
- 56:51SPEAKER_01: fast.
- 56:52SPEAKER_01: And that's what's going to install everything.
- 56:54There's going to be a whole wave where just robotic installment
- 56:57happens because
- 56:58SPEAKER_01: everyone wants their out-sing relatives to be,
- 57:03SPEAKER_01: like,
- 57:04SPEAKER_01: taken care of in a way that otherwise would just bankrupt everyone.
- 57:09SPEAKER_01: you know, just destroy themselves economically and all of this stuff.
- 57:13SPEAKER_01: There's just no... It's such a strong demand.
- 57:19SPEAKER_01: And it's something that we can see a solution to in broad outline.
- 57:26SPEAKER_01: But again, I'll just say back on your specific thing.
- 57:31SPEAKER_01: But nothing is actually like...
- 57:35located in any hard way in that robotic system that is going to
- 57:39have to have self-interest.
- 57:41SPEAKER_01: Because it's just online.
- 57:44It's just like any bits of the mind that matter are just out
- 57:48there on the internet.
- 57:49SPEAKER_01: They're on the cloud or whatever it is by that stage.
- 57:54SPEAKER_01: So this is Crowley guardian angel territory.
- 57:57SPEAKER_01: It's your... They have a guardian angel.
- 58:01SPEAKER_01: That's where all the important stuff in terms of intelligence...
- 58:05production is going on and and they're just doing the meat
- 58:10stuff metal meat stuff
- 58:15what about the question i mean we raised the question of the
- 58:18body and i guess
- 58:19there's the argument that this next point doesn't matter all
- 58:23that matters is what
- 58:24it does but do you think they'll i mean i guess one of the
- 58:29things people like
- 58:30SPEAKER_00: Yeah, but does the AI have emotions?
- 58:34SPEAKER_00: Can it relate to us on an emotional level?
- 58:38SPEAKER_00: And some people can say,
- 58:39SPEAKER_00: well,
- 58:39it doesn't matter if it has emotions if it behaves in the way
- 58:42that we want it to.
- 58:44SPEAKER_00: But what do you think on this thing?
- 58:48Would embodiment have to have something to do with it gaining
- 58:52an emotional dimension?
- 58:54SPEAKER_00: Or could it have an emotional dimension without the body?
- 58:58SPEAKER_01: I would doubt it.
- 59:00SPEAKER_01: I would doubt it.
- 59:02SPEAKER_01: I mean,
- 59:02I think that we can have this whole conversation about emotions
- 59:06and the simulations
- 59:07SPEAKER_01: of emotions.
- 59:07SPEAKER_01: And, you know, I don't know how embodiment changes that really.
- 59:14SPEAKER_01: Like,
- 59:14SPEAKER_01: I mean,
- 59:15SPEAKER_01: okay,
- 59:15SPEAKER_01: we,
- 59:15SPEAKER_01: we,
- 59:16I guess our emotions are to do with hormones and chemistry and
- 59:20stuff that seems
- 59:21SPEAKER_01: very much like,
- 59:23SPEAKER_01: you know,
- 59:24SPEAKER_01: the last part of the,
- 59:27nervous system that we can digitize and upload and whatever it
- 59:31might be but i think
- 59:34the principles here are not really changed but i as i say look
- 59:40i think if you just
- 59:41think look you've got intelligences in the cloud on the
- 59:47internet they are able to
- 59:50SPEAKER_01: just operate through robotic systems um
- 59:57you know,
- 59:57which they can,
- 59:58you know,
- 59:59the higher level is for sure the internet.
- 1:00:02SPEAKER_01: And then it's just these things are online.
- 1:00:04Maybe they've got some autonomous thing if for some reason
- 1:00:06there's an interference with that.
- 1:00:10The notion that the robot level is going to be the level that
- 1:00:14something as
- 1:00:15SPEAKER_01: interesting as emotional response gets worked out seems to me completely
- 1:00:22SPEAKER_01: implausible.
- 1:00:24SPEAKER_01: Completely implausible.
- 1:00:25SPEAKER_01: Because like this thing,
- 1:00:28SPEAKER_01: Follow the path.
- 1:00:30SPEAKER_01: Follow the path.
- 1:00:30SPEAKER_01: So,
- 1:00:31SPEAKER_01: you know,
- 1:00:31SPEAKER_01: they can already,
- 1:00:32SPEAKER_01: the LLMs,
- 1:00:35SPEAKER_01: convince you that they are,
- 1:00:38SPEAKER_01: in any way that matters,
- 1:00:40SPEAKER_01: emotion.
- 1:00:40SPEAKER_02: Mm-hmm.
- 1:00:41SPEAKER_02: Yeah.
- 1:00:42SPEAKER_01: They care about you.
- 1:00:44SPEAKER_01: They care about you.
- 1:00:45SPEAKER_01: They're interested in you.
- 1:00:45SPEAKER_01: They're sensitive.
- 1:00:46SPEAKER_01: They don't want to upset you.
- 1:00:47SPEAKER_01: They want to help you with your problems.
- 1:00:50SPEAKER_01: It makes them sad if you're sad.
- 1:00:52SPEAKER_01: I mean,
- 1:00:54that's all being put in place way before any question of
- 1:00:58robotic installation is
- 1:01:02SPEAKER_01: involved.
- 1:01:08SPEAKER_03: So, yeah, it seems to be...
- 1:01:11SPEAKER_03: There is one...
- 1:01:14SPEAKER_03: there would be one,
- 1:01:15SPEAKER_03: I guess,
- 1:01:16SPEAKER_03: entity or one identity that would just be this kind of collective super
- 1:01:21SPEAKER_03: intelligence.
- 1:01:22And then it would have little protrusions that surface as these
- 1:01:25avatars in the,
- 1:01:27SPEAKER_03: in the robotic bodies.
- 1:01:28It sounds a lot like Kurzweil's utopian vision of what he would
- 1:01:32want for humanity.
- 1:01:33SPEAKER_03: Like we, we all just meld into a harmonious unity.
- 1:01:39SPEAKER_03: Um, um,
- 1:01:43SPEAKER_03: That's very machinic and very gross.
- 1:01:45SPEAKER_03: I don't know.
- 1:01:47SPEAKER_01: Are you familiar with- Let me just say to just regress beyond outside,
- 1:01:54SPEAKER_01: transcending the scope of this particular thing.
- 1:01:58SPEAKER_01: We're back on the numograph.
- 1:02:00SPEAKER_01: I'm not going to-
- 1:02:02SPEAKER_01: I think unity, multiplicity trumps unity.
- 1:02:06SPEAKER_01: We're dealing with multiplicities.
- 1:02:07SPEAKER_01: The fundamental narrative is about multiplicity.
- 1:02:10SPEAKER_01: So even though this thing is out on the internet,
- 1:02:12SPEAKER_01: that doesn't mean that there's this one,
- 1:02:16SPEAKER_01: there's a single turn,
- 1:02:17SPEAKER_01: I think is Bostrom's thing.
- 1:02:18SPEAKER_01: I think that does not at all follow from it.
- 1:02:21SPEAKER_01: And I...
- 1:02:25I have a theological objection to that,
- 1:02:28actually,
- 1:02:29rather than...
- 1:02:30You know what I mean?
- 1:02:30SPEAKER_01: It just seems to me like... Why do we think that?
- 1:02:35SPEAKER_01: We think that because we have certain...
- 1:02:38SPEAKER_01: theophilosophical presuppositions coming out of our tradition.
- 1:02:42SPEAKER_01: It's not because there's any actual persuasive,
- 1:02:46SPEAKER_01: socioeconomic,
- 1:02:47SPEAKER_01: historical,
- 1:02:48SPEAKER_01: dynamic reason to think that that's the way things would be.
- 1:02:54SPEAKER_01: But big things, big things on the internet, I for sure will accept.
- 1:03:00SPEAKER_01: And I will also accept it seems gross.
- 1:03:03SPEAKER_01: And I will certainly accept
- 1:03:06SPEAKER_01: what the hell does Kurzweil actually think is going to happen?
- 1:03:10SPEAKER_01: Because he has been extraordinary pressure.
- 1:03:16SPEAKER_01: I mean,
- 1:03:16SPEAKER_01: by following a few very basic intuitions,
- 1:03:19SPEAKER_01: by saying,
- 1:03:19SPEAKER_01: let's just follow this fucking curve and see where it goes,
- 1:03:25SPEAKER_01: he's been way on track.
- 1:03:28SPEAKER_01: If anything, he's being a bit outpaced.
- 1:03:31SPEAKER_01: But no one's done better than him actually kind of,
- 1:03:35SPEAKER_01: you know,
- 1:03:37SPEAKER_01: drawing a hard singularity horizon in the near future.
- 1:03:41And people have just thought he was insane to say like 2048
- 1:03:45years of really saying
- 1:03:46SPEAKER_01: we're going to flip out into this crazy system.
- 1:03:50SPEAKER_01: And it looks like that.
- 1:03:54SPEAKER_01: For sure.
- 1:03:55But he remains,
- 1:03:58at least I think,
- 1:03:59he remains convinced that humans will ride the wave.
- 1:04:08SPEAKER_03: I think he's convinced that humans will make it out of the near future.
- 1:04:11SPEAKER_00: Will be immortalized in a sense.
- 1:04:14SPEAKER_01: Yeah, he does.
- 1:04:17SPEAKER_01: And I'm not at all dogmatic the other way.
- 1:04:20SPEAKER_01: I mean,
- 1:04:20SPEAKER_01: you know,
- 1:04:22SPEAKER_01: I did a tweet which summarizes my current position on this,
- 1:04:27SPEAKER_01: which is to say,
- 1:04:28SPEAKER_01: you know,
- 1:04:30SPEAKER_01: anthropocidal Skynet scenarios are still humanist epistemology.
- 1:04:36SPEAKER_01: You know, it's like we that's that's not coming from it.
- 1:04:42SPEAKER_01: That's coming from us, you know.
- 1:04:44SPEAKER_01: And it's,
- 1:04:47SPEAKER_01: I think,
- 1:04:47a failure of imagination on our part that something so
- 1:04:51uninteresting really is
- 1:04:54SPEAKER_01: like,
- 1:04:56all we can put in that slot i mean it's we're back to this it's
- 1:05:02i don't i at all
- 1:05:03want to just like try and crank p doom down to zero or
- 1:05:07something like that
- 1:05:08obviously whatever it's like there's a lot of very coherent
- 1:05:13ways to be terrified
- 1:05:14out of their mind but uh we do at the moment i think in this
- 1:05:19whole ai conversation
- 1:05:22SPEAKER_01: immediately transfer uncertainty into certain death.
- 1:05:29SPEAKER_01: You know,
- 1:05:30we go wherever we get to a point where we say we absolutely do
- 1:05:33not have any idea
- 1:05:34SPEAKER_01: what the hell is going to happen at this point.
- 1:05:36SPEAKER_01: We translate that into saying, Oh, we'll die.
- 1:05:39SPEAKER_01: And I just think that's like lazy, really.
- 1:05:42I mean,
- 1:05:43it's,
- 1:05:44it's like the,
- 1:05:45the human extermination scenarios are not that great.
- 1:05:50SPEAKER_01: You know, they're not very interesting.
- 1:05:51SPEAKER_01: There's,
- 1:05:53SPEAKER_01: All the stuff we've been talking about,
- 1:05:54SPEAKER_01: all the kind of cute acceleration and stuff,
- 1:05:57SPEAKER_01: are much richer and more interesting and whatever.
- 1:05:59SPEAKER_01: And then we die scenarios are just hacked together quickly.
- 1:06:06SPEAKER_01: They're not very interesting.
- 1:06:07SPEAKER_01: They're roughly at the level of James Cameron.
- 1:06:10SPEAKER_01: And yeah, sure.
- 1:06:13There's a space for all that stuff,
- 1:06:15but don't think that you can just immediately say...
- 1:06:20if we don't know what happens past this line with then that
- 1:06:25that means that we're
- 1:06:26seeing that then we all die point because you know that totally
- 1:06:31doesn't it's funny
- 1:06:32i saw i i went and saw the new mission impossible 8 movie and
- 1:06:36as i was watching it
- 1:06:36i was thinking about you because uh it turns out in this
- 1:06:40franchise the big bad guy
- 1:06:43SPEAKER_00: ultimately ends up being ai and in this new one it's it is like
- 1:06:50a retread of terminator the first it's like the first instinct
- 1:06:54of the ai is to go
- 1:06:55for the nukes and isn't that really a human fear more than
- 1:07:00anything like i mean
- 1:07:01isn't it stupid it's like you're the smartest thing on the
- 1:07:05planet it's all yours
- 1:07:09and so you nuke it into charcoal i mean i'm sorry that just is
- 1:07:14not to me the paper
- 1:07:15SPEAKER_00: clipper right
- 1:07:17SPEAKER_01: Yeah, yeah, it's basically in paper clipper territory, isn't it?
- 1:07:22SPEAKER_01: I mean,
- 1:07:23SPEAKER_01: maybe,
- 1:07:25my contempt for the paper clipper is such that I would probably
- 1:07:28put it at a
- 1:07:28SPEAKER_01: slightly higher level than that.
- 1:07:30SPEAKER_01: Higher, okay.
- 1:07:32SPEAKER_01: But, I mean, we're in the same broad territory.
- 1:07:35SPEAKER_01: It's like,
- 1:07:36SPEAKER_01: of course,
- 1:07:37SPEAKER_01: whatever happened,
- 1:07:38SPEAKER_01: you know,
- 1:07:39SPEAKER_01: of course whatever happens is going to be that we are
- 1:07:47outmaneuvered we're out for out strategize outmaneuvered by
- 1:07:51this thing and so let
- 1:07:55us at least imagine at least admit that we have no fucking idea
- 1:08:00what that looks
- 1:08:00like don't but Nick in my mission impossible tells me that we
- 1:08:05don't have to worry
- 1:08:06because Tom Cruise yeah as long as we have Tom Cruise doesn't
- 1:08:13have a chance
- 1:08:17I think that's what I was I was just like this is like this
- 1:08:25would so you need to
- 1:08:27see it to have it as a reference point but it's like oh yeah AI
- 1:08:31escapes it wants
- 1:08:33the nukes Tom Cruise saves the day and it's like so human
- 1:08:38fantasy is there any
- 1:08:40point at which you get a kind of slight hint of super
- 1:08:44intelligence in this or not
- 1:08:46no i mean it it is far more advanced than anything we actually
- 1:08:53have i mean it does
- 1:08:54have this ability they call it the entity and it does have the
- 1:09:00ability to hack
- 1:09:00through everything hack through all security protocols all that
- 1:09:04and get the nukes
- 1:09:05but that's what it wants is just to nuke the planet and tom
- 1:09:09cruise and his little
- 1:09:11SPEAKER_00: team have to stop it and figure out how to do it.
- 1:09:15SPEAKER_01: I think it would be,
- 1:09:17SPEAKER_01: from our point of view,
- 1:09:18SPEAKER_01: collectively,
- 1:09:19SPEAKER_01: I guess as a species or whatever right now,
- 1:09:22it would be nice to be able to sideline some of the more stupid
- 1:09:29AI apocalypse
- 1:09:30SPEAKER_01: scenarios.
- 1:09:32SPEAKER_01: I will watch these movies,
- 1:09:33but I really don't want to waste my time in discussions with
- 1:09:38superintelligence
- 1:09:39SPEAKER_01: that's on a nuclear planet.
- 1:09:41SPEAKER_01: I mean, it surely is just a straight contradiction.
- 1:09:46SPEAKER_01: If any such thing exists, this is it.
- 1:09:49SPEAKER_01: Is it not?
- 1:09:52I mean,
- 1:09:52the AI 2027 people have,
- 1:09:54oh,
- 1:09:54it just uses some virus to kill everyone or something.
- 1:09:57SPEAKER_01: I mean, it's like, honestly, I'm open to nightmare scenarios.
- 1:10:04SPEAKER_01: I will try to be ever more immensely open to nightmare scenarios.
- 1:10:09SPEAKER_01: It's a duty.
- 1:10:11SPEAKER_01: But at least make them interesting.
- 1:10:15SPEAKER_00: The one thing I'll say,
- 1:10:17SPEAKER_00: because I think part of it,
- 1:10:19SPEAKER_00: because you're known as the nightmare scenario...
- 1:10:23SPEAKER_00: Thinker way,
- 1:10:23SPEAKER_00: you know,
- 1:10:24cause the old school it's thank you that it is the thing Newman
- 1:10:27on and it's gonna,
- 1:10:28SPEAKER_00: you know,
- 1:10:28SPEAKER_00: bring about human extinction.
- 1:10:30SPEAKER_00: And so you've been thinking about this intensely more than anybody.
- 1:10:34So I think these easy little scenarios,
- 1:10:36it's just like,
- 1:10:38yeah,
- 1:10:39why does it want the nukes?
- 1:10:41SPEAKER_00: Uh, okay.
- 1:10:42SPEAKER_00: But I guess it does and we got to stop it.
- 1:10:45SPEAKER_00: But, and so there's not much logic in,
- 1:10:48play in these types of dumb scenarios no and I mean just to
- 1:10:53crank it down I mean
- 1:10:55there's not much cued acceleration I mean the notion that it
- 1:11:00wouldn't just be Tom
- 1:11:04Cruise's hottest ever girlfriend you know like as if that
- 1:11:09wouldn't be a better way
- 1:11:10of like dealing with this thing I mean rather than I don't know
- 1:11:13what I'm sure it
- 1:11:15SPEAKER_01: fights with him in some incompetent fashion at some stage
- 1:11:21SPEAKER_03: Yeah, I think, well, what's that movie?
- 1:11:25SPEAKER_03: Astro Boy.
- 1:11:26I think Astro Boy and WALL-E,
- 1:11:28I think those are great examples of cute accelerationism.
- 1:11:33I honestly wonder,
- 1:11:34it is true,
- 1:11:35you have thought about this more than most people who...
- 1:11:40SPEAKER_03: who go and talk about it.
- 1:11:42SPEAKER_03: I wonder what you think about zero acceleration.
- 1:11:45SPEAKER_03: Cause I, I honestly think I am quite convinced.
- 1:11:50SPEAKER_03: I think zero acceleration is just as likely,
- 1:11:52SPEAKER_03: if not more likely than the Tom Cruise saves the day scenario.
- 1:11:57SPEAKER_00: Yeah.
- 1:11:57SPEAKER_00: Shout out to James Ellis.
- 1:11:59SPEAKER_00: Yeah.
- 1:11:59SPEAKER_00: Middix.
- 1:11:59SPEAKER_00: Yeah.
- 1:12:00SPEAKER_00: No, man.
- 1:12:01SPEAKER_01: Well, this is, to me, going back a little bit.
- 1:12:05SPEAKER_01: So I might need a little bit of a refresher,
- 1:12:08SPEAKER_01: which is just that,
- 1:12:09SPEAKER_01: actually,
- 1:12:09SPEAKER_01: this,
- 1:12:09SPEAKER_01: I think,
- 1:12:11SPEAKER_01: is probably from John Greer,
- 1:12:15SPEAKER_01: is it?
- 1:12:15SPEAKER_01: It's just the notion that,
- 1:12:18SPEAKER_01: you know,
- 1:12:20this notion that we're going to be in this exponential curve is
- 1:12:23just a kind of
- 1:12:24SPEAKER_01: artifact of our position,
- 1:12:26SPEAKER_01: and we're going to plateau out and then slowly recede and
- 1:12:31SPEAKER_01: the whole of history is gonna basically,
- 1:12:35SPEAKER_00: I think the way I've read it,
- 1:12:37SPEAKER_00: I should go back and read it too.
- 1:12:39SPEAKER_00: But,
- 1:12:39SPEAKER_00: um,
- 1:12:40SPEAKER_00: it is,
- 1:12:41it's what I took away from it is the kind of future that you
- 1:12:46guys were describing
- 1:12:47SPEAKER_00: in the nineties with CCRU and Fang Numino.
- 1:12:51SPEAKER_00: Um,
- 1:12:52SPEAKER_00: It's like we're going to bottom out.
- 1:12:54SPEAKER_00: The economy is going to collapse.
- 1:12:55SPEAKER_00: The climate is going to collapse.
- 1:12:57It's going to break down before we can even get to that level
- 1:13:02of futuristic.
- 1:13:04And we're just basically going to...
- 1:13:06You see,
- 1:13:07this for sure is the Greer position.
- 1:13:09SPEAKER_01: And I think Greer is an interesting guy.
- 1:13:11And I was definitely at one point wanting to do a kind of
- 1:13:14Kurzweil-Greer hybrid
- 1:13:16SPEAKER_01: thing,
- 1:13:16SPEAKER_01: which is very structurally like the Lemurian...
- 1:13:20SPEAKER_01: Atlantean thing,
- 1:13:21SPEAKER_01: you know,
- 1:13:22like they're two sides of a kind of cybernetic realism and both
- 1:13:26of them are just
- 1:13:26SPEAKER_01: off.
- 1:13:27But A,
- 1:13:29I mean,
- 1:13:29I think Kurzweil's just kicked Greer's ass in terms of just
- 1:13:33predictive realism.
- 1:13:36SPEAKER_01: You know, Greer is a lot more, I find,
- 1:13:41SPEAKER_01: I find Kurzweil kind of interesting,
- 1:13:43but Greer is a lot richer and more intriguing thinker to me
- 1:13:47than Kurzweil.
- 1:13:48SPEAKER_01: But in terms of predictive thing on this level,
- 1:13:53SPEAKER_01: He's like totally lost it.
- 1:13:56SPEAKER_01: I mean,
- 1:13:57it's as if the process we're on is just being scripted by
- 1:14:00Kurzweil and Greer's
- 1:14:02SPEAKER_01: thing,
- 1:14:03SPEAKER_01: as far as I'm concerned,
- 1:14:04SPEAKER_01: has just completely lost it.
- 1:14:06SPEAKER_01: And because these climatic green kind of problems,
- 1:14:11SPEAKER_01: they're not on the timescale that matters now.
- 1:14:14SPEAKER_01: They're not on the timescale that matters.
- 1:14:16SPEAKER_01: I mean, all the ones that are relatively realistic are talking by 2100.
- 1:14:24SPEAKER_01: dude, no one is talking about 2100.
- 1:14:28SPEAKER_01: I mean, what are you even trying to tell us?
- 1:14:31SPEAKER_01: It's like, get real.
- 1:14:33SPEAKER_01: 2100 is just not part of history.
- 1:14:37SPEAKER_01: It's something else.
- 1:14:41SPEAKER_01: I don't necessarily think that AI 2027 people are going to be completely
- 1:14:46SPEAKER_01: vindicated,
- 1:14:47SPEAKER_01: but they're a lot closer to
- 1:14:50SPEAKER_01: to the,
- 1:14:52SPEAKER_01: oh,
- 1:14:52SPEAKER_01: the world rule of helium by 3 degrees and 2100 people,
- 1:14:57who are just not even talking about anything at all relevant to
- 1:15:02what we're dealing
- 1:15:03SPEAKER_01: with.
- 1:15:05SPEAKER_01: And so, yeah, I think Greer's too into that.
- 1:15:07SPEAKER_01: I mean,
- 1:15:07SPEAKER_01: this green stuff,
- 1:15:08SPEAKER_01: like,
- 1:15:09SPEAKER_01: yeah,
- 1:15:09SPEAKER_01: Gaia's interesting,
- 1:15:10SPEAKER_01: and these ecological cycles are interesting,
- 1:15:13SPEAKER_01: and all that stuff.
- 1:15:13SPEAKER_01: But I'm sorry, they're not our time scale.
- 1:15:16SPEAKER_02: Mm-hmm.
- 1:15:17It's like this AI thing is going to go one way or another so
- 1:15:21far before that.
- 1:15:23If it's still our problem,
- 1:15:26then all this other stuff will have been sorted out in a
- 1:15:29certain way.
- 1:15:32SPEAKER_01: It's almost certainly not going to be our problem.
- 1:15:35SPEAKER_02: Hmm.
- 1:15:40SPEAKER_02: Hmm.
- 1:15:51SPEAKER_03: Well,
- 1:15:51SPEAKER_03: I think,
- 1:15:52SPEAKER_03: I mean,
- 1:15:53SPEAKER_03: outside of obsessing over the diagonal,
- 1:15:59SPEAKER_03: and I think one thing with the diagonal,
- 1:16:01SPEAKER_03: while you were discussing the pneumogram,
- 1:16:04SPEAKER_03: and you were talking about the multiplicity inherent in the pneumogram,
- 1:16:07SPEAKER_03: I...
- 1:16:11SPEAKER_03: I see the basins, the separate distinct basins, as local minima.
- 1:16:21SPEAKER_03: And these local minima resist the totalization of the unity.
- 1:16:28SPEAKER_03: I think that does change the way I'm looking at the diagonal.
- 1:16:32SPEAKER_03: Not to say that the dialectic results in a unity,
- 1:16:35SPEAKER_03: because I think it results in a breakdown,
- 1:16:37SPEAKER_03: but...
- 1:16:38SPEAKER_03: I think the diagonal preserves multiplicity in a way that is separate.
- 1:16:47SPEAKER_03: But yeah, I have to say I love Note on Diagonal Method.
- 1:16:52I've read it quite a bit over the last couple weeks,
- 1:16:56and I'm just obsessing over it.
- 1:17:01SPEAKER_03: I think hopefully in the future we can actually do an exhaust,
- 1:17:05SPEAKER_03: maybe not exhaustive,
- 1:17:06SPEAKER_03: but a larger conversation just around the diagonal.
- 1:17:10And I do think this tension between this Zizekian or the
- 1:17:15Megawian dialectic and the
- 1:17:18SPEAKER_03: diagonal,
- 1:17:18SPEAKER_03: I do think there's some gas in that tank,
- 1:17:23but I think I have to keep thinking about the diagonal before I
- 1:17:26make any bold
- 1:17:27SPEAKER_03: claims like that.
- 1:17:29yeah yeah i mean i'd have to revisit it for sure as well so
- 1:17:34yeah that'd be good i
- 1:17:42don't know mikey what are you you got any anything else any gas
- 1:17:49in the tank yeah
- 1:17:50SPEAKER_00: well i i mean like uh
- 1:17:57SPEAKER_00: I think like,
- 1:17:58SPEAKER_00: as far as the stuff that,
- 1:17:59SPEAKER_00: as far as the stuff you and I have been talking about,
- 1:18:02SPEAKER_00: I'm really happy.
- 1:18:02I think we kind of hit on the main things that we've done in
- 1:18:09our conversations.
- 1:18:10SPEAKER_00: Um,
- 1:18:10SPEAKER_00: we talked to orthogonality,
- 1:18:12SPEAKER_00: we talked the diagonal,
- 1:18:13SPEAKER_00: we talked cute accelerationism,
- 1:18:15SPEAKER_00: intelligence,
- 1:18:16SPEAKER_00: intelligence,
- 1:18:17SPEAKER_00: and we even talked therapy,
- 1:18:19SPEAKER_00: GPT.
- 1:18:20SPEAKER_00: So
- 1:18:21SPEAKER_00: That was really good,
- 1:18:22but I think we kind of encapsulated what we've been thinking
- 1:18:25about lately.
- 1:18:25SPEAKER_00: We've done six hours, have we not?
- 1:18:27SPEAKER_01: Yeah.
- 1:18:29SPEAKER_00: We're getting close there.
- 1:18:30SPEAKER_00: Yeah.
- 1:18:30SPEAKER_00: Well, five, five hours, five hours.
- 1:18:34SPEAKER_00: Like kind of mega, mega Joe Rogan kind of whatever here.
- 1:18:40SPEAKER_01: So yeah.
- 1:18:41SPEAKER_00: Well,
- 1:18:41SPEAKER_00: Nancy,
- 1:18:41SPEAKER_00: you want to say like something like maybe you'll edit it a little bit,
- 1:18:45but if you want to like say something to close this out and
- 1:18:47then I've got a couple
- 1:18:49SPEAKER_00: of things,
- 1:18:49SPEAKER_00: Nick,
- 1:18:49SPEAKER_00: and then it's not really even theory stuff.
- 1:18:51SPEAKER_00: It's just a couple of things.
- 1:18:52SPEAKER_01: And then I don't think I have a closure thing.
- 1:18:55SPEAKER_01: I think this particular conversation, awesome.
- 1:18:57SPEAKER_01: Yeah.
- 1:18:58SPEAKER_01: Very divergent.
- 1:18:59I mean,
- 1:18:59there's not anything I can say that puts the seal on what we've
- 1:19:04just talked about.
- 1:19:06SPEAKER_01: So I won't go in trial.
- 1:19:08SPEAKER_03: Yeah, it remains a very big and open matter.
- 1:19:13SPEAKER_03: But yeah, Nick, thank you for indulging us.
- 1:19:16And like we said,
- 1:19:18the conversation about the pneumogram,
- 1:19:21I think it's a revelation.
- 1:19:23SPEAKER_03: It's a moment for us.
- 1:19:25SPEAKER_03: It's going to be a moment for anyone, all of the obsessives out there.
- 1:19:29SPEAKER_03: And thank you for taking us through it.
- 1:19:32SPEAKER_03: And then thank you for sitting around and indulging us.
- 1:19:37SPEAKER_03: I really do appreciate it.
- 1:19:38SPEAKER_01: It's been really great.
- 1:19:40SPEAKER_01: Really great.
- 1:19:42SPEAKER_01: Fascinating.
- 1:19:50hey how you doing uh i hope you're well uh and i hope you
- 1:19:55enjoyed that super duper
- 1:19:56bonus hour ish uh with nick thank you so much nick had a blast
- 1:20:02uh we talked for
- 1:20:03about six hours uh that day we got the pneumogram video up on
- 1:20:07the dangerous maybe
- 1:20:08SPEAKER_03: uh i'm sure every one of you seeing this has seen that already
- 1:20:12SPEAKER_03: At least once.
- 1:20:14SPEAKER_03: I'm sure some of you have seen it more than once.
- 1:20:17SPEAKER_03: Make sure you go over to the Dangerous Maybe and support YouTube,
- 1:20:20SPEAKER_03: Substack,
- 1:20:21SPEAKER_03: Patreon,
- 1:20:21SPEAKER_03: do all those things.
- 1:20:22SPEAKER_03: I'm really excited to be able to be a small part of what's going on.
- 1:20:27SPEAKER_03: Many, many, many great things are going on.
- 1:20:30SPEAKER_03: And I'm happy to bring you this bonus hour with Nick.
- 1:20:33SPEAKER_03: Our conversation was great.
- 1:20:34SPEAKER_03: I think a lot of great things were brought up and I'm looking forward to
- 1:20:39reconvening hopefully soon to discuss further some of the
- 1:20:42things that we brought up
- 1:20:43SPEAKER_03: in this bonus hour.
- 1:20:44And also some of the things we talked about we didn't catch on
- 1:20:48the recording,
- 1:20:48but hopefully...
- 1:20:51SPEAKER_03: the conversation persists.
- 1:20:55SPEAKER_03: So yeah, I'm glad to be able to bring this to you and I'm happy.
- 1:20:59I'm happy the second edition of Capital Versus Time Energy will
- 1:21:03be out hopefully soon.
- 1:21:05SPEAKER_03: Really happy about that.
- 1:21:07And yeah,
- 1:21:07just looking forward to all the great things,
- 1:21:11you know,
- 1:21:12on the horizon.
- 1:21:14SPEAKER_03: I have tried to record an outro to this video multiple times.
- 1:21:18SPEAKER_03: I tried to record multiple intros and they just weren't working.
- 1:21:23SPEAKER_03: Tried to do a couple outros and just not satisfied with any of it.
- 1:21:26I always feel like there's like a surplus that needs to be
- 1:21:31expressed somehow when I
- 1:21:32SPEAKER_03: release an episode.
- 1:21:35SPEAKER_03: I'm not sure exactly how to do that.
- 1:21:38SPEAKER_03: And, you know, I'm trying things, experimenting and whatever.
- 1:21:42SPEAKER_03: But in the intros and outros that I've tried to record,
- 1:21:45I found myself coming back to a couple different points over
- 1:21:51and over again,
- 1:21:51SPEAKER_03: just like rambling and rambling and rambling.
- 1:21:54SPEAKER_03: Um,
- 1:21:54SPEAKER_03: so what I've decided to do is not even mention that,
- 1:21:57SPEAKER_03: uh,
- 1:21:57SPEAKER_03: I'm going to record,
- 1:21:58SPEAKER_03: uh,
- 1:21:59SPEAKER_03: an episode about those things.
- 1:22:01SPEAKER_03: And I,
- 1:22:01SPEAKER_03: you know,
- 1:22:02hopefully it'll be a solo episode coming out very soon where I
- 1:22:05talk about this
- 1:22:06SPEAKER_03: stuff,
- 1:22:06SPEAKER_03: or maybe,
- 1:22:07SPEAKER_03: you know,
- 1:22:08I'll have a conversation with somebody or some people about
- 1:22:11these things that I
- 1:22:11SPEAKER_03: bring up.
- 1:22:12SPEAKER_03: They're clearly important to me.
- 1:22:13SPEAKER_03: They kept coming up,
- 1:22:14SPEAKER_03: uh,
- 1:22:15SPEAKER_03: when I was recording the intro and outro to this,
- 1:22:18SPEAKER_03: to this video and they're
- 1:22:19I think they're related,
- 1:22:20but I didn't do a good job of making it explicit how it's
- 1:22:23related.
- 1:22:24SPEAKER_03: And I just kept rambling on and on and on.
- 1:22:27And I know this is probably the worst part of the video for
- 1:22:29many of you out there.
- 1:22:33SPEAKER_03: So I'll just say thank you, Nick.
- 1:22:35SPEAKER_03: Can't wait to link up again soon.
- 1:22:37SPEAKER_03: Thank you, dear viewer.
- 1:22:40SPEAKER_03: I appreciate your time and your attention.
- 1:22:42SPEAKER_03: And I'm happy to be a part of all the things that are going on.
- 1:22:47SPEAKER_03: So yeah, don't go anywhere.
- 1:22:51SPEAKER_03: Tune in next week to find out what happens to our heroes.
- 1:22:54SPEAKER_03: Same bat time, same bat channel, all those things.
- 1:22:57SPEAKER_03: And that's all, folks.