WEBVTT - The AI Industry Must Stop Doom Trolling w/ Cal Newport

0:00:02.480 --> 0:00:07.280
<v Speaker 1>Also media, Hello and welcome to Better Offline. I'm, of

0:00:07.320 --> 0:00:21.800
<v Speaker 1>course your host ed Zitron. That's right, I'm back all

0:00:21.800 --> 0:00:23.000
<v Speaker 1>of you. And to be clear, when I said I

0:00:23.040 --> 0:00:25.079
<v Speaker 1>was having wisdom teeth surgery, I meant I was having

0:00:25.120 --> 0:00:27.720
<v Speaker 1>more of them added to make me more powerful. The

0:00:27.760 --> 0:00:30.440
<v Speaker 1>operation was the success. I apologize also for doing this

0:00:30.480 --> 0:00:33.120
<v Speaker 1>week in reverse with the longer episode on the Friday

0:00:33.120 --> 0:00:35.320
<v Speaker 1>and the monologue on the welth I guess Tuesday. I

0:00:35.360 --> 0:00:37.640
<v Speaker 1>guess I had to put out open ayes financials at

0:00:37.680 --> 0:00:39.879
<v Speaker 1>the last minute, and well, here we are. Subscribe to

0:00:39.880 --> 0:00:43.400
<v Speaker 1>the newsletter ideally the premium support your local zitron and

0:00:43.600 --> 0:00:46.239
<v Speaker 1>as ever, if you have anyone else's financials, you can

0:00:46.240 --> 0:00:49.599
<v Speaker 1>catch me on ezitron Dot seventy six on signal. It's

0:00:49.680 --> 0:00:52.239
<v Speaker 1>going to be in the episode notes. Anyway, Today, now

0:00:52.240 --> 0:00:54.680
<v Speaker 1>that I'm done drinking milkshakes, I'm joined by the wonderful

0:00:54.680 --> 0:00:58.760
<v Speaker 1>computer science professor and writer Cal Newport to talk about AI,

0:00:58.880 --> 0:01:01.920
<v Speaker 1>but specifically to talk about what he calls doom trolling

0:01:02.160 --> 0:01:04.240
<v Speaker 1>in one of the best pieces I have read on

0:01:04.360 --> 0:01:05.200
<v Speaker 1>AI ever.

0:01:06.720 --> 0:01:10.680
<v Speaker 2>Cal Well, first of all, I'm disappointed I've sent you

0:01:10.720 --> 0:01:14.760
<v Speaker 2>my financials multiple times in your signal chat and yet

0:01:14.760 --> 0:01:17.080
<v Speaker 2>you have not written ten thousand words about them yet.

0:01:17.200 --> 0:01:18.919
<v Speaker 2>So I don't know what you're doing here.

0:01:19.640 --> 0:01:22.880
<v Speaker 1>You haven't lost eleven billion dollars or anything. You need

0:01:22.959 --> 0:01:25.960
<v Speaker 1>to get those numbers down. No, but I write about

0:01:26.000 --> 0:01:26.760
<v Speaker 1>bad companies.

0:01:26.880 --> 0:01:27.640
<v Speaker 2>I wrote it off though.

0:01:27.640 --> 0:01:30.480
<v Speaker 1>It's okay, Yeah, you moved it to an SPV.

0:01:30.880 --> 0:01:33.240
<v Speaker 2>That's right. I may have spent eleven billion dollars, but

0:01:33.280 --> 0:01:36.600
<v Speaker 2>really I was profitable. It's just a matter exactly you

0:01:36.640 --> 0:01:37.240
<v Speaker 2>look at it.

0:01:37.840 --> 0:01:40.880
<v Speaker 1>But let's talk doom trolling, because you have this fantastic

0:01:40.920 --> 0:01:44.120
<v Speaker 1>piece that talks about the fact that all these companies Anthropic,

0:01:44.280 --> 0:01:47.440
<v Speaker 1>Open AI, even Deep Mind to an extent, just keep

0:01:47.560 --> 0:01:51.040
<v Speaker 1>selling things based on the destruction that AI will allegedly cause,

0:01:51.280 --> 0:01:53.320
<v Speaker 1>though they appear to be talking about another kind of

0:01:53.320 --> 0:01:55.000
<v Speaker 1>technology entirely when you think about him.

0:01:55.320 --> 0:01:58.360
<v Speaker 2>Yeah, I mean, look, I think we have forgotten are

0:01:58.400 --> 0:02:01.960
<v Speaker 2>overlooking how strange this is, right, because what we have here,

0:02:02.040 --> 0:02:05.720
<v Speaker 2>let's just let's abstract away from the particular products here

0:02:05.720 --> 0:02:07.680
<v Speaker 2>and just say what do we have here? More abstractly,

0:02:07.680 --> 0:02:12.120
<v Speaker 2>as a business story, we have consumer product companies who

0:02:12.200 --> 0:02:16.080
<v Speaker 2>are on a regular basis through interviews and official white

0:02:16.080 --> 0:02:21.519
<v Speaker 2>paper releases trying to terrify their consumers about the harms

0:02:21.520 --> 0:02:23.800
<v Speaker 2>that their products are going to cause. And if this

0:02:23.960 --> 0:02:28.040
<v Speaker 2>was any of the industry, this would be completely mind boggling.

0:02:28.120 --> 0:02:30.120
<v Speaker 2>The example I had in the in the New York

0:02:30.120 --> 0:02:32.240
<v Speaker 2>Times op ed is like if the CEO of Ford,

0:02:32.840 --> 0:02:35.040
<v Speaker 2>You'll put out a white paper and was like, we're

0:02:35.080 --> 0:02:37.440
<v Speaker 2>really worried that our f one fifties are going to

0:02:37.440 --> 0:02:40.280
<v Speaker 2>start catching on fire and burning people to death. But

0:02:40.320 --> 0:02:44.320
<v Speaker 2>you know, automotive technology, we can't impede it, and we're

0:02:44.320 --> 0:02:45.359
<v Speaker 2>just going to get back to making them.

0:02:45.320 --> 0:02:47.360
<v Speaker 1>We're like, we've got to go two thousand miles an

0:02:47.400 --> 0:02:50.600
<v Speaker 1>hour at some point in the future. Yeah, we would

0:02:50.680 --> 0:02:52.520
<v Speaker 1>just ban Yeah, we'd ban it.

0:02:52.600 --> 0:02:54.200
<v Speaker 2>No, what's consumer say, No one would buy?

0:02:54.480 --> 0:02:57.880
<v Speaker 1>Yeah, just it's really it is something that's gen I

0:02:57.880 --> 0:03:00.000
<v Speaker 1>loved the term doom trolling as well because it really

0:03:00.000 --> 0:03:03.280
<v Speaker 1>he frames it well because it's it's all part of

0:03:03.320 --> 0:03:06.320
<v Speaker 1>a larger complex than never talking about what they're building today.

0:03:06.560 --> 0:03:10.240
<v Speaker 1>It's always this theoretical, scary thing that might it's always

0:03:10.440 --> 0:03:12.720
<v Speaker 1>two years or two minutes away, depending on who you're

0:03:12.760 --> 0:03:13.200
<v Speaker 1>talking to.

0:03:13.400 --> 0:03:15.480
<v Speaker 2>Yeah, So what I did was, I said, let's actually

0:03:15.480 --> 0:03:17.959
<v Speaker 2>think this through because I'm also a digital ethicist. It's like,

0:03:18.000 --> 0:03:20.679
<v Speaker 2>let's actually think this through with our ethics had on, right, Like,

0:03:20.760 --> 0:03:23.280
<v Speaker 2>let's think through what particularly anthropic I think is the

0:03:23.280 --> 0:03:26.760
<v Speaker 2>most systematic in this propaganda campaign that you know, I

0:03:26.760 --> 0:03:29.200
<v Speaker 2>analogized that to a cat leaving a dead bird on

0:03:29.240 --> 0:03:32.000
<v Speaker 2>your doorstep and walking away. They just every two to

0:03:32.080 --> 0:03:36.080
<v Speaker 2>three weeks want to drop some sort of doom troll

0:03:36.200 --> 0:03:38.120
<v Speaker 2>document to get people freaked out, and then they just

0:03:38.160 --> 0:03:39.680
<v Speaker 2>back away and go back to what they're doing. And

0:03:39.800 --> 0:03:41.160
<v Speaker 2>I kind of got fed up with this, so I said,

0:03:41.200 --> 0:03:45.840
<v Speaker 2>let's do a moral analysis about this strategy. And I said, look,

0:03:45.840 --> 0:03:49.880
<v Speaker 2>there's only two options here. Option number one. They really

0:03:49.920 --> 0:03:52.160
<v Speaker 2>do think that technology they're building has put them on

0:03:52.200 --> 0:03:54.840
<v Speaker 2>a trajectory to and the best case, destroying the economy

0:03:54.920 --> 0:03:56.960
<v Speaker 2>or in the worst case, wiping out humanity, in which

0:03:57.000 --> 0:04:00.840
<v Speaker 2>case every ethical system of you know, any merit would

0:04:00.840 --> 0:04:03.440
<v Speaker 2>say you have no other choice but to stop immediately

0:04:03.440 --> 0:04:05.440
<v Speaker 2>what you're doing and use all of your time, energy

0:04:05.440 --> 0:04:07.200
<v Speaker 2>and treasure to try to stop other people from doing

0:04:07.280 --> 0:04:09.720
<v Speaker 2>the same. The second case is you don't fully believe it,

0:04:10.280 --> 0:04:13.480
<v Speaker 2>which means you are laundering the anxiety of hundreds of

0:04:13.480 --> 0:04:15.840
<v Speaker 2>millions of people to try to make the bank balances

0:04:15.840 --> 0:04:19.520
<v Speaker 2>of a vanishingly small number of original stockholders larger, which

0:04:19.520 --> 0:04:22.640
<v Speaker 2>I think is equally as morally monstrous. There is no

0:04:22.839 --> 0:04:27.520
<v Speaker 2>moral analysis that says trying to frequently terrify people about

0:04:27.560 --> 0:04:30.240
<v Speaker 2>the things you're building while making no adjustments to the

0:04:30.240 --> 0:04:32.600
<v Speaker 2>things you're building, there's no moral calculus that says that

0:04:32.680 --> 0:04:36.039
<v Speaker 2>makes sense. But because AI is new and people don't

0:04:36.120 --> 0:04:39.080
<v Speaker 2>understand the technology, and it's shiny, and we saw James

0:04:39.080 --> 0:04:42.760
<v Speaker 2>Cameron's movies, we somehow treat this different. And I say

0:04:42.760 --> 0:04:45.920
<v Speaker 2>it's not different. This is a product that they are selling.

0:04:46.000 --> 0:04:50.400
<v Speaker 2>They are not the stewards of an inevitable technology. They

0:04:50.440 --> 0:04:53.000
<v Speaker 2>are building products according to business plans and trying to

0:04:53.080 --> 0:04:56.200
<v Speaker 2>sell them. They need to act like consumer product companies.

0:04:56.320 --> 0:04:59.560
<v Speaker 2>We need to demand that they act like consumer product companies.

0:05:00.520 --> 0:05:03.640
<v Speaker 1>Well, another thing that you have in the piece is

0:05:03.680 --> 0:05:07.960
<v Speaker 1>that it's unclear, like that dichotomy is, it isn't clear

0:05:08.760 --> 0:05:11.520
<v Speaker 1>why they're even doing this to other than to make money.

0:05:11.520 --> 0:05:13.600
<v Speaker 1>And also they're not acting as if any of this

0:05:13.640 --> 0:05:16.960
<v Speaker 1>is true. They're not doing anything that suggests that they're

0:05:16.960 --> 0:05:20.880
<v Speaker 1>actually concerned. You mentioned Daria Amida's thing about how oh

0:05:20.960 --> 0:05:23.320
<v Speaker 1>we might need to pause in AI or we should

0:05:23.320 --> 0:05:25.960
<v Speaker 1>have that option. But that isn't actually what he was suggesting.

0:05:26.000 --> 0:05:27.640
<v Speaker 1>He was suggesting something entirely different.

0:05:27.760 --> 0:05:30.160
<v Speaker 2>Well, what they said was and they got headlines I

0:05:30.360 --> 0:05:32.120
<v Speaker 2>had to look up for the factchecker. They got every

0:05:32.120 --> 0:05:35.440
<v Speaker 2>major publication to say anthropic calls for worldwide pause and

0:05:35.520 --> 0:05:38.000
<v Speaker 2>AI development. You read that section of their white paper,

0:05:38.279 --> 0:05:40.720
<v Speaker 2>which by the way, reads like sci fi that a

0:05:40.839 --> 0:05:45.000
<v Speaker 2>ninth grader wrote, right, it's not. It's pseudo scientific at best.

0:05:45.040 --> 0:05:47.560
<v Speaker 2>It's not even really scientific. It's it's I find their

0:05:47.560 --> 0:05:50.120
<v Speaker 2>white paper sort of laughable. But they said in it,

0:05:50.640 --> 0:05:52.960
<v Speaker 2>we should do a worldwide pause because the thing we're

0:05:52.960 --> 0:05:56.480
<v Speaker 2>building is so dangerous dot dot dot. But we can't

0:05:56.480 --> 0:05:58.640
<v Speaker 2>do that unless everybody in the world does the same thing,

0:05:58.640 --> 0:06:01.560
<v Speaker 2>including China essentially, and if that's not the case, we

0:06:01.600 --> 0:06:03.360
<v Speaker 2>have to just keep going full speed. So what they

0:06:03.360 --> 0:06:05.440
<v Speaker 2>said was we should do worldwide pause, but of course

0:06:05.440 --> 0:06:08.720
<v Speaker 2>we're not going to. It's really what they were saying there.

0:06:08.720 --> 0:06:10.800
<v Speaker 2>But the point is, all right. So the report we're

0:06:10.800 --> 0:06:12.760
<v Speaker 2>talking about here came out a few weeks ago. If

0:06:13.040 --> 0:06:14.920
<v Speaker 2>you haven't read it, this is basically the report where

0:06:14.960 --> 0:06:17.560
<v Speaker 2>they just out of nowhere wrote a white paper with

0:06:17.600 --> 0:06:19.920
<v Speaker 2>a bunch of charts. They always have charts where they

0:06:19.920 --> 0:06:24.400
<v Speaker 2>were saying, yeah, pictures and an animation of.

0:06:25.000 --> 0:06:27.400
<v Speaker 1>I was just thinking of the little replicating thing. We'll

0:06:27.400 --> 0:06:29.760
<v Speaker 1>get to self learning AI in a second.

0:06:29.760 --> 0:06:33.720
<v Speaker 2>But the report basically said, huh, cloud code, which is

0:06:33.760 --> 0:06:35.799
<v Speaker 2>a harness on top of an LLM for for helping

0:06:35.839 --> 0:06:37.200
<v Speaker 2>to run computer program.

0:06:37.640 --> 0:06:39.680
<v Speaker 1>Just to be clear, a harness is basically a series

0:06:39.720 --> 0:06:41.719
<v Speaker 1>of scripts that runs on top of the LM.

0:06:41.960 --> 0:06:44.920
<v Speaker 2>It's a computer program that prompts an LLM and then

0:06:46.279 --> 0:06:49.160
<v Speaker 2>parses the response to LM gives and then actually takes

0:06:49.200 --> 0:06:51.440
<v Speaker 2>actions on your computer based on what it reads. Right,

0:06:51.480 --> 0:06:54.680
<v Speaker 2>So it's yeah, and that's a whole other conversation about

0:06:54.680 --> 0:06:56.360
<v Speaker 2>the coding world and how much is the harness and

0:06:56.360 --> 0:06:58.839
<v Speaker 2>how much of the LM. But they basically said, this tool,

0:06:59.560 --> 0:07:01.640
<v Speaker 2>which is a a harness, it's an LM, it's some

0:07:01.680 --> 0:07:04.320
<v Speaker 2>other things. This tool that computer software developers are using

0:07:04.400 --> 0:07:07.480
<v Speaker 2>is getting really good, and you know that could be.

0:07:07.760 --> 0:07:10.400
<v Speaker 2>We think that it might be. Looks like it's getting

0:07:10.400 --> 0:07:13.320
<v Speaker 2>closer to being able to improve itself. And if it

0:07:13.360 --> 0:07:15.760
<v Speaker 2>does recursive self improvement, which by the way, is that

0:07:16.000 --> 0:07:18.640
<v Speaker 2>it's at the core of philosophical superintelligence, right, the sort

0:07:18.640 --> 0:07:21.760
<v Speaker 2>of a Silicon Valley cult, like thinking about computers replacing us.

0:07:22.440 --> 0:07:25.360
<v Speaker 2>If it does, hey, maybe that'll be good for humanity.

0:07:25.400 --> 0:07:28.080
<v Speaker 2>It'll cure all diseases. But also we could lose control

0:07:28.120 --> 0:07:30.720
<v Speaker 2>of it. Thank you, have a nice day. And they

0:07:30.720 --> 0:07:33.360
<v Speaker 2>walked away. They just put out that report. That's the report.

0:07:34.600 --> 0:07:36.320
<v Speaker 1>And the thing that I like about it as well

0:07:36.400 --> 0:07:38.720
<v Speaker 1>is I think that we're in the next big thing

0:07:38.800 --> 0:07:41.920
<v Speaker 1>is going to be RSI. Because so just to be

0:07:41.960 --> 0:07:44.720
<v Speaker 1>clear for the listeners, our recursive self improvement is just

0:07:44.840 --> 0:07:48.680
<v Speaker 1>AI that builds itself. It is a theoretical concept. It

0:07:48.720 --> 0:07:52.160
<v Speaker 1>has not happened yet. But mysteriously within the same six

0:07:52.280 --> 0:07:55.720
<v Speaker 1>seven day period this came out and Sam Moltman said

0:07:55.760 --> 0:07:59.200
<v Speaker 1>to his staff in a Slack message he knew would leak,

0:07:59.640 --> 0:08:02.600
<v Speaker 1>Oh yeah, yeah, we might delay our IPO if we

0:08:02.680 --> 0:08:05.320
<v Speaker 1>have signs of RSI, if we have signs of AI

0:08:05.400 --> 0:08:08.160
<v Speaker 1>that can build itself. Meta a year ago said in

0:08:08.240 --> 0:08:12.280
<v Speaker 1>Personal Superintelligence, we're seeing the early signs. And they're doing

0:08:12.360 --> 0:08:16.640
<v Speaker 1>this because, in my opinion, maybe disagreek how they've run

0:08:16.680 --> 0:08:18.720
<v Speaker 1>out of ideas to the point that they're literally going

0:08:18.760 --> 0:08:20.960
<v Speaker 1>back to what Sam Moltman said in twenty nineteen. Yeah,

0:08:21.000 --> 0:08:23.960
<v Speaker 1>we'll just ask the AI what to do. Yeah, lo'st

0:08:23.960 --> 0:08:26.560
<v Speaker 1>the AI out build. These guys have so few ideas that,

0:08:26.680 --> 0:08:28.760
<v Speaker 1>just like Christ, can't we just make the AI.

0:08:28.640 --> 0:08:31.200
<v Speaker 2>Do it OURSI is to get out of jail free

0:08:31.200 --> 0:08:34.679
<v Speaker 2>card from a technical expertise standpoint. So you know, we

0:08:35.080 --> 0:08:37.079
<v Speaker 2>saw a lot of this a year or two ago

0:08:37.880 --> 0:08:40.240
<v Speaker 2>when we had like Project twenty twenty seven. We had

0:08:40.240 --> 0:08:45.600
<v Speaker 2>these sort of post GPT for warnings of you know,

0:08:45.760 --> 0:08:48.760
<v Speaker 2>existential crisises. It's a get out of jail free card

0:08:48.760 --> 0:08:52.160
<v Speaker 2>because if you ask, like the authors of twenty twenty seven, well,

0:08:52.200 --> 0:08:55.640
<v Speaker 2>explain to me what a system that is smarter than

0:08:55.720 --> 0:08:57.960
<v Speaker 2>humans and can fire the missiles and over what does

0:08:58.000 --> 0:09:00.560
<v Speaker 2>it look like? How you architect it? Like, what what

0:09:00.679 --> 0:09:02.880
<v Speaker 2>allows it to do it? And their answer is we

0:09:02.960 --> 0:09:06.319
<v Speaker 2>don't have to figure that out. The AI will right now.

0:09:06.320 --> 0:09:10.520
<v Speaker 2>This is perfect not the history problem. We could history

0:09:10.600 --> 0:09:12.880
<v Speaker 2>lesson this this although this goes all the way back

0:09:12.920 --> 0:09:15.400
<v Speaker 2>to the nineteen sixties. So there's a there's a statistician

0:09:15.480 --> 0:09:19.440
<v Speaker 2>applied mathematician named I. J. Good and he wrote a paper.

0:09:19.480 --> 0:09:21.600
<v Speaker 2>It's really like an information theory paper, but he wrote

0:09:21.640 --> 0:09:25.000
<v Speaker 2>a paper where he said, we, if we think about it,

0:09:25.080 --> 0:09:28.120
<v Speaker 2>called it ultra intelligence. Really, all you have to do

0:09:28.960 --> 0:09:32.480
<v Speaker 2>is build a machine smart enough to improve itself, and

0:09:32.480 --> 0:09:34.720
<v Speaker 2>then that machine will improve itsself. That machin will improve itself,

0:09:34.720 --> 0:09:36.440
<v Speaker 2>and then we'll get to ultra intelligence. So we don't

0:09:36.440 --> 0:09:38.800
<v Speaker 2>have to figure out how to do it. The danger

0:09:38.880 --> 0:09:40.959
<v Speaker 2>lies in just making a machine fast enough. So it's

0:09:41.000 --> 0:09:44.920
<v Speaker 2>like a bit of a semi thought experiment that became

0:09:45.000 --> 0:09:48.480
<v Speaker 2>the er document of what I call like the philosophical

0:09:48.520 --> 0:09:52.800
<v Speaker 2>superintelligence cultists. So this sort of where the philosophers took

0:09:52.800 --> 0:09:56.440
<v Speaker 2>this idea away from the technologist and began turning it

0:09:56.480 --> 0:09:59.160
<v Speaker 2>into a sort of digital erascholasticism, where they sort of

0:09:59.160 --> 0:10:02.959
<v Speaker 2>abstracted away all technical details and just started building list

0:10:03.000 --> 0:10:04.839
<v Speaker 2>of list of list of lists of all the different

0:10:04.840 --> 0:10:06.240
<v Speaker 2>things that could happen. They just did a bunch of

0:10:06.240 --> 0:10:10.160
<v Speaker 2>thought experiments about super intelligence and recursive self improvement. They

0:10:10.240 --> 0:10:14.640
<v Speaker 2>then merged with the utopian singularity people that were coming

0:10:14.640 --> 0:10:16.560
<v Speaker 2>out of sort of Kurzweil's orbit, and you kind of

0:10:16.559 --> 0:10:20.160
<v Speaker 2>created this group that was non technical. This is not engineers,

0:10:20.440 --> 0:10:23.319
<v Speaker 2>this philosopher's ethicist and sort of gadflies and you know,

0:10:23.559 --> 0:10:26.600
<v Speaker 2>you know, speculative fiction writers and fan fiction writers like Yukowski,

0:10:26.960 --> 0:10:30.720
<v Speaker 2>and they kind of created this sort of dumorous utopian cult.

0:10:31.600 --> 0:10:35.160
<v Speaker 2>And then what happened is that group was like, we

0:10:35.200 --> 0:10:38.880
<v Speaker 2>should create a nonprofit to look at these ideas, and

0:10:38.960 --> 0:10:41.040
<v Speaker 2>we'll take sam Altman and put them in charge. We'll

0:10:41.080 --> 0:10:44.160
<v Speaker 2>call this Open Ai. And then suddenly they were like,

0:10:44.240 --> 0:10:46.600
<v Speaker 2>oh my god, this jin Ai thing is working better

0:10:46.679 --> 0:10:49.920
<v Speaker 2>than we thought, and we have now a one of

0:10:49.920 --> 0:10:52.240
<v Speaker 2>the you know, someone coming out of this general kind

0:10:52.280 --> 0:10:56.520
<v Speaker 2>of cultish let's eschatological you know, in times could come

0:10:56.520 --> 0:10:59.880
<v Speaker 2>through computers, was suddenly and accidentally in charge of the

0:11:00.000 --> 0:11:03.080
<v Speaker 2>most important company in AI. And one of his you know,

0:11:03.600 --> 0:11:06.520
<v Speaker 2>lieutenants left to start his own company. And his name

0:11:06.600 --> 0:11:09.880
<v Speaker 2>is Dario Amaday. So these the two major players we

0:11:10.040 --> 0:11:13.679
<v Speaker 2>have in AI right now, come out of philosophical superintelligence,

0:11:14.040 --> 0:11:17.640
<v Speaker 2>which uses this idea of RSI, and it's like a

0:11:17.720 --> 0:11:20.320
<v Speaker 2>it's like a parable, it's like an abstraction. This is

0:11:20.360 --> 0:11:22.880
<v Speaker 2>not like a serious it wasn't like a technical concern.

0:11:22.960 --> 0:11:25.640
<v Speaker 2>It was a like a thought experiment, and they are

0:11:25.679 --> 0:11:29.080
<v Speaker 2>completely influencing the like the daily lives and psychological well

0:11:29.080 --> 0:11:31.440
<v Speaker 2>being of hundreds of millions, and like the economic health

0:11:31.480 --> 0:11:33.280
<v Speaker 2>of the entire world. So it's it's a really an

0:11:33.280 --> 0:11:36.880
<v Speaker 2>interesting world we've ended up in now. And all of

0:11:36.920 --> 0:11:39.520
<v Speaker 2>which is to say, I don't think we're talking enough

0:11:39.520 --> 0:11:41.959
<v Speaker 2>about how weird these people are. I think we keep

0:11:42.040 --> 0:11:46.080
<v Speaker 2>steering the conversation back to the technology itself. We keep

0:11:46.120 --> 0:11:48.840
<v Speaker 2>steering the conversation back to like, well, could it take

0:11:48.880 --> 0:11:51.600
<v Speaker 2>over or not? And we need to spend some more

0:11:51.600 --> 0:11:54.000
<v Speaker 2>time saying, wait, who are these people and what are

0:11:54.040 --> 0:11:57.720
<v Speaker 2>they doing? They're doom trolling. This makes no sense. No

0:11:57.880 --> 0:12:00.480
<v Speaker 2>company ever in the history of companies was trying to

0:12:00.520 --> 0:12:03.440
<v Speaker 2>troll their own customers to be afraid of their products.

0:12:03.120 --> 0:12:07.120
<v Speaker 2>They're treating their technology like a religious object. Like, we

0:12:07.440 --> 0:12:11.400
<v Speaker 2>are underestimating the weirdness of these people, and when we

0:12:11.480 --> 0:12:15.400
<v Speaker 2>put that into focus, it completely changes I think our

0:12:15.480 --> 0:12:18.160
<v Speaker 2>own psychological relationship with what's going on right now.

0:12:19.320 --> 0:12:23.640
<v Speaker 1>Yeah, and I fully agree that it is almost I

0:12:23.679 --> 0:12:27.320
<v Speaker 1>think it's a distraction as well from the actual reality

0:12:27.360 --> 0:12:29.280
<v Speaker 1>of what these things are, and I think it's a

0:12:29.360 --> 0:12:31.720
<v Speaker 1>deliberate one. I naturally fall on the side of just

0:12:31.720 --> 0:12:34.679
<v Speaker 1>thinking these people are deeply cynical. I've met people who

0:12:34.760 --> 0:12:37.640
<v Speaker 1>think that Dario amade believes this stuff. I actually don't.

0:12:37.760 --> 0:12:39.840
<v Speaker 1>I don't think they believe a fucking thing. I think

0:12:39.880 --> 0:12:43.200
<v Speaker 1>that however you feel is however you feel. But I

0:12:43.240 --> 0:12:47.679
<v Speaker 1>think that they correctly realize that they could very easily

0:12:47.720 --> 0:12:50.600
<v Speaker 1>make people talk about something that doesn't exist just by

0:12:50.640 --> 0:12:53.000
<v Speaker 1>mentioning it. How I saw a bunch of headlines that

0:12:53.040 --> 0:12:56.480
<v Speaker 1>were like, Wow, anthropics seize the path to self building AI.

0:12:57.280 --> 0:12:58.000
<v Speaker 2>No, they don't.

0:12:58.480 --> 0:13:00.200
<v Speaker 1>Other than the fact they had a single set and

0:13:00.240 --> 0:13:03.200
<v Speaker 1>saying I see the path to this. They don't have

0:13:03.240 --> 0:13:06.360
<v Speaker 1>any proof, they don't have anything, and yeah, you're right.

0:13:06.440 --> 0:13:10.320
<v Speaker 1>These people are weird. I have seen people on XT,

0:13:10.480 --> 0:13:15.439
<v Speaker 1>everything app Twitter actually posting that we are in AI

0:13:15.520 --> 0:13:18.480
<v Speaker 1>twenty twenty seven, that everything matches up with AI twenty

0:13:18.480 --> 0:13:21.120
<v Speaker 1>twenty seven. I want to be clear that and the

0:13:21.160 --> 0:13:23.480
<v Speaker 1>reason I mentioned this is I just brought this up,

0:13:23.679 --> 0:13:26.280
<v Speaker 1>is that in early twenty twenty six, in AI twenty

0:13:26.280 --> 0:13:31.000
<v Speaker 1>twenty seven, they say that Agent one has already started

0:13:31.040 --> 0:13:35.160
<v Speaker 1>doing its own research. That's not happening. None of that's happening.

0:13:35.160 --> 0:13:37.319
<v Speaker 1>In fact, none of this is happening. These people are

0:13:37.360 --> 0:13:41.240
<v Speaker 1>just weird fantasists. I think. While there's the poison of

0:13:41.280 --> 0:13:44.040
<v Speaker 1>the philosophers in them. I just think that Warrio Ama

0:13:44.160 --> 0:13:47.160
<v Speaker 1>Day and Sam Altman both I think that they both

0:13:47.200 --> 0:13:51.000
<v Speaker 1>realize that journalists series ly con the investors reasonly con

0:13:51.080 --> 0:13:54.319
<v Speaker 1>that you can just say shit and anyone will print

0:13:54.360 --> 0:13:56.920
<v Speaker 1>it and believe you because you're a CEO.

0:13:57.720 --> 0:14:01.160
<v Speaker 2>I don't know how to answer that question, because the

0:14:01.160 --> 0:14:05.400
<v Speaker 2>more I look into how quasi religious and cultish this

0:14:05.480 --> 0:14:10.440
<v Speaker 2>sort of philosophical superintelligence community is, the more I get

0:14:10.480 --> 0:14:14.320
<v Speaker 2>confused about what these people actually believe. So go spend

0:14:14.360 --> 0:14:17.080
<v Speaker 2>time in Silicon Valley, Go spend time.

0:14:17.440 --> 0:14:20.760
<v Speaker 1>To be clear, when I say they, I'm referring specifically

0:14:20.760 --> 0:14:23.760
<v Speaker 1>to Amidei and Olman. Well, so they could they fully

0:14:23.760 --> 0:14:25.680
<v Speaker 1>believe that there are tons of people in the valley

0:14:25.800 --> 0:14:28.480
<v Speaker 1>that believe is that they believe that this is the

0:14:28.520 --> 0:14:31.680
<v Speaker 1>path to that Yeah, I think just to be abundantly clear,

0:14:31.960 --> 0:14:34.480
<v Speaker 1>I don't think everyone is cynical. I think they've been

0:14:34.520 --> 0:14:37.560
<v Speaker 1>easily conned. I believe think that they are cynical.

0:14:37.640 --> 0:14:40.560
<v Speaker 2>Now I could I could believe that as and I

0:14:40.560 --> 0:14:43.760
<v Speaker 2>could believe that as they their companies got larger and

0:14:43.800 --> 0:14:48.240
<v Speaker 2>their leadership roles got larger, that they have stepped away

0:14:48.280 --> 0:14:51.280
<v Speaker 2>like oh this is this was a little naive. That's possible.

0:14:51.320 --> 0:14:54.280
<v Speaker 2>But also I saw a news report today that Dario

0:14:54.360 --> 0:14:55.600
<v Speaker 2>I don't know if this is true, so I'm just

0:14:55.640 --> 0:14:57.440
<v Speaker 2>this a headline. I haven't read yet that he has

0:14:57.480 --> 0:15:01.360
<v Speaker 2>like one direct report. Yes, so I don't think you know,

0:15:01.600 --> 0:15:04.280
<v Speaker 2>he's he's necessarily in the war room with a thousand

0:15:04.360 --> 0:15:06.840
<v Speaker 2>engineers and like doing a lot of business strategy here too.

0:15:06.840 --> 0:15:09.360
<v Speaker 2>So it's possible that these guys are more isolated than

0:15:10.120 --> 0:15:13.080
<v Speaker 2>than we think. I don't know, but I mean, if

0:15:13.080 --> 0:15:13.520
<v Speaker 2>he has.

0:15:13.400 --> 0:15:15.680
<v Speaker 1>One direct report, he's not doing anything. He's just what

0:15:15.800 --> 0:15:17.840
<v Speaker 1>he's going to meetings. Sorry what you're saying.

0:15:18.080 --> 0:15:20.160
<v Speaker 2>Yeah, but but but coming back to it, like what

0:15:20.160 --> 0:15:22.400
<v Speaker 2>what are the whether they how much they believe or not?

0:15:23.200 --> 0:15:25.080
<v Speaker 2>And I hope they don't because again, I think it

0:15:25.120 --> 0:15:27.360
<v Speaker 2>would be morally monstrous if you really believe this that

0:15:27.400 --> 0:15:30.000
<v Speaker 2>you're like, yeah, but I want to win an IPO,

0:15:30.200 --> 0:15:32.840
<v Speaker 2>like I want to make more exactly, Yeah, I think

0:15:32.840 --> 0:15:35.520
<v Speaker 2>that would be morally monstrous, right, But like if you

0:15:35.520 --> 0:15:39.120
<v Speaker 2>think about what the benefits would would be. They've this

0:15:39.240 --> 0:15:44.120
<v Speaker 2>has led to a sort of discourse environment essentially, where

0:15:44.240 --> 0:15:47.760
<v Speaker 2>you have a normal technology that's that's advancing along a

0:15:47.800 --> 0:15:50.680
<v Speaker 2>jagged frontier, which is what happens with like new technologies.

0:15:51.160 --> 0:15:53.480
<v Speaker 2>There's like certain applications where they feel like they're making

0:15:53.520 --> 0:15:55.640
<v Speaker 2>progress and a lot of other applications where they're stuck.

0:15:55.720 --> 0:15:58.600
<v Speaker 2>And so like they've been making progress with these these

0:15:58.640 --> 0:16:02.120
<v Speaker 2>coding systems that that's kind of a scenario which there's

0:16:02.120 --> 0:16:04.760
<v Speaker 2>a lot of room for play with these type of systems.

0:16:05.600 --> 0:16:07.400
<v Speaker 2>A lot of other things like non coding agents, et

0:16:07.440 --> 0:16:10.920
<v Speaker 2>Cetera's not going so well. In math, they're making progress,

0:16:10.960 --> 0:16:12.640
<v Speaker 2>but that's such a narrow field that it's not like

0:16:12.720 --> 0:16:15.200
<v Speaker 2>it's not economically relevant to them. So you could see

0:16:15.200 --> 0:16:18.840
<v Speaker 2>this as just like a normal technology. And in chat

0:16:19.280 --> 0:16:22.440
<v Speaker 2>chat search which is basically you know, running Google search

0:16:22.480 --> 0:16:25.440
<v Speaker 2>results through a language model, the natural language search basically,

0:16:25.640 --> 0:16:28.200
<v Speaker 2>so like places they're making progress, other places they're stuck.

0:16:28.240 --> 0:16:31.480
<v Speaker 2>That's like an interesting normal technology story. But if you

0:16:31.600 --> 0:16:35.400
<v Speaker 2>layer on this other sort of eschological story about we

0:16:35.440 --> 0:16:37.760
<v Speaker 2>are on our way to utopia or dystopia and we

0:16:37.840 --> 0:16:40.600
<v Speaker 2>keep trying to keep you like upset about that. Think

0:16:40.640 --> 0:16:42.760
<v Speaker 2>about the way now when you're online, how people think

0:16:42.800 --> 0:16:46.920
<v Speaker 2>and talk about AI. It's all about if something made,

0:16:46.960 --> 0:16:49.160
<v Speaker 2>if there's some new feature that works, if there's some

0:16:49.240 --> 0:16:52.440
<v Speaker 2>new chart that goes better, that's evidence for the prophecy

0:16:52.520 --> 0:16:57.080
<v Speaker 2>is true. And then so then that that puts so

0:16:57.200 --> 0:16:59.080
<v Speaker 2>much more power. Like if you otherwise just came to

0:16:59.120 --> 0:17:02.240
<v Speaker 2>me and said like, hey, we have this like advance

0:17:02.320 --> 0:17:04.600
<v Speaker 2>in our coding harnesses, I'd be like, I don't care.

0:17:04.640 --> 0:17:07.080
<v Speaker 2>I'm not a software developer. But if you say, oh,

0:17:07.240 --> 0:17:10.920
<v Speaker 2>that means that the prophecy of the RSI gods is true,

0:17:10.960 --> 0:17:14.720
<v Speaker 2>Now suddenly that is really eventful, and I'm really excited

0:17:14.840 --> 0:17:17.800
<v Speaker 2>or scared. But I'm giving a lot of import to

0:17:17.920 --> 0:17:21.560
<v Speaker 2>your company. So I do think it generates way more

0:17:21.680 --> 0:17:25.440
<v Speaker 2>sense of import on these otherwise, like I think very

0:17:25.480 --> 0:17:29.200
<v Speaker 2>often very minor and jagged improvements. I mean, I think

0:17:29.200 --> 0:17:31.040
<v Speaker 2>with other companies, if you kept putting out graphs that

0:17:31.080 --> 0:17:33.080
<v Speaker 2>you created on measures you created, I was like, I

0:17:33.080 --> 0:17:34.800
<v Speaker 2>don't know what that means. Tell me when like, my

0:17:34.880 --> 0:17:37.360
<v Speaker 2>business is going to be doing better? Right, So there

0:17:37.440 --> 0:17:40.119
<v Speaker 2>is a huge advantage regardless of intention. There is a

0:17:40.200 --> 0:17:43.520
<v Speaker 2>huge advantage to have a doom troll because as long

0:17:43.560 --> 0:17:48.520
<v Speaker 2>as people are are terrified or extremely excited about this future,

0:17:48.560 --> 0:17:50.960
<v Speaker 2>and then there's a secondary advantage. This is cynical, but

0:17:50.960 --> 0:17:52.280
<v Speaker 2>I just want to throw it out here ed because

0:17:52.280 --> 0:17:54.760
<v Speaker 2>I'm sensing this more. I think there's a whole other

0:17:54.880 --> 0:17:58.080
<v Speaker 2>substrata of sort of like tech adjacent people who are

0:17:58.200 --> 0:18:01.000
<v Speaker 2>who are you know, online or not, who got caught

0:18:01.080 --> 0:18:04.320
<v Speaker 2>up a little bit in that type of hype type

0:18:04.400 --> 0:18:08.160
<v Speaker 2>vision and now their pride is engaged, and so now

0:18:08.280 --> 0:18:12.159
<v Speaker 2>they will interpret I encountered this all the time, not

0:18:12.200 --> 0:18:15.760
<v Speaker 2>even skepticism, just literally saying oh, I don't think that tool.

0:18:15.880 --> 0:18:18.159
<v Speaker 2>People are struggling to use that tool well, or this

0:18:18.240 --> 0:18:20.920
<v Speaker 2>might this is not a very profitable model here, They'll

0:18:20.960 --> 0:18:23.000
<v Speaker 2>have to find another model. They interpret it as like, well,

0:18:23.000 --> 0:18:26.680
<v Speaker 2>this is embarrassing because I was just telling my siblings

0:18:26.720 --> 0:18:29.320
<v Speaker 2>and everyone at dinner last night about because I used

0:18:29.320 --> 0:18:32.760
<v Speaker 2>cloud code, and I think it's how nine thousand and

0:18:32.800 --> 0:18:36.480
<v Speaker 2>I was just so excited about it. It's cognitive dissidents,

0:18:36.520 --> 0:18:38.679
<v Speaker 2>and I'm embarrassed. And so now I am a warrior

0:18:38.680 --> 0:18:40.720
<v Speaker 2>for the prophecy. So I think you also have that

0:18:40.840 --> 0:18:44.120
<v Speaker 2>substrate who are like, yeah, but what about and then

0:18:44.160 --> 0:18:46.239
<v Speaker 2>you know some metric that went up or something like that,

0:18:46.320 --> 0:18:48.199
<v Speaker 2>and that's way and their benefit as well. You have

0:18:48.200 --> 0:18:50.399
<v Speaker 2>a soldiers of marketers that are trying to basically protect

0:18:50.440 --> 0:19:00.520
<v Speaker 2>their pride.

0:19:03.040 --> 0:19:06.480
<v Speaker 1>Yeah, you've absolutely nailed it. It's the pride thing again,

0:19:06.640 --> 0:19:09.639
<v Speaker 1>hadn't thought of that, But you're right the way that

0:19:09.720 --> 0:19:12.119
<v Speaker 1>they react when you give that. I mean, I just

0:19:12.200 --> 0:19:17.400
<v Speaker 1>published Open Ais Financials and the Pretzels. They're contorting themselves

0:19:17.400 --> 0:19:22.360
<v Speaker 1>into the ways in which it's Silicon value has always

0:19:22.400 --> 0:19:30.159
<v Speaker 1>been this proudly atheistic, meritocratic, rugged pragmatism world, or at

0:19:30.200 --> 0:19:32.399
<v Speaker 1>least this is how they sold themselves. It's not about

0:19:32.440 --> 0:19:35.160
<v Speaker 1>your feelings, it's about can you ship co? Is your

0:19:35.200 --> 0:19:38.600
<v Speaker 1>software good? Does it make money? But something about this

0:19:38.720 --> 0:19:43.240
<v Speaker 1>is really entered into the realm of religiosity, not even faith,

0:19:43.400 --> 0:19:47.520
<v Speaker 1>because faith is a much stronger and more fundamentally sound thing.

0:19:47.880 --> 0:19:53.119
<v Speaker 1>This is religious cultism. But the God is capital Like

0:19:53.160 --> 0:19:56.280
<v Speaker 1>you can say it's Ai, but the God is capitalism.

0:19:56.640 --> 0:19:59.480
<v Speaker 1>It is this idea that the big money people who

0:19:59.520 --> 0:20:02.160
<v Speaker 1>are constantly threatening you, like they just they're like, yeah,

0:20:02.160 --> 0:20:05.520
<v Speaker 1>it's gonna kill you or it's gonna take everything. I

0:20:05.520 --> 0:20:08.159
<v Speaker 1>guess it's Pascal's wager as well. It's well, if I

0:20:08.200 --> 0:20:09.800
<v Speaker 1>don't know if this is the case, but if they're

0:20:09.800 --> 0:20:12.400
<v Speaker 1>all saying it, and if I'm wrong, the Roco's bassilisk

0:20:12.440 --> 0:20:17.240
<v Speaker 1>will drop an anvil on me. I guess it's it's frustrating.

0:20:17.280 --> 0:20:20.560
<v Speaker 1>But also I don't know what these people will do

0:20:20.920 --> 0:20:23.159
<v Speaker 1>if the bubble, when the bubble burst. I don't know

0:20:23.160 --> 0:20:26.000
<v Speaker 1>why I'm saying if when it bursts, I don't know

0:20:26.040 --> 0:20:28.840
<v Speaker 1>what these people do because their worldview is built around

0:20:29.800 --> 0:20:33.800
<v Speaker 1>existing outside of reality, almost like that just this chatbot

0:20:34.720 --> 0:20:38.600
<v Speaker 1>will become God will become autonomous, and because I was

0:20:38.600 --> 0:20:42.719
<v Speaker 1>there saying it will become autonomous, I will be given something.

0:20:42.760 --> 0:20:48.520
<v Speaker 1>It's unclear what the reward is other than simpering, and

0:20:48.600 --> 0:20:51.479
<v Speaker 1>there's something in it for them, that's the thing. It's like,

0:20:51.520 --> 0:20:55.520
<v Speaker 1>I guess maybe it's community. They're finding. It's replace AI

0:20:55.680 --> 0:20:58.400
<v Speaker 1>with God, just like with church, and they would mock

0:20:58.800 --> 0:21:01.080
<v Speaker 1>whoever was talking about it. It would mark anyone who

0:21:01.119 --> 0:21:04.680
<v Speaker 1>took a sign of well, this is proof that God exists.

0:21:04.880 --> 0:21:06.600
<v Speaker 1>But they would look at a meter study and they go,

0:21:06.640 --> 0:21:09.000
<v Speaker 1>well that's proof that AGI is coming. Yeah, and it's

0:21:09.040 --> 0:21:09.840
<v Speaker 1>so strange.

0:21:10.200 --> 0:21:12.760
<v Speaker 2>Yeah, it's like, hey, look at how complicated the structure

0:21:12.800 --> 0:21:17.040
<v Speaker 2>of the leaf is. It's Paaley intelligent design, like God exists.

0:21:17.040 --> 0:21:20.080
<v Speaker 2>Because of the argue from the design, they would mock that, right, Yeah,

0:21:20.119 --> 0:21:23.359
<v Speaker 2>now they're two seconds. Yeah, but like, hey, we can

0:21:23.480 --> 0:21:26.840
<v Speaker 2>draw a unicorn in ticks using like a GBT four prompt. Yeah,

0:21:26.920 --> 0:21:29.200
<v Speaker 2>the AI god is coming.

0:21:30.160 --> 0:21:33.280
<v Speaker 1>But it's it's so strange as well. I've been writing

0:21:33.280 --> 0:21:36.600
<v Speaker 1>about Silicon Valley bubble recently, this idea that basically the

0:21:36.640 --> 0:21:40.320
<v Speaker 1>valley itself is a bubble inflating that will burst going

0:21:40.359 --> 0:21:42.119
<v Speaker 1>back ten to fifteen years because they ran out of

0:21:42.160 --> 0:21:45.479
<v Speaker 1>new ideas, that they've not had anything good, not had

0:21:45.520 --> 0:21:47.280
<v Speaker 1>a new smartphone or a new app store or a

0:21:47.280 --> 0:21:49.440
<v Speaker 1>new cloud computing in a while. And I think that

0:21:49.440 --> 0:21:51.480
<v Speaker 1>that's what's fed into this as well, because you've got

0:21:51.480 --> 0:21:53.600
<v Speaker 1>that kind of desperation that we've run out of big

0:21:53.640 --> 0:22:01.040
<v Speaker 1>ideas and then added this layer of fearmongering plus childish philosophy. Ye,

0:22:01.359 --> 0:22:04.520
<v Speaker 1>very like, very much like this shit must have hit

0:22:04.600 --> 0:22:07.240
<v Speaker 1>so this would hit so hard if I was twelve. Yeah,

0:22:07.320 --> 0:22:10.400
<v Speaker 1>it goes to it like holy shit. Yeah, Oh I'm

0:22:10.440 --> 0:22:13.280
<v Speaker 1>using this computer today and tomorrow it will be ultra intelligent.

0:22:13.359 --> 0:22:18.200
<v Speaker 1>I'm so smart for finding this first. It's deeply sad.

0:22:18.240 --> 0:22:20.840
<v Speaker 1>I didn't think I could be more embarrassed by Silicon Valley,

0:22:21.240 --> 0:22:23.399
<v Speaker 1>but the fact that these people are being this weird.

0:22:23.520 --> 0:22:25.680
<v Speaker 1>Is that I like, we need to call them weird.

0:22:26.200 --> 0:22:28.760
<v Speaker 1>It's actually kind of unsettling as well. It's unsettling the

0:22:29.680 --> 0:22:31.800
<v Speaker 1>also that you are one of the few people just

0:22:31.840 --> 0:22:35.119
<v Speaker 1>saying it that everyone else is acting like this is

0:22:35.280 --> 0:22:39.080
<v Speaker 1>normal or rational or in any way logical.

0:22:39.320 --> 0:22:42.440
<v Speaker 2>Well, I'm trying to activate a bigger crowd here, right

0:22:42.520 --> 0:22:44.439
<v Speaker 2>and good. Here's what I think we need. I think

0:22:44.480 --> 0:22:46.720
<v Speaker 2>you're you're used to the word embarrassing is key because

0:22:46.800 --> 0:22:50.199
<v Speaker 2>I'm the trend I'm seeing which I don't like. Is

0:22:50.240 --> 0:22:54.080
<v Speaker 2>this idea of the people who instinctually do not like

0:22:54.480 --> 0:22:57.520
<v Speaker 2>what's going on with these AI companies are trying to

0:22:57.800 --> 0:23:01.040
<v Speaker 2>match the intensity and high of what they're saying, but

0:23:01.119 --> 0:23:06.239
<v Speaker 2>going the other direction, right, And so they're like they're

0:23:06.280 --> 0:23:11.680
<v Speaker 2>destroying everything, right, Like the environment will all be destroyed

0:23:11.680 --> 0:23:15.240
<v Speaker 2>by the data centers, and all education is going to end.

0:23:15.280 --> 0:23:17.560
<v Speaker 2>And the idea being is like, I need to be

0:23:17.600 --> 0:23:22.159
<v Speaker 2>as like as intense about all of the terribleness that

0:23:22.200 --> 0:23:24.320
<v Speaker 2>your technology will cause as you are about like the

0:23:24.400 --> 0:23:26.520
<v Speaker 2>good it will cause. Where I don't think that's playing

0:23:26.560 --> 0:23:30.440
<v Speaker 2>into their hands. I think the right reaction is ridicule

0:23:30.440 --> 0:23:33.439
<v Speaker 2>and embarrassment. This is where things need be if you

0:23:33.520 --> 0:23:37.400
<v Speaker 2>read anthropic white paper, you read these things, it's we're

0:23:37.400 --> 0:23:40.360
<v Speaker 2>now going to go through we envision four possible futures

0:23:40.359 --> 0:23:42.040
<v Speaker 2>for where this technology could come, and then they just

0:23:42.119 --> 0:23:44.880
<v Speaker 2>make up like four or twelve year old sci fi.

0:23:45.359 --> 0:23:48.840
<v Speaker 2>Maybe the technology will take over, but it will be

0:23:48.840 --> 0:23:51.840
<v Speaker 2>in more of a sort of a church Hillian mold,

0:23:51.920 --> 0:23:53.960
<v Speaker 2>and we'll wear top hats and that we could imagine

0:23:54.000 --> 0:23:56.560
<v Speaker 2>a council of humans, but they would probably be like

0:23:56.600 --> 0:23:59.520
<v Speaker 2>one third humans and two third robots. And you know,

0:23:59.600 --> 0:24:02.040
<v Speaker 2>like just like work, it's just it's just nonsense, right,

0:24:02.080 --> 0:24:05.080
<v Speaker 2>Like this's goofy. It's not it's not philosophy, it's not

0:24:05.359 --> 0:24:09.320
<v Speaker 2>it's not technology, it's not political science. It's just goofy. Right,

0:24:09.320 --> 0:24:10.600
<v Speaker 2>And I think that, and I said this in the

0:24:10.640 --> 0:24:13.520
<v Speaker 2>op ed. I said, the zeitgeist on this should could

0:24:13.560 --> 0:24:16.679
<v Speaker 2>change really quickly, right where it's like, all right, Dario,

0:24:17.200 --> 0:24:19.320
<v Speaker 2>is the agi in the room with you right now?

0:24:20.000 --> 0:24:22.199
<v Speaker 2>Can you see the agi in the room with your right?

0:24:22.359 --> 0:24:25.479
<v Speaker 2>Cause it's that's like, that's the reaction I think we

0:24:25.560 --> 0:24:30.680
<v Speaker 2>need plus forget that justify why I'm spending a thousand

0:24:30.680 --> 0:24:32.280
<v Speaker 2>dollars a month on tokens, Like I don't want to

0:24:32.320 --> 0:24:35.800
<v Speaker 2>hear your report about you know, hey Ford, I don't

0:24:35.800 --> 0:24:38.560
<v Speaker 2>need your report about the future in which we'll have

0:24:38.600 --> 0:24:42.560
<v Speaker 2>transformers that will help us police our streets. I am

0:24:42.600 --> 0:24:45.399
<v Speaker 2>spending this much money on gas for my f one fifty, Like,

0:24:45.440 --> 0:24:46.760
<v Speaker 2>tell me, like why that's worth it?

0:24:46.840 --> 0:24:47.000
<v Speaker 1>Right?

0:24:47.040 --> 0:24:49.920
<v Speaker 2>Like, just continually bring it back to you have really

0:24:49.960 --> 0:24:52.719
<v Speaker 2>three products you're selling right now. Great, sell a product,

0:24:52.760 --> 0:24:54.600
<v Speaker 2>it's interest, it's interesting technology, You're make it a good

0:24:54.600 --> 0:24:57.360
<v Speaker 2>product out of it. Sell me what are the benefits?

0:24:57.359 --> 0:24:59.680
<v Speaker 2>Why is it worth the money? And I will take

0:24:59.720 --> 0:25:01.760
<v Speaker 2>for and if you better say it's not going to

0:25:01.800 --> 0:25:04.159
<v Speaker 2>harm anyone, because otherwise fix it. So it doesn't like

0:25:04.160 --> 0:25:04.880
<v Speaker 2>that should.

0:25:04.560 --> 0:25:06.600
<v Speaker 1>Just be a yeah, that's your job to fix.

0:25:06.680 --> 0:25:08.320
<v Speaker 2>That's your job to fix, and you don't. And just

0:25:08.320 --> 0:25:09.959
<v Speaker 2>to throw in another moral argument I put in the

0:25:09.960 --> 0:25:13.159
<v Speaker 2>New York Times up ed it is not actually a

0:25:13.359 --> 0:25:16.159
<v Speaker 2>moral argument to say, there's this thing we want to

0:25:16.160 --> 0:25:18.200
<v Speaker 2>do that's going to cause potentially a lot of harm

0:25:18.200 --> 0:25:21.680
<v Speaker 2>and destruction. But that guy's going to do it, and

0:25:21.720 --> 0:25:23.399
<v Speaker 2>as long as he's doing it, I get to do

0:25:23.440 --> 0:25:26.720
<v Speaker 2>it too. That is what my eight year old says, right,

0:25:26.840 --> 0:25:29.720
<v Speaker 2>he's like, I know, but like he stole the cookies.

0:25:29.760 --> 0:25:31.359
<v Speaker 2>Why can't I steal it too? So I do not

0:25:31.480 --> 0:25:34.119
<v Speaker 2>understand this argument. It was in the latest anthropic paper.

0:25:35.000 --> 0:25:38.240
<v Speaker 2>We're building this technology, it might come alive and we

0:25:38.320 --> 0:25:40.720
<v Speaker 2>might lose control of it. But how can we stop

0:25:40.840 --> 0:25:43.600
<v Speaker 2>because China is also building the technology? And what if

0:25:43.640 --> 0:25:45.520
<v Speaker 2>they build it and it comes alive and comes control.

0:25:45.760 --> 0:25:47.480
<v Speaker 2>Why don't we get to do it too? What's the

0:25:47.520 --> 0:25:49.960
<v Speaker 2>moral argument there? This is not like a nuclear deterrence

0:25:50.040 --> 0:25:54.159
<v Speaker 2>argument that if you somehow destroy humanity first, that like

0:25:54.440 --> 0:25:55.520
<v Speaker 2>what's the advantage there?

0:25:55.560 --> 0:25:55.720
<v Speaker 1>Right?

0:25:55.760 --> 0:25:57.959
<v Speaker 2>Like what game have you want? So I don't understand that.

0:25:58.080 --> 0:25:59.719
<v Speaker 2>But my bigger point here is like, yeah, I think

0:25:59.760 --> 0:26:01.879
<v Speaker 2>we need to activate. I'm working on a big article

0:26:01.920 --> 0:26:05.720
<v Speaker 2>now for a trade magazine for engineers. Actually where I'm

0:26:05.760 --> 0:26:08.520
<v Speaker 2>making this case. It's a cred of cure. You need

0:26:08.520 --> 0:26:12.800
<v Speaker 2>to stand up and say that's nonsense. Let's talk about

0:26:12.840 --> 0:26:15.439
<v Speaker 2>your real technology. I know about this because I think

0:26:15.480 --> 0:26:17.240
<v Speaker 2>there's a lot of people like me who are on

0:26:17.280 --> 0:26:18.919
<v Speaker 2>the wrong coast. We're not on the West coast. We

0:26:18.920 --> 0:26:21.639
<v Speaker 2>don't we haven't gone to those cult meetings who are like,

0:26:21.680 --> 0:26:23.880
<v Speaker 2>there's interesting technology here, and I want to hear about

0:26:23.920 --> 0:26:25.720
<v Speaker 2>the products. And this is actually you and I maybe

0:26:25.720 --> 0:26:27.520
<v Speaker 2>even differ a little bit. I en that piece by

0:26:27.560 --> 0:26:30.919
<v Speaker 2>saying I'm pissed off because this should have been for

0:26:31.000 --> 0:26:33.840
<v Speaker 2>a nerd like me, an exciting time. This is a

0:26:33.880 --> 0:26:36.159
<v Speaker 2>cool technology. It's hard to make products out of it,

0:26:36.200 --> 0:26:38.280
<v Speaker 2>but we should have been like it, kind of excited about.

0:26:38.560 --> 0:26:40.679
<v Speaker 2>Oh I never thought about doing that with this. Oh

0:26:40.720 --> 0:26:44.480
<v Speaker 2>isn't that cool? What's going on? Instead? Everyone is terrified,

0:26:44.520 --> 0:26:47.560
<v Speaker 2>like you ruin this moment because of God knows why.

0:26:47.600 --> 0:26:49.000
<v Speaker 2>We don't even have to get into it. So I

0:26:49.000 --> 0:26:51.359
<v Speaker 2>think the right response from like actual engineers who are

0:26:51.400 --> 0:26:54.000
<v Speaker 2>not part of that cult. And you know, you got

0:26:54.000 --> 0:26:55.920
<v Speaker 2>to swallow your pride. I get it. You were really

0:26:55.960 --> 0:26:58.280
<v Speaker 2>excited to your cousin after you use claude code like

0:26:58.280 --> 0:27:00.600
<v Speaker 2>that's okay, swallow your pride a little bit and say

0:27:00.640 --> 0:27:04.440
<v Speaker 2>this is doom trolling and it's embarrassing. Can we talk

0:27:04.480 --> 0:27:07.160
<v Speaker 2>about the real technology, like a normal product. There's cool

0:27:07.160 --> 0:27:09.560
<v Speaker 2>things you can build. Some stuff isn't working. I want

0:27:09.560 --> 0:27:12.159
<v Speaker 2>to know if my stock market portfolio is going to crash, Like,

0:27:12.880 --> 0:27:17.640
<v Speaker 2>let's just be normal. Stop the weird stuff. Just stop it,

0:27:17.720 --> 0:27:20.199
<v Speaker 2>like you're a real company trying to sell real things.

0:27:20.640 --> 0:27:23.800
<v Speaker 2>Let's talk about products like why I want to spark

0:27:23.800 --> 0:27:25.760
<v Speaker 2>a revolution to bring things back to normal.

0:27:26.720 --> 0:27:29.040
<v Speaker 1>I fully agree, and I even said it was on

0:27:29.119 --> 0:27:32.320
<v Speaker 1>Chris Hayes his show a couple of weeks ago, and

0:27:32.400 --> 0:27:35.520
<v Speaker 1>I said, I think we should legally ban boosters from

0:27:35.600 --> 0:27:38.679
<v Speaker 1>using the future tents. I don't think they should be

0:27:38.680 --> 0:27:41.560
<v Speaker 1>allowed to say if either. I don't think they should

0:27:41.560 --> 0:27:44.280
<v Speaker 1>be able to say will. I think they should only

0:27:44.320 --> 0:27:46.520
<v Speaker 1>be allowed to talk about what's coming out today. You

0:27:46.520 --> 0:27:48.160
<v Speaker 1>can go two weeks in the future, Max.

0:27:48.440 --> 0:27:48.840
<v Speaker 2>Yeah.

0:27:49.000 --> 0:27:52.280
<v Speaker 1>I think that because when you read any story about

0:27:52.359 --> 0:27:55.080
<v Speaker 1>AI that's positive, it always speaks in the future tents.

0:27:55.080 --> 0:27:57.360
<v Speaker 1>It does not speak about what's happening now. And when

0:27:57.400 --> 0:27:59.520
<v Speaker 1>they do speak about what's happening now, it's the veguest

0:27:59.560 --> 0:28:01.919
<v Speaker 1>stuff in the world. It's like it can write or

0:28:02.080 --> 0:28:05.200
<v Speaker 1>it can autonomously write software. Is the software good? No?

0:28:05.359 --> 0:28:08.120
<v Speaker 1>Is it stable? No? Can you rely? Can you even

0:28:08.119 --> 0:28:09.720
<v Speaker 1>actually do anything with it if you don't know what

0:28:09.720 --> 0:28:12.959
<v Speaker 1>you're doing? No? Is the code overlea for Bose, Yes, Like,

0:28:13.040 --> 0:28:14.679
<v Speaker 1>it's like there are tons of problems with it, but

0:28:15.000 --> 0:28:19.960
<v Speaker 1>one sentence allows the generalities to grow. And also, I

0:28:20.000 --> 0:28:22.040
<v Speaker 1>think there's a very obvious reason they don't talk about

0:28:22.080 --> 0:28:24.439
<v Speaker 1>the present because it's kind of boring. It is like,

0:28:24.640 --> 0:28:28.719
<v Speaker 1>it's just in comparison to a big, sexy autonomous robot.

0:28:28.920 --> 0:28:33.320
<v Speaker 1>It's incredibly boring to say. But also the business doesn't work.

0:28:34.000 --> 0:28:36.240
<v Speaker 1>The business is like, the business doesn't function if they

0:28:36.240 --> 0:28:38.880
<v Speaker 1>have to talk about why it costs one thousand dollars

0:28:38.880 --> 0:28:41.480
<v Speaker 1>of API tokens to do something. They have to talk

0:28:41.480 --> 0:28:43.160
<v Speaker 1>about their business and they have to talk about what

0:28:43.200 --> 0:28:45.680
<v Speaker 1>it does today, and the problem is with that is

0:28:46.240 --> 0:28:48.800
<v Speaker 1>so my position on this is I don't mind on

0:28:48.920 --> 0:28:51.360
<v Speaker 1>device models. I actually think that's all that's going to

0:28:51.400 --> 0:28:54.440
<v Speaker 1>remain of LM's I think maybe we'll get some rag

0:28:54.480 --> 0:28:58.600
<v Speaker 1>search stuff here and there. But I don't have an

0:28:58.680 --> 0:29:02.400
<v Speaker 1>animous with the non data center ones. But I do

0:29:02.480 --> 0:29:06.480
<v Speaker 1>agree that there needs to be that people do need

0:29:06.520 --> 0:29:09.120
<v Speaker 1>to stop fighting fire with fire. As far as being

0:29:09.120 --> 0:29:11.520
<v Speaker 1>like if they're gonna dooma, we're gonna say that all

0:29:11.600 --> 0:29:14.240
<v Speaker 1>data centers are the worst, which they are. The water

0:29:14.320 --> 0:29:17.400
<v Speaker 1>story is murky, but the environmental story is very real,

0:29:17.520 --> 0:29:19.520
<v Speaker 1>especially with the behind the meat of gas turbines.

0:29:19.560 --> 0:29:21.800
<v Speaker 2>But and the economic story, which you've you've been doing

0:29:21.760 --> 0:29:24.840
<v Speaker 2>to is there's so much more money coming into building

0:29:24.840 --> 0:29:27.080
<v Speaker 2>these centers and their centers being built and that's going

0:29:27.160 --> 0:29:28.520
<v Speaker 2>to be its own is.

0:29:28.720 --> 0:29:32.520
<v Speaker 1>But also just just fundamentally, these weirdos are trying to

0:29:32.560 --> 0:29:35.720
<v Speaker 1>make you scared of cloud software. Like that's that's my

0:29:35.920 --> 0:29:38.040
<v Speaker 1>thing that I keep up. It's like if such an

0:29:38.040 --> 0:29:40.640
<v Speaker 1>adela every two weeks is like, You'll never fucking believe

0:29:40.640 --> 0:29:43.120
<v Speaker 1>what Microsoft Work's going to do next mate, You're never

0:29:43.120 --> 0:29:45.920
<v Speaker 1>gonna believe what powerbi is going to do with the dashboard.

0:29:46.080 --> 0:29:49.320
<v Speaker 1>I saw a dashboard on Powerbi and I'm pretty sure

0:29:49.360 --> 0:29:51.840
<v Speaker 1>next week it's gonna it's gonna try and kill someone.

0:29:52.120 --> 0:29:54.640
<v Speaker 1>You'd be like, you sound like a lunatic man. You

0:29:54.760 --> 0:29:59.080
<v Speaker 1>sound like a crazy person. This is an insane Sundhar Peshai. Yeah,

0:29:59.120 --> 0:30:03.800
<v Speaker 1>Google Search is oh any in any week now? You

0:30:04.000 --> 0:30:05.640
<v Speaker 1>just laugh at it. You would laugh at him.

0:30:05.680 --> 0:30:11.800
<v Speaker 2>But the two in the Microsoft guys are are he's

0:30:11.840 --> 0:30:14.840
<v Speaker 2>trying to because he's not He's trying to use the

0:30:14.960 --> 0:30:19.160
<v Speaker 2>Dario sam playbook, but he's not good at it, and

0:30:19.200 --> 0:30:23.440
<v Speaker 2>so it's kind of he's like, I think, uh, you know,

0:30:23.600 --> 0:30:26.440
<v Speaker 2>Ai is going to It's it's not only going to

0:30:26.480 --> 0:30:29.600
<v Speaker 2>take over all jobs, it's going to create new jobs

0:30:29.640 --> 0:30:33.520
<v Speaker 2>just to automate them, and it's going to do it Thursday.

0:30:33.720 --> 0:30:35.440
<v Speaker 2>He'll do a Thursday. How about that? Is that? Guys?

0:30:35.560 --> 0:30:36.160
<v Speaker 2>That is that?

0:30:36.200 --> 0:30:37.720
<v Speaker 1>I mean, it's like, do you like that? He's turning

0:30:37.760 --> 0:30:40.239
<v Speaker 1>a big dial, this is AI on it. It's just like,

0:30:40.280 --> 0:30:43.000
<v Speaker 1>I do you like this? I love it well because

0:30:43.240 --> 0:30:45.800
<v Speaker 1>he'll he'll even talk and be like, yeah, twelve to

0:30:45.840 --> 0:30:49.640
<v Speaker 1>eighteen months away, we're going to have super intelligence, but

0:30:49.720 --> 0:30:52.800
<v Speaker 1>also anthropics too expense if we can't pay the money more.

0:30:53.360 --> 0:30:55.760
<v Speaker 1>But also this is normal and it's not going to

0:30:55.800 --> 0:30:57.640
<v Speaker 1>take jobs, but it's going to take all the jobs

0:30:57.640 --> 0:30:59.960
<v Speaker 1>in two years, but not today, but not next week,

0:31:00.120 --> 0:31:03.000
<v Speaker 1>maybe in six months. And it's just the Yeah, they're

0:31:03.040 --> 0:31:07.000
<v Speaker 1>using the Dario playbook with everything because they're just making

0:31:07.040 --> 0:31:10.640
<v Speaker 1>the same things and they don't work particularly well. It's

0:31:10.680 --> 0:31:13.840
<v Speaker 1>just so these people are so weird. Now you've said it.

0:31:13.880 --> 0:31:15.120
<v Speaker 1>Now that's all I can think about.

0:31:15.360 --> 0:31:17.760
<v Speaker 2>So here's the logic. I also want people to avoid

0:31:17.760 --> 0:31:19.400
<v Speaker 2>it because I came up into what you were saying

0:31:19.440 --> 0:31:21.640
<v Speaker 2>there and I feel like maybe Chris did this a

0:31:21.640 --> 0:31:23.320
<v Speaker 2>little bit in your interview. I have to go back

0:31:23.360 --> 0:31:26.800
<v Speaker 2>and listen to it. But this this logic, and I

0:31:26.880 --> 0:31:29.000
<v Speaker 2>get it, and I'll give an explanation for it from

0:31:29.000 --> 0:31:31.400
<v Speaker 2>a psychological perspective in a second. But there's this logic

0:31:31.600 --> 0:31:35.280
<v Speaker 2>of if there's something I can point to where AI

0:31:35.440 --> 0:31:37.840
<v Speaker 2>did something that I thought was like impressive or interesting

0:31:37.880 --> 0:31:42.200
<v Speaker 2>or better than I thought, every prophecy is true, yes,

0:31:43.000 --> 0:31:46.440
<v Speaker 2>and I think I think it's I don't know psychological

0:31:46.480 --> 0:31:49.480
<v Speaker 2>self preservation or I don't know what it is. But

0:31:49.600 --> 0:31:51.720
<v Speaker 2>I don't under this. This is again from like a

0:31:51.800 --> 0:31:54.720
<v Speaker 2>logic point of view. I teach logic to undergraduates, like

0:31:54.760 --> 0:31:58.560
<v Speaker 2>propositional logic. I don't understand. This is a it's an

0:31:58.680 --> 0:32:02.520
<v Speaker 2>argument that's missing premises. Right, So they're going from the

0:32:02.560 --> 0:32:05.760
<v Speaker 2>premise of hey, it's and it's almost always the same thing.

0:32:06.080 --> 0:32:11.160
<v Speaker 2>It's almost always a non engineer using cloud code because

0:32:11.200 --> 0:32:14.400
<v Speaker 2>it's really fun. It's like a slot machine. It especially

0:32:14.640 --> 0:32:16.880
<v Speaker 2>your you know, it just like creates code. It works

0:32:17.280 --> 0:32:21.720
<v Speaker 2>for simple things, and you're like this, it's really fun.

0:32:21.920 --> 0:32:23.760
<v Speaker 2>Like I've talked I know people who are like they're

0:32:24.040 --> 0:32:25.760
<v Speaker 2>they're like, this is my new hobby. Like it's it's

0:32:25.800 --> 0:32:27.760
<v Speaker 2>really fun if you're into that type of thing. Right,

0:32:27.920 --> 0:32:31.120
<v Speaker 2>we talked about this earlier. It's like model trains for engineers.

0:32:31.440 --> 0:32:33.320
<v Speaker 2>You do that like that was really cool, and then

0:32:33.360 --> 0:32:35.760
<v Speaker 2>you translate that experience of like I really enjoyed that

0:32:36.360 --> 0:32:40.120
<v Speaker 2>to you know, I wonder what salt mind the AI

0:32:40.200 --> 0:32:42.640
<v Speaker 2>overlords are going to assign me to. And it's a

0:32:42.880 --> 0:32:47.880
<v Speaker 2>weird that that leap of logic. How can we just

0:32:47.960 --> 0:32:50.800
<v Speaker 2>keep it as like, Okay, this was really cool that

0:32:50.880 --> 0:32:53.440
<v Speaker 2>you built this dashboard. We don't have to leap from

0:32:53.440 --> 0:32:56.880
<v Speaker 2>that to therefore everything Dario Amadae said must be true.

0:32:56.920 --> 0:33:01.160
<v Speaker 2>It's not this binary balance on a folk rum where

0:33:01.160 --> 0:33:04.120
<v Speaker 2>we're just looking for like one one little piece of

0:33:04.120 --> 0:33:05.880
<v Speaker 2>evidence will tip it one way or the other. So

0:33:06.040 --> 0:33:08.440
<v Speaker 2>I think we need to be careful about these elaborations.

0:33:08.760 --> 0:33:12.120
<v Speaker 2>It's ad jagged technology. Some things make more progress than others.

0:33:12.120 --> 0:33:14.000
<v Speaker 2>Some things are cool, some things are more narrow than

0:33:14.840 --> 0:33:17.120
<v Speaker 2>you would think. But like, you know, does see I

0:33:18.160 --> 0:33:21.360
<v Speaker 2>built a really cool web based game in Fable five,

0:33:21.400 --> 0:33:23.360
<v Speaker 2>which actually is probably just copying the source code of

0:33:23.360 --> 0:33:26.320
<v Speaker 2>a similar game. Yeah, does that mean like recursive self

0:33:26.360 --> 0:33:28.920
<v Speaker 2>improvement is coming? So what I'm hoping we can do

0:33:29.240 --> 0:33:33.400
<v Speaker 2>is stop making these leaps of logic where if you

0:33:33.720 --> 0:33:36.840
<v Speaker 2>encounter any sort of feature or have any sort of

0:33:36.880 --> 0:33:39.920
<v Speaker 2>positive experience with an AI product, that you then leap

0:33:40.040 --> 0:33:43.960
<v Speaker 2>immediately to the conclusion that everything involved in the prophecy

0:33:44.000 --> 0:33:46.480
<v Speaker 2>about the future of AI must be true, or to

0:33:46.560 --> 0:33:48.520
<v Speaker 2>put someone like you and me now it has to

0:33:48.520 --> 0:33:50.800
<v Speaker 2>be put on the defense of of I did something

0:33:50.840 --> 0:33:52.240
<v Speaker 2>positive with AI, all right?

0:33:52.440 --> 0:33:52.480
<v Speaker 1>Ed?

0:33:53.080 --> 0:33:55.480
<v Speaker 2>Doesn't that now mean everything anyone has ever said about

0:33:55.520 --> 0:33:58.800
<v Speaker 2>AI is true? Convinced me I'm wrong. I think that

0:33:58.960 --> 0:34:03.120
<v Speaker 2>logic is back words. That these grand proclamations of we're

0:34:03.120 --> 0:34:05.880
<v Speaker 2>going to have self improving AI, we're going to automate

0:34:05.920 --> 0:34:10.759
<v Speaker 2>all jobs use to Carl Sagan quote, those remarkable claims

0:34:10.840 --> 0:34:13.960
<v Speaker 2>require a remarkable amount of evidence. Why is everyone so

0:34:14.040 --> 0:34:16.960
<v Speaker 2>quick of like this product, what you're working on and

0:34:17.000 --> 0:34:19.080
<v Speaker 2>putting billions of dollars on a course? You're going to

0:34:19.160 --> 0:34:21.600
<v Speaker 2>keep adding new features or finding things to improve? That

0:34:21.719 --> 0:34:24.000
<v Speaker 2>is a given a course? What technology in the history

0:34:24.000 --> 0:34:26.239
<v Speaker 2>of world would you have no improvements on? If you're

0:34:26.280 --> 0:34:28.879
<v Speaker 2>spending you know, ten to twenty billion dollars a year

0:34:28.880 --> 0:34:30.800
<v Speaker 2>in R and D, of course it's going to have improvements.

0:34:31.719 --> 0:34:33.000
<v Speaker 2>Why are we leaping from that?

0:34:33.000 --> 0:34:33.080
<v Speaker 1>That?

0:34:33.200 --> 0:34:36.800
<v Speaker 2>Like that means that anything you say, any speculative future

0:34:36.840 --> 0:34:39.520
<v Speaker 2>you give about, you know, some sci fi future, is

0:34:39.600 --> 0:34:42.200
<v Speaker 2>now the default. Unless we can prove otherwise, so we

0:34:42.480 --> 0:34:44.600
<v Speaker 2>got to stop that logic too. We should be able

0:34:44.640 --> 0:34:48.480
<v Speaker 2>to discuss, enjoy, or get around the implications of particular

0:34:48.520 --> 0:34:51.360
<v Speaker 2>AI features without having to have that shift to a

0:34:51.360 --> 0:34:54.720
<v Speaker 2>conversation about RSI or widespread job automation.

0:34:55.560 --> 0:34:57.239
<v Speaker 1>I think what it is. It comes back to the

0:34:57.280 --> 0:35:01.800
<v Speaker 1>pride thing. I think the the There's also something about

0:35:01.960 --> 0:35:06.279
<v Speaker 1>lms and the way they make people feel like I'm

0:35:06.320 --> 0:35:09.200
<v Speaker 1>on the spectrum autistic people have probably worked that out

0:35:09.200 --> 0:35:12.640
<v Speaker 1>from the third ten thousand word article. But I don't

0:35:12.640 --> 0:35:15.920
<v Speaker 1>feel anything when they're nice to me. I feel absolutely nothing.

0:35:16.080 --> 0:35:17.680
<v Speaker 1>When the chat bot is like, Wow, that was a

0:35:17.680 --> 0:35:20.640
<v Speaker 1>great idea. I don't have I don't respect it. I

0:35:20.680 --> 0:35:23.279
<v Speaker 1>don't care what it thinks because it doesn't think. I

0:35:23.320 --> 0:35:26.040
<v Speaker 1>don't see anything about what it says. If it tells

0:35:26.080 --> 0:35:28.399
<v Speaker 1>me I have a good idea, I'm like, I don't care.

0:35:28.480 --> 0:35:30.960
<v Speaker 1>I feel nothing. I think that a lot of people

0:35:31.120 --> 0:35:33.280
<v Speaker 1>do feel something. I think when they see these things

0:35:33.360 --> 0:35:38.319
<v Speaker 1>compliment them and do an impression of brainstorming with them,

0:35:38.360 --> 0:35:40.360
<v Speaker 1>which is really just talking to yourself in the mirror,

0:35:40.400 --> 0:35:44.560
<v Speaker 1>which can help. That can help. I've done it. I

0:35:44.640 --> 0:35:47.440
<v Speaker 1>think that it makes them feel good and also your

0:35:47.480 --> 0:35:51.640
<v Speaker 1>claud code example. I think that people feel very good

0:35:51.680 --> 0:35:53.960
<v Speaker 1>about that, and they feel very proficient, and it makes

0:35:54.000 --> 0:35:56.760
<v Speaker 1>them feel capable at something perhaps they were incapable of before.

0:35:57.800 --> 0:36:01.680
<v Speaker 1>And thus if you say to them, well, okay, that's

0:36:01.760 --> 0:36:04.600
<v Speaker 1>just that's fine, they go, no, no, I'm a big boy.

0:36:04.840 --> 0:36:07.520
<v Speaker 1>I think there's something kind of almost childish about it,

0:36:07.600 --> 0:36:12.480
<v Speaker 1>like no, I'm special, I'm special. Why aren't you impressed with?

0:36:13.400 --> 0:36:18.279
<v Speaker 1>And it's something that models are built to do. The ingratiation,

0:36:18.640 --> 0:36:22.279
<v Speaker 1>I think is important. But also I think that the

0:36:22.320 --> 0:36:24.319
<v Speaker 1>tech industry has done a really good job over the

0:36:24.360 --> 0:36:27.200
<v Speaker 1>years of convincing us that everything gets better and better

0:36:27.239 --> 0:36:29.560
<v Speaker 1>and better, even though it doesn't. Even though a lot

0:36:29.600 --> 0:36:33.360
<v Speaker 1>of Google Search has got worse, Facebook's got worse, Twitter's

0:36:33.400 --> 0:36:36.120
<v Speaker 1>got worse. Honestly, argue, many parts of Blue Sky have

0:36:36.200 --> 0:36:39.200
<v Speaker 1>got worse. It's less reliable. At least, the experience of

0:36:39.320 --> 0:36:41.640
<v Speaker 1>using the computer has got worse. We've hit a wall

0:36:41.640 --> 0:36:44.839
<v Speaker 1>with innovation in general. But this thing, because it can

0:36:44.920 --> 0:36:48.200
<v Speaker 1>kind of resemble anything if you want it to, I

0:36:48.280 --> 0:36:51.880
<v Speaker 1>think touches something dark in some people, not evil, but

0:36:51.960 --> 0:36:56.279
<v Speaker 1>just like something very vulnerable and kind of weak, kind

0:36:56.280 --> 0:36:59.319
<v Speaker 1>of gullible, where they say well, if I'm getting this

0:36:59.400 --> 0:37:02.359
<v Speaker 1>out of this, then I'm part of the future and

0:37:02.440 --> 0:37:04.480
<v Speaker 1>I need someone to be impressed. But it's this is

0:37:04.520 --> 0:37:07.239
<v Speaker 1>now a part of my identity, and that's I think

0:37:07.280 --> 0:37:09.719
<v Speaker 1>that that might also be the differentiator here as well.

0:37:10.320 --> 0:37:12.600
<v Speaker 1>I don't remember. Take sure you had four or wars

0:37:12.719 --> 0:37:15.880
<v Speaker 1>about gaming consoles. I'm sure you had, you have Android

0:37:15.920 --> 0:37:21.080
<v Speaker 1>versus iOS whatever, not like this, nothing like this, And

0:37:21.120 --> 0:37:23.400
<v Speaker 1>I think that it touches something in people.

0:37:23.680 --> 0:37:26.239
<v Speaker 2>Well, so I think there's a whole class of people,

0:37:26.320 --> 0:37:28.879
<v Speaker 2>or that's what's going on, especially the sort of bro

0:37:29.000 --> 0:37:32.279
<v Speaker 2>adjacent tech proadjacent people on you know X, who are

0:37:32.480 --> 0:37:34.600
<v Speaker 2>who are always yelling well, but what about this? What

0:37:34.640 --> 0:37:35.000
<v Speaker 2>about this?

0:37:35.160 --> 0:37:35.319
<v Speaker 1>Right?

0:37:36.120 --> 0:37:38.000
<v Speaker 2>And then I think the the other thing that happens

0:37:38.000 --> 0:37:39.479
<v Speaker 2>for a lot of people, especially who are a little

0:37:39.480 --> 0:37:42.680
<v Speaker 2>bit more tech distance, is it's a way of dealing

0:37:42.760 --> 0:37:45.680
<v Speaker 2>with anxiety. Right, Yeah, And if you think about this,

0:37:45.800 --> 0:37:47.839
<v Speaker 2>and like we've talked about COVID before, you and I,

0:37:47.920 --> 0:37:50.759
<v Speaker 2>but I think it offers some good you know, analogies

0:37:50.800 --> 0:37:54.000
<v Speaker 2>here is like one of the ways when you know

0:37:54.080 --> 0:37:56.560
<v Speaker 2>the virus was spreading but hadn't quite got to the US,

0:37:56.680 --> 0:37:58.640
<v Speaker 2>or it was just coming or the hospitals were overloaded

0:37:58.640 --> 0:38:00.440
<v Speaker 2>in Italy or whatever. It's like one of the ways

0:38:00.440 --> 0:38:02.360
<v Speaker 2>to deal with it psychologically was to get out in

0:38:02.400 --> 0:38:04.640
<v Speaker 2>front of it and be like, this is going to

0:38:04.719 --> 0:38:07.759
<v Speaker 2>be really bad, Like I you know, I'm on it,

0:38:08.160 --> 0:38:10.839
<v Speaker 2>and you know, you feel more prepared, like I'm not.

0:38:11.040 --> 0:38:12.719
<v Speaker 2>I'm not going to be caught off guard. I'll be

0:38:12.800 --> 0:38:15.239
<v Speaker 2>the one telling people how bad this is. And like

0:38:15.280 --> 0:38:17.400
<v Speaker 2>somehow that does help you a little bit deal with

0:38:17.440 --> 0:38:19.799
<v Speaker 2>like the anxiety of like impending dread. And I think

0:38:19.800 --> 0:38:21.480
<v Speaker 2>there's a lot of that going on with this as well,

0:38:21.680 --> 0:38:24.880
<v Speaker 2>is like I'm anxious about this doomtroll lamp and doomtrolled

0:38:24.920 --> 0:38:26.440
<v Speaker 2>so many times, I'm going to kind of get out

0:38:26.480 --> 0:38:28.200
<v Speaker 2>in front of it. Now if I'm the one saying,

0:38:28.560 --> 0:38:30.799
<v Speaker 2>well wait a second, I get so many I am

0:38:30.880 --> 0:38:34.319
<v Speaker 2>for a lot of people now, they're like stress pacifier

0:38:34.520 --> 0:38:37.840
<v Speaker 2>of like well what about this go right? Like I

0:38:37.880 --> 0:38:40.200
<v Speaker 2>saw this thing that upset me, Like doesn't that mean

0:38:40.239 --> 0:38:41.759
<v Speaker 2>that means? I think that means it is going to

0:38:41.800 --> 0:38:43.520
<v Speaker 2>take it over? And then they're kind of hoping that

0:38:43.960 --> 0:38:45.319
<v Speaker 2>me or someone else will come in and be like

0:38:45.800 --> 0:38:48.279
<v Speaker 2>you don't really it's okay, Like that's not as big

0:38:48.320 --> 0:38:49.759
<v Speaker 2>of a deal as it seems, where so I think

0:38:49.800 --> 0:38:52.480
<v Speaker 2>both are going on. It's either like this gives me

0:38:52.480 --> 0:38:54.719
<v Speaker 2>meaning like I'm a part of the future if you're

0:38:54.719 --> 0:38:56.880
<v Speaker 2>like using cloud code, and if you're just using like

0:38:56.960 --> 0:38:59.600
<v Speaker 2>chat GPT for natural language search, you might just be

0:39:00.200 --> 0:39:03.360
<v Speaker 2>really anxious because you know, you hear these earnest podcasts

0:39:03.360 --> 0:39:05.520
<v Speaker 2>where people are interviewing people about you know, we'll have

0:39:05.560 --> 0:39:08.120
<v Speaker 2>to eat our dogs when the jobs are gone, and

0:39:08.320 --> 0:39:10.520
<v Speaker 2>you're like, well, if I get out in front of it,

0:39:10.560 --> 0:39:12.799
<v Speaker 2>at least I feel like I have some autonomy in

0:39:12.880 --> 0:39:15.520
<v Speaker 2>this dark thing that's happening. And so it's like common psychology.

0:39:15.600 --> 0:39:20.080
<v Speaker 2>Both of the psychologies equal dollar and cents for a Again,

0:39:20.120 --> 0:39:22.640
<v Speaker 2>I keep coming back to remember is a vanishingly small

0:39:22.719 --> 0:39:26.040
<v Speaker 2>number of early investors who have large equity positions in

0:39:26.080 --> 0:39:28.799
<v Speaker 2>these companies. That is, who benefits from all of that.

0:39:28.960 --> 0:39:32.120
<v Speaker 2>They will benefit immensely because no matter what that psychology is,

0:39:32.640 --> 0:39:34.600
<v Speaker 2>this company is at the full crumb of the future,

0:39:35.280 --> 0:39:38.040
<v Speaker 2>and that's worthy of investing in, even if their revenue

0:39:38.080 --> 0:39:42.520
<v Speaker 2>story is murky, even if their product story is there's

0:39:42.560 --> 0:39:45.680
<v Speaker 2>no moat and it is I mean, I don't want

0:39:45.719 --> 0:39:48.640
<v Speaker 2>to get into the economic weeds. But this is like

0:39:48.719 --> 0:39:52.160
<v Speaker 2>a Clayton Christiansen fever dream what is happening with these

0:39:52.280 --> 0:39:55.839
<v Speaker 2>companies and their products. So he had this idea of,

0:39:56.400 --> 0:39:59.120
<v Speaker 2>you know, talking about disruption from below that the problem

0:39:59.200 --> 0:40:00.960
<v Speaker 2>is is you're working on high end version of a

0:40:00.960 --> 0:40:03.120
<v Speaker 2>product and you don't want to undermine your market, so

0:40:03.160 --> 0:40:05.080
<v Speaker 2>you try to keep it expensive and high end, and

0:40:05.120 --> 0:40:07.160
<v Speaker 2>then someone comes in from below with a different way

0:40:07.160 --> 0:40:09.200
<v Speaker 2>of doing it this much cheaper, and you're screwed because

0:40:09.239 --> 0:40:12.440
<v Speaker 2>you ignored that market. All of these use the use

0:40:12.480 --> 0:40:14.799
<v Speaker 2>cases that are useful now that people are spending money on,

0:40:14.840 --> 0:40:18.400
<v Speaker 2>like software development tools and natural language search. Here's the problem.

0:40:18.880 --> 0:40:21.920
<v Speaker 2>You don't need a ten trillion parameter language model to

0:40:21.960 --> 0:40:24.200
<v Speaker 2>do this, and this is why they're so these companies

0:40:24.239 --> 0:40:25.920
<v Speaker 2>are so screwed, and I think are trying to IPO

0:40:26.000 --> 0:40:28.680
<v Speaker 2>as fast as they can. Is that like someone's going

0:40:28.760 --> 0:40:31.560
<v Speaker 2>to come along and say, we're using a twenty billion

0:40:31.640 --> 0:40:34.960
<v Speaker 2>parameter open source model with a really smart coding harness

0:40:35.000 --> 0:40:36.680
<v Speaker 2>that we just coded ourselves on top of it and

0:40:36.680 --> 0:40:38.719
<v Speaker 2>tweaked for the last two years, and you know what,

0:40:38.840 --> 0:40:41.239
<v Speaker 2>people like it better than claud code. And it's a

0:40:41.360 --> 0:40:44.680
<v Speaker 2>hundredth of the price right or natural language search. You

0:40:44.719 --> 0:40:51.239
<v Speaker 2>don't need claud fable five to summarize Google search results right.

0:40:51.320 --> 0:40:53.960
<v Speaker 2>And the economics make no sense. But a small model could,

0:40:54.000 --> 0:40:56.200
<v Speaker 2>a non chip model could. That is why they're in

0:40:56.200 --> 0:40:57.200
<v Speaker 2>trouble economically.

0:41:12.560 --> 0:41:15.520
<v Speaker 1>I also think that there is and I do think

0:41:15.520 --> 0:41:18.160
<v Speaker 1>there is a deeply cynical thing here. Well, let me

0:41:18.200 --> 0:41:22.120
<v Speaker 1>reframe them. I think that people have to find ways

0:41:22.160 --> 0:41:26.080
<v Speaker 1>to rationalize capitalism. I think that the only way that

0:41:26.120 --> 0:41:28.680
<v Speaker 1>they could the amount of times I hear a week

0:41:28.760 --> 0:41:31.360
<v Speaker 1>of just like Amazon Web service is just like Uber's,

0:41:31.480 --> 0:41:34.960
<v Speaker 1>just like this. I have restated that point a hundred times.

0:41:35.000 --> 0:41:39.719
<v Speaker 1>People sometimes listen, people sometimes don't. But people need to

0:41:39.800 --> 0:41:44.000
<v Speaker 1>find a way, I think to rationalize the massive expenditures,

0:41:44.080 --> 0:41:48.320
<v Speaker 1>and the current products do not provide that rationalization. Nothing

0:41:48.360 --> 0:41:51.600
<v Speaker 1>about chat, GBT or Claude or clock code or fable

0:41:51.680 --> 0:41:56.400
<v Speaker 1>or whatever actually rationalizes this. Doesn't make it. A trillion

0:41:56.440 --> 0:41:59.360
<v Speaker 1>dollars is too much money for anything outside of like

0:42:00.239 --> 0:42:02.879
<v Speaker 1>healthcare and housing. In fact, probably could have taken care

0:42:02.880 --> 0:42:08.400
<v Speaker 1>of that really easily. But nevertheless, people see that and

0:42:08.480 --> 0:42:13.040
<v Speaker 1>they don't want to deal with the natural kind of

0:42:13.080 --> 0:42:16.280
<v Speaker 1>dissonance they get, which is wow, are all the companies

0:42:16.360 --> 0:42:18.200
<v Speaker 1>run by people who don't know what's going on, and

0:42:18.239 --> 0:42:20.320
<v Speaker 1>the answer is yes, Like, I don't think that there's

0:42:20.440 --> 0:42:23.040
<v Speaker 1>actually a grand conspiracy. Oh my god. The emails are

0:42:23.040 --> 0:42:25.239
<v Speaker 1>get every week, people being like, it's surveillance. It's a

0:42:25.280 --> 0:42:28.280
<v Speaker 1>secret surveillance state. It's secret. It's yeah, it's so secret.

0:42:28.320 --> 0:42:31.239
<v Speaker 1>It's on CNBC every day. Oh oh, they're all they're

0:42:31.239 --> 0:42:34.359
<v Speaker 1>doing secret data centers. Yeah, it's called Oracle, it's being

0:42:35.000 --> 0:42:37.120
<v Speaker 1>it's called Oracle. That's what Oracle does. Oracle has a

0:42:37.200 --> 0:42:39.480
<v Speaker 1>huge class. Like, that's very well known. What do you like?

0:42:40.520 --> 0:42:42.880
<v Speaker 1>I think people need there to be a conspiracy because

0:42:42.880 --> 0:42:45.040
<v Speaker 1>otherwise the answer is which is what I believe, which

0:42:45.040 --> 0:42:47.719
<v Speaker 1>is none of these companies had a plan. I think

0:42:47.800 --> 0:42:49.880
<v Speaker 1>Sam Mortman and Daria Amedy thought we're gonna throw a

0:42:49.880 --> 0:42:53.880
<v Speaker 1>bunch of money at this and then God will come out. Yeah, money, money,

0:42:54.120 --> 0:42:57.640
<v Speaker 1>so much money, and then they will have asi agi

0:42:57.800 --> 0:42:59.880
<v Speaker 1>whatever they call it. And then it happened, and then

0:43:00.040 --> 0:43:02.160
<v Speaker 1>ye didn't. They hit the diminishing returns at the end

0:43:02.200 --> 0:43:04.920
<v Speaker 1>twenty twenty four, and since then they've been kind of

0:43:05.160 --> 0:43:08.640
<v Speaker 1>harnessing non LLM stuff on top of LLMS in an

0:43:08.640 --> 0:43:10.160
<v Speaker 1>attempt to make them do stuff.

0:43:10.360 --> 0:43:12.880
<v Speaker 2>But that's been so successful. By the way, think about

0:43:13.080 --> 0:43:15.480
<v Speaker 2>the doom trolling. This is another advantage of it. They've

0:43:15.520 --> 0:43:19.479
<v Speaker 2>completely offfuscated exactly that point that you just made, which

0:43:19.520 --> 0:43:22.319
<v Speaker 2>is that pre training scaling again had We've talked about

0:43:22.320 --> 0:43:24.680
<v Speaker 2>this on the show, but has had diminishing returns, you know,

0:43:25.200 --> 0:43:28.120
<v Speaker 2>starting after twenty twenty four, and so all of the

0:43:28.160 --> 0:43:30.680
<v Speaker 2>improvements that have come since then has been post training

0:43:30.719 --> 0:43:33.960
<v Speaker 2>pre train models and building hand coding harnesses with more

0:43:34.000 --> 0:43:37.160
<v Speaker 2>and more special cases plugged into it to find more

0:43:37.200 --> 0:43:40.000
<v Speaker 2>places for them to be useful, which is a completely

0:43:40.000 --> 0:43:43.120
<v Speaker 2>different trajectory than twenty twenty three where they were like, no,

0:43:43.200 --> 0:43:45.160
<v Speaker 2>we'll just keep pre trained scaling these things till we

0:43:45.160 --> 0:43:48.160
<v Speaker 2>have AGI. And they've created a sense of like, well

0:43:48.160 --> 0:43:51.719
<v Speaker 2>AI has been drastically improving, hasn't it. And you said, well,

0:43:51.719 --> 0:43:53.480
<v Speaker 2>wait a second, Like if you told me in twenty

0:43:53.560 --> 0:43:55.360
<v Speaker 2>twenty two, you know when you're like, we're going to

0:43:55.400 --> 0:43:57.680
<v Speaker 2>scale our way to AGI, and our vision of AGI

0:43:57.960 --> 0:43:59.240
<v Speaker 2>is software development tools.

0:44:00.080 --> 0:44:03.920
<v Speaker 1>Yeah, Like it's just dreadfully boring and compact for.

0:44:04.000 --> 0:44:06.080
<v Speaker 2>Natural language Google. But you're like, but it's not just

0:44:06.120 --> 0:44:08.640
<v Speaker 2>like a feature that like Google would just kind of

0:44:08.800 --> 0:44:10.840
<v Speaker 2>otherwise would have just kept adding and tweaking in a

0:44:10.920 --> 0:44:13.160
<v Speaker 2>very like economically reasonable way.

0:44:13.160 --> 0:44:17.120
<v Speaker 1>I'm saying forever. It's like, yeah, I thought, like the

0:44:17.200 --> 0:44:20.799
<v Speaker 1>reason that people like LLLM search is pretty simple that

0:44:20.960 --> 0:44:23.960
<v Speaker 1>it responds to search how people have always used search,

0:44:24.000 --> 0:44:26.680
<v Speaker 1>which is questions you're like, hey, what about this and this.

0:44:27.120 --> 0:44:30.319
<v Speaker 1>I've used AI mode with Google maybe three times, and

0:44:30.400 --> 0:44:33.120
<v Speaker 1>I always use it as like a trawling expedition to say, like,

0:44:33.719 --> 0:44:36.360
<v Speaker 1>has anyone ever said anything like this? And then it

0:44:36.360 --> 0:44:38.440
<v Speaker 1>would give me a bunch of fake citations like just

0:44:38.520 --> 0:44:41.200
<v Speaker 1>that happens every time. But then I'll find one weird

0:44:41.239 --> 0:44:44.520
<v Speaker 1>PDF and It'll be like, oh cool, But isn't that

0:44:45.560 --> 0:44:48.240
<v Speaker 1>if you didn't if elms didn't exist, I would expect

0:44:48.400 --> 0:44:50.680
<v Speaker 1>Google Search to do that by now, like that would

0:44:50.800 --> 0:44:55.000
<v Speaker 1>that feels like a natural progression, But it's just normal

0:44:55.280 --> 0:44:57.560
<v Speaker 1>kind of boring. So I always say it's like, oh,

0:44:57.600 --> 0:44:59.719
<v Speaker 1>if they called it library models, it wouldn't be that

0:44:59.719 --> 0:45:01.719
<v Speaker 1>big deal, but it would be a more accurate, like

0:45:01.760 --> 0:45:03.479
<v Speaker 1>you just choose a regular name.

0:45:04.040 --> 0:45:04.120
<v Speaker 3>This.

0:45:04.400 --> 0:45:08.080
<v Speaker 1>People weren't freaking out about platform as a service tools,

0:45:08.520 --> 0:45:10.719
<v Speaker 1>then they hear all this bullshit about Kubernettes, which is

0:45:10.719 --> 0:45:13.440
<v Speaker 1>way more fun to say in large language models. People

0:45:14.239 --> 0:45:17.600
<v Speaker 1>have taken this. And the thing is, I don't think

0:45:17.640 --> 0:45:20.600
<v Speaker 1>that the strategy was always we're going to doom troll like.

0:45:20.600 --> 0:45:22.600
<v Speaker 1>I don't think they sat and thought about it very much.

0:45:23.160 --> 0:45:25.200
<v Speaker 1>I think they just kind of oath maxed their way.

0:45:25.080 --> 0:45:27.360
<v Speaker 1>They just bumbled into it and they're like, it's scary,

0:45:27.480 --> 0:45:29.919
<v Speaker 1>and then they saw how well it worked. I think

0:45:30.120 --> 0:45:33.200
<v Speaker 1>I genuinely think on an ethical level, it's one of

0:45:33.280 --> 0:45:37.000
<v Speaker 1>the most disgraceful things anyone's ever done technology Technologically. I

0:45:37.040 --> 0:45:40.560
<v Speaker 1>think it's just on top of the obfuscation of what

0:45:40.600 --> 0:45:43.840
<v Speaker 1>these things actually do, it's horrible. It's a it's a

0:45:44.040 --> 0:45:46.960
<v Speaker 1>cruel thing to do. I hear from people all the

0:45:47.000 --> 0:45:48.759
<v Speaker 1>times being like, am I going to lose my job

0:45:48.800 --> 0:45:51.400
<v Speaker 1>to this? I'm like, why is my boss telling me

0:45:51.440 --> 0:45:52.680
<v Speaker 1>I'm going to lose my job to this?

0:45:52.719 --> 0:45:53.040
<v Speaker 2>Why?

0:45:53.239 --> 0:45:56.040
<v Speaker 1>What is this? Why is it different? Why? Am And

0:45:56.239 --> 0:45:58.759
<v Speaker 1>I think when they hear all the doomerism and they

0:45:58.840 --> 0:46:00.640
<v Speaker 1>use the product and they use claw code and they go, oh,

0:46:00.640 --> 0:46:05.080
<v Speaker 1>it boffed out an open source CRM, they're like, oh god,

0:46:05.120 --> 0:46:07.600
<v Speaker 1>they're right. Imagine what someone who kind of knew what

0:46:07.600 --> 0:46:10.560
<v Speaker 1>they were doing could do. Ooh, I'm scared. Maybe I

0:46:10.600 --> 0:46:13.080
<v Speaker 1>don't think that. I think that they shouldn't be able

0:46:13.120 --> 0:46:15.399
<v Speaker 1>to sleep at night. I think it's an evil way

0:46:15.400 --> 0:46:15.840
<v Speaker 1>to market.

0:46:15.960 --> 0:46:17.759
<v Speaker 2>I would say the way I would think of I

0:46:17.760 --> 0:46:19.960
<v Speaker 2>put on like my like my Bentham hat. Here, right,

0:46:20.040 --> 0:46:23.880
<v Speaker 2>let's use like a utilitarian calculus the net harm that

0:46:23.960 --> 0:46:27.000
<v Speaker 2>doom trolling as a communication strategy is caused in terms

0:46:27.120 --> 0:46:31.239
<v Speaker 2>of mental health, right, just like people feeling bad, The

0:46:31.280 --> 0:46:35.000
<v Speaker 2>net harm caused by doom trolling far far, far, far

0:46:35.120 --> 0:46:39.640
<v Speaker 2>far exceeds any the grand total of benefits that this

0:46:39.719 --> 0:46:42.040
<v Speaker 2>AI technology is produced so far, Like, I don't see

0:46:42.040 --> 0:46:44.239
<v Speaker 2>any other way for that calculus to work out. You know,

0:46:44.239 --> 0:46:46.840
<v Speaker 2>I was just reading one of the mini papers that

0:46:46.880 --> 0:46:49.239
<v Speaker 2>are actually now you know, surveying and studying AI in

0:46:49.280 --> 0:46:52.359
<v Speaker 2>the workforce, and they they looked at this paper and

0:46:52.400 --> 0:46:55.120
<v Speaker 2>they asked people that these software developers, like, estimate how

0:46:55.160 --> 0:46:58.719
<v Speaker 2>much time you've saved or whatever, and it was, you know,

0:46:58.840 --> 0:47:02.960
<v Speaker 2>eleven hours a week we're saving. And then they asked them,

0:47:03.160 --> 0:47:06.160
<v Speaker 2>is your company performing better at all? Because this made

0:47:06.160 --> 0:47:08.480
<v Speaker 2>any impact on your bottom line? And it was like

0:47:09.080 --> 0:47:11.640
<v Speaker 2>a vanishingly small percent that was like, yeah, there's anything

0:47:11.640 --> 0:47:14.840
<v Speaker 2>different on the performance was that worth or making like

0:47:14.880 --> 0:47:19.080
<v Speaker 2>software developers maybe by some measures one point two times

0:47:19.400 --> 0:47:24.880
<v Speaker 2>more efficient, Yeah, worth causing. I think the negative toll

0:47:24.920 --> 0:47:28.000
<v Speaker 2>of this psychologically is like getting close to like COVID territory.

0:47:28.120 --> 0:47:34.399
<v Speaker 2>It's like population wide, you are stressing out everybody day

0:47:34.440 --> 0:47:37.120
<v Speaker 2>after day. They've made the last two years for anyone

0:47:37.120 --> 0:47:41.560
<v Speaker 2>who's following the news psychologically miserable. That is a real harm.

0:47:42.040 --> 0:47:44.720
<v Speaker 2>That is a real measurable harm. Like I really hate

0:47:44.719 --> 0:47:47.160
<v Speaker 2>that I have to deal with this like time and again.

0:47:48.280 --> 0:47:50.520
<v Speaker 2>It reminds me of having to deal with the pandemic,

0:47:50.600 --> 0:47:54.280
<v Speaker 2>which was like a like a psychologically fraught time because

0:47:54.320 --> 0:47:57.480
<v Speaker 2>every day you were being bombarded. But at this time,

0:47:58.040 --> 0:48:00.920
<v Speaker 2>you know, a, there's not a virus that's infecting a

0:48:00.960 --> 0:48:04.160
<v Speaker 2>lot of people, and be the people who are like

0:48:04.239 --> 0:48:06.680
<v Speaker 2>warning about all the doom. In this analogy, it would

0:48:06.680 --> 0:48:08.040
<v Speaker 2>be as if like they were the ones in the

0:48:08.120 --> 0:48:10.400
<v Speaker 2>lab doing the research that was producing the virus. Like

0:48:10.400 --> 0:48:13.000
<v Speaker 2>that's the craziest thing about it, right, I mean imagine

0:48:13.000 --> 0:48:17.160
<v Speaker 2>if in like twenty nineteen someone was like these viruses,

0:48:17.280 --> 0:48:20.960
<v Speaker 2>coronaviruses doing gaina function research could really really like infect

0:48:20.960 --> 0:48:22.879
<v Speaker 2>the whole world Like a course, they're going to get out.

0:48:22.880 --> 0:48:26.680
<v Speaker 2>We can't contain them. Oh, by the way, we're curiously

0:48:26.840 --> 0:48:29.680
<v Speaker 2>doing this gain of function research on these viruses, and

0:48:29.680 --> 0:48:31.640
<v Speaker 2>we're not going to stop because you know the Chinese

0:48:31.680 --> 0:48:35.520
<v Speaker 2>well in this case, but whatever, yeah, someone else is direction,

0:48:36.040 --> 0:48:37.960
<v Speaker 2>someone else might do it. So like what can we do,

0:48:38.280 --> 0:48:41.960
<v Speaker 2>like stop building the viruses then? Anyways, So I think

0:48:42.520 --> 0:48:44.600
<v Speaker 2>we have to account for the net psychological harm, and

0:48:44.600 --> 0:48:46.440
<v Speaker 2>I think, I mean, I don't think our current government

0:48:46.480 --> 0:48:48.160
<v Speaker 2>can do this, but I think that that's something where

0:48:48.160 --> 0:48:52.600
<v Speaker 2>the government should, in an abstract world be involved, Like, oh,

0:48:52.800 --> 0:48:56.120
<v Speaker 2>there's a small number of people who are causing a

0:48:56.239 --> 0:48:58.680
<v Speaker 2>huge amount of distress and anxiety around our entire country.

0:48:58.719 --> 0:49:00.759
<v Speaker 2>Of course we're going to get in the game and

0:49:00.800 --> 0:49:02.960
<v Speaker 2>we're going to pause all your model deployments and tell

0:49:03.120 --> 0:49:06.759
<v Speaker 2>like you either convince us that that's not the case

0:49:06.800 --> 0:49:08.319
<v Speaker 2>and you admit like, yeah, we were kind of making

0:49:08.320 --> 0:49:09.879
<v Speaker 2>that up, or if you don't want to say thathing

0:49:09.960 --> 0:49:12.440
<v Speaker 2>like no, you don't get a release's product. You're causing

0:49:12.640 --> 0:49:18.520
<v Speaker 2>mass pro mass unhappiness and damage right to our entire country.

0:49:18.640 --> 0:49:20.920
<v Speaker 2>You have to account for that. You don't get to

0:49:21.000 --> 0:49:23.520
<v Speaker 2>just do that. So does I think the harms? So

0:49:23.680 --> 0:49:28.839
<v Speaker 2>far have exponentially outweighed the tangible benefits.

0:49:29.160 --> 0:49:31.720
<v Speaker 1>I'm just also going to be honest, Yeah, the model

0:49:31.800 --> 0:49:34.319
<v Speaker 1>AI companies are terrible. I also think the media is

0:49:34.360 --> 0:49:37.040
<v Speaker 1>categorically failed here and should feel fucking ashamed of themselves.

0:49:37.120 --> 0:49:38.680
<v Speaker 1>I think the media is as much.

0:49:38.520 --> 0:49:41.200
<v Speaker 2>To blame because I think I'm gonna put it. I'm

0:49:41.200 --> 0:49:43.160
<v Speaker 2>gonna put a fundamental blame on venture capital. I'll give

0:49:43.160 --> 0:49:45.680
<v Speaker 2>you my quick pitch on that. Yes, it's because, I

0:49:45.680 --> 0:49:48.719
<v Speaker 2>mean you have to understand the economics of especially high

0:49:48.719 --> 0:49:51.920
<v Speaker 2>tech venture capital coming out of Silicon Valley. The whole model,

0:49:51.960 --> 0:49:53.799
<v Speaker 2>and again this is all just because you have, like

0:49:53.840 --> 0:49:55.960
<v Speaker 2>some institutional investors in rich people that you want to

0:49:55.960 --> 0:49:58.839
<v Speaker 2>make Richard. That's it, Like, that's the whole underlying goal, right,

0:49:59.360 --> 0:50:02.759
<v Speaker 2>And the their whole model is we have to find

0:50:03.320 --> 0:50:08.200
<v Speaker 2>new areas where there's something to be monetized, in which

0:50:08.200 --> 0:50:10.440
<v Speaker 2>there's possibilities for unicorns. So we have to find new

0:50:10.520 --> 0:50:12.880
<v Speaker 2>territories where you have the possibility of getting in early

0:50:13.400 --> 0:50:15.600
<v Speaker 2>and having an investment turn out to be a multi

0:50:15.600 --> 0:50:16.480
<v Speaker 2>billion dollar company.

0:50:16.600 --> 0:50:16.759
<v Speaker 1>Right.

0:50:17.360 --> 0:50:21.400
<v Speaker 2>The web that's what started in the nineties with the web,

0:50:21.560 --> 0:50:23.319
<v Speaker 2>and most of those bets failed because of the first

0:50:23.320 --> 0:50:26.160
<v Speaker 2>dot Com crash, a couple paid off like Amazon. Then

0:50:26.200 --> 0:50:28.840
<v Speaker 2>you had Web two. It turned out attention was a

0:50:28.880 --> 0:50:31.000
<v Speaker 2>resource you could monetize, so it's like discovering a new

0:50:31.160 --> 0:50:33.759
<v Speaker 2>minroller oil. There was a huge fortune to be had.

0:50:33.760 --> 0:50:35.879
<v Speaker 2>The venture capitals could get in there early. They had

0:50:35.920 --> 0:50:39.640
<v Speaker 2>massive returns betting on meta, betting on you know, the

0:50:39.719 --> 0:50:42.440
<v Speaker 2>social media companies. Right then they're like, we need another

0:50:42.480 --> 0:50:44.960
<v Speaker 2>because you have to keep finding every time you raise

0:50:44.960 --> 0:50:47.640
<v Speaker 2>a new fund, you need a giant new resource that

0:50:47.680 --> 0:50:49.239
<v Speaker 2>you can monetize to get to early. So the next

0:50:49.280 --> 0:50:52.839
<v Speaker 2>thing they tried was Crypto. Crypto was just you know,

0:50:52.920 --> 0:50:57.520
<v Speaker 2>Andreessen Horowitz saying, there's no more unicorns left in social

0:50:58.000 --> 0:51:00.040
<v Speaker 2>We're going to try to create a new field in

0:51:00.080 --> 0:51:02.680
<v Speaker 2>which we can create unicorns. And they really flogged crypto.

0:51:02.760 --> 0:51:05.000
<v Speaker 2>This is the future, This is the future. Yeah, yeah,

0:51:05.040 --> 0:51:07.279
<v Speaker 2>and it wasn't, especially for software. And I got yelled

0:51:07.280 --> 0:51:09.759
<v Speaker 2>at for saying this, but I'm literally an expert on this.

0:51:10.160 --> 0:51:12.319
<v Speaker 2>My doctorate's in the theory of distributed systems. It's a

0:51:12.360 --> 0:51:14.759
<v Speaker 2>stupid way to build a distributed system, and I was right.

0:51:14.800 --> 0:51:16.839
<v Speaker 2>And then no one built the systems that way, so

0:51:16.880 --> 0:51:20.440
<v Speaker 2>that kind of failed. This was the next territory. And

0:51:20.480 --> 0:51:23.240
<v Speaker 2>it doesn't say this technology is not interesting or potentially useful,

0:51:23.239 --> 0:51:25.960
<v Speaker 2>but as you said, it's not trillion dollar useful, not yet.

0:51:26.480 --> 0:51:28.760
<v Speaker 2>So why did all that money come in is because

0:51:30.200 --> 0:51:33.160
<v Speaker 2>we're talking about one street in one city. We're talking

0:51:33.200 --> 0:51:35.560
<v Speaker 2>like sand Hill Road in Palo Alto. Like the people

0:51:35.600 --> 0:51:37.960
<v Speaker 2>on that one street are like, we need to find

0:51:38.000 --> 0:51:40.240
<v Speaker 2>a new territory that can produce, you know, one hundred

0:51:40.239 --> 0:51:42.280
<v Speaker 2>billion dollar companies so that we can get the returns

0:51:42.320 --> 0:51:44.279
<v Speaker 2>that we expect to keep our lifestyle up it And.

0:51:44.320 --> 0:51:46.520
<v Speaker 1>That's just what I've I've been saying this for years.

0:51:46.520 --> 0:51:49.319
<v Speaker 1>It's the rock combobbles. They're out of big ideas, and

0:51:49.440 --> 0:51:52.480
<v Speaker 1>AI was the panacea. It was the It was the

0:51:52.560 --> 0:51:55.040
<v Speaker 1>API would be the way to build new startups, be

0:51:55.120 --> 0:51:57.640
<v Speaker 1>the way for enterprise companies to sell more software with

0:51:57.719 --> 0:52:00.800
<v Speaker 1>new add ons. It would be the way for business

0:52:00.800 --> 0:52:03.000
<v Speaker 1>software to take off consumer software. It was meant to

0:52:03.000 --> 0:52:06.480
<v Speaker 1>be the solution because they're doing exactly the same thing

0:52:06.560 --> 0:52:09.240
<v Speaker 1>is that AI companies are doing, saying yeah, the AI

0:52:09.320 --> 0:52:11.560
<v Speaker 1>will just fix this, what are we going to build

0:52:11.560 --> 0:52:13.720
<v Speaker 1>with this? Or the AI will work out what's the product?

0:52:13.760 --> 0:52:16.680
<v Speaker 1>AI will work it out. It's honestly a kind of

0:52:16.719 --> 0:52:18.160
<v Speaker 1>a pathetic culture.

0:52:18.360 --> 0:52:20.759
<v Speaker 2>I don't think they even care about that. I think, like, okay,

0:52:20.800 --> 0:52:23.200
<v Speaker 2>if we put the money in this, now, what's the

0:52:23.239 --> 0:52:24.600
<v Speaker 2>strike price going to be at the IPO.

0:52:25.600 --> 0:52:27.399
<v Speaker 1>That's whine I'm getting at that.

0:52:28.400 --> 0:52:30.120
<v Speaker 2>Works from the end or not they're they're like, we

0:52:30.480 --> 0:52:31.759
<v Speaker 2>want a SpaceX, look at that.

0:52:31.880 --> 0:52:35.280
<v Speaker 1>But the theoretical that I'm saying is that the idea

0:52:35.320 --> 0:52:37.800
<v Speaker 1>of large language models as an API or a consumer

0:52:37.880 --> 0:52:39.879
<v Speaker 1>product was what they were excited about. Because you can

0:52:39.920 --> 0:52:43.480
<v Speaker 1>just extrapolate the various things. But I must, I must

0:52:43.560 --> 0:52:46.239
<v Speaker 1>get back to it. I think the media is responsible.

0:52:46.760 --> 0:52:49.160
<v Speaker 1>I one hundred percent believe the media is responsible. They

0:52:49.160 --> 0:52:52.040
<v Speaker 1>are the ones that at the Times Ezra Cline, how

0:52:52.120 --> 0:52:55.080
<v Speaker 1>much AGI bullshit is that guy shoveled? How many time

0:52:55.239 --> 0:52:58.560
<v Speaker 1>Bernie Sanders going out there and talking to Claude. I

0:52:58.600 --> 0:53:01.320
<v Speaker 1>realized that's politics. But they put on YouTube for a reason.

0:53:01.640 --> 0:53:04.239
<v Speaker 1>How many different outlets have talked about the job apocalypse

0:53:04.239 --> 0:53:08.280
<v Speaker 1>that does not exist or the saspocalypse claiming that software

0:53:08.320 --> 0:53:11.880
<v Speaker 1>companies are being sold off because of AI disruption that

0:53:11.920 --> 0:53:16.280
<v Speaker 1>doesn't exist. The media bought and sold this. In twenty

0:53:16.320 --> 0:53:18.719
<v Speaker 1>twenty three, there was Guardian headline that said Samuel was

0:53:18.760 --> 0:53:23.799
<v Speaker 1>a little bit scared. It's just without that, none of

0:53:23.840 --> 0:53:26.560
<v Speaker 1>this was possible. And perhaps they were just being a

0:53:26.600 --> 0:53:30.400
<v Speaker 1>megaphone for real anxieties from these companies. Well, perhaps they

0:53:30.400 --> 0:53:33.680
<v Speaker 1>were just looking for a narrative that would get clicks.

0:53:34.239 --> 0:53:37.160
<v Speaker 2>I just think, yeah, I don't think it's clix. I

0:53:37.280 --> 0:53:42.279
<v Speaker 2>think they underestimated the weirdness of these people, and they

0:53:42.280 --> 0:53:44.640
<v Speaker 2>took them at phase. This is my you know, having

0:53:44.840 --> 0:53:47.040
<v Speaker 2>mini feet within the world of media. What I really

0:53:47.040 --> 0:53:51.799
<v Speaker 2>think happened is it didn't cross their minds that these

0:53:51.800 --> 0:53:57.120
<v Speaker 2>people are super strange cultist weirdos. They don't know about Yukowski,

0:53:57.160 --> 0:53:59.239
<v Speaker 2>they don't know about the rationalists. They don't know about

0:53:59.280 --> 0:54:02.040
<v Speaker 2>the excescummuit that came out of the rationalists and then

0:54:02.080 --> 0:54:04.839
<v Speaker 2>merged with the technological singularities, and that they also think

0:54:04.840 --> 0:54:07.840
<v Speaker 2>they're going to live forever by harvesting the blood of

0:54:07.920 --> 0:54:10.640
<v Speaker 2>children and all. They were just like, these are engineers.

0:54:11.080 --> 0:54:12.239
<v Speaker 2>I don't know about technology.

0:54:12.320 --> 0:54:12.560
<v Speaker 1>They do.

0:54:12.640 --> 0:54:15.839
<v Speaker 2>They're telling me this. Why would they lie? I've heard

0:54:15.880 --> 0:54:20.000
<v Speaker 2>that word for word from multiple early on, multiple journalists,

0:54:20.239 --> 0:54:22.040
<v Speaker 2>and I think it was all being seen through the

0:54:22.040 --> 0:54:25.080
<v Speaker 2>frame of COVID, where that was in the recent history

0:54:25.560 --> 0:54:27.480
<v Speaker 2>if the viral I don't know anything about viruses, but

0:54:27.520 --> 0:54:29.719
<v Speaker 2>the virologists were like trusted, Like this is something I

0:54:29.760 --> 0:54:33.719
<v Speaker 2>can't directly observe right because there's not very many people sick,

0:54:33.719 --> 0:54:36.160
<v Speaker 2>but like trust us, this is going to spread and

0:54:36.160 --> 0:54:37.960
<v Speaker 2>this is a really big deal. And like all trucks,

0:54:38.239 --> 0:54:40.279
<v Speaker 2>the virologists know what they talk about. Why would they lie?

0:54:40.840 --> 0:54:42.239
<v Speaker 2>And I don't want to miss that story. I think

0:54:42.280 --> 0:54:45.759
<v Speaker 2>there was a lot of that going on. Of these

0:54:45.760 --> 0:54:47.880
<v Speaker 2>are the people building the technic because I heard this

0:54:47.960 --> 0:54:50.480
<v Speaker 2>phrase from so many people. These are the people building

0:54:50.560 --> 0:54:53.759
<v Speaker 2>the technology. They know about it best. I'm not a

0:54:53.800 --> 0:54:57.000
<v Speaker 2>computer scientist. I mean I am, but if I'm being

0:54:57.040 --> 0:54:59.359
<v Speaker 2>the journalist, and this is why, like the people who

0:54:59.360 --> 0:55:02.680
<v Speaker 2>were not, where is the resistance is coming from is

0:55:02.760 --> 0:55:05.640
<v Speaker 2>computer scientists like me who don't live near Silicon Valley,

0:55:06.080 --> 0:55:08.560
<v Speaker 2>And well, I do understand this technology and what the

0:55:08.560 --> 0:55:10.440
<v Speaker 2>hell are you talking about? But I think for the

0:55:10.520 --> 0:55:13.080
<v Speaker 2>journalist it was and now they're just and this is

0:55:13.080 --> 0:55:15.520
<v Speaker 2>why I wrote that off ed. I'm trying to spread

0:55:15.560 --> 0:55:19.719
<v Speaker 2>the word more. These guys are weird. They're weird, and

0:55:19.840 --> 0:55:21.359
<v Speaker 2>by the way, they're mainly guys too, because I think

0:55:21.360 --> 0:55:22.960
<v Speaker 2>women were gonna put up with this weird You know,

0:55:23.000 --> 0:55:24.239
<v Speaker 2>I'm gonna go in my room and put with my

0:55:24.280 --> 0:55:27.839
<v Speaker 2>Star Wars toys nonsense, right, I don't want to Who

0:55:27.920 --> 0:55:30.520
<v Speaker 2>knows what that tells us about our our gender. But

0:55:30.520 --> 0:55:33.280
<v Speaker 2>it's not great that it's all men doing this craziness.

0:55:33.760 --> 0:55:37.040
<v Speaker 1>It's tim Tim ree Gamber who isn't particularly sold on this,

0:55:37.160 --> 0:55:38.319
<v Speaker 1>for example, like.

0:55:39.040 --> 0:55:42.160
<v Speaker 2>Emily Bender, like a lot of the hardcore skeptics are

0:55:42.200 --> 0:55:44.400
<v Speaker 2>like you boys with.

0:55:44.280 --> 0:55:46.080
<v Speaker 1>Your Star Wars.

0:55:46.160 --> 0:55:50.480
<v Speaker 2>Yeah, you don't have friends, do you? Is this is

0:55:50.520 --> 0:55:51.880
<v Speaker 2>this related the fantasy football?

0:55:52.040 --> 0:55:52.160
<v Speaker 1>What?

0:55:52.560 --> 0:55:53.319
<v Speaker 2>Guys? What is this?

0:55:53.400 --> 0:55:55.960
<v Speaker 1>Right? Which? What's the funny thing is, though there is

0:55:56.000 --> 0:56:00.080
<v Speaker 1>something to that where it's is a very male technology,

0:56:00.320 --> 0:56:02.560
<v Speaker 1>and I do wonder like it's not something I think

0:56:02.600 --> 0:56:05.680
<v Speaker 1>two white guys can necessarily get into in any any

0:56:05.719 --> 0:56:08.880
<v Speaker 1>depth without a woman present, so that we could actually

0:56:08.960 --> 0:56:12.120
<v Speaker 1>have a rounded conversation. But it is interesting that pretty

0:56:12.160 --> 0:56:15.319
<v Speaker 1>much everyone running this stuff is a guy, and the

0:56:15.360 --> 0:56:19.440
<v Speaker 1>majority of them are white fellas, and they're all claiming

0:56:19.520 --> 0:56:22.240
<v Speaker 1>that everything they build is amazing. I don't know, perhaps

0:56:22.239 --> 0:56:25.279
<v Speaker 1>there's something going on that too, It's just.

0:56:25.480 --> 0:56:28.520
<v Speaker 2>Is there something Yeah, white guys are good at? Is

0:56:28.560 --> 0:56:30.839
<v Speaker 2>thinking everything we do is amazing. We are pretty well

0:56:30.880 --> 0:56:33.680
<v Speaker 2>trained or we are pretty well trained at that. But

0:56:33.840 --> 0:56:35.480
<v Speaker 2>that's why I do think it's important why I wrote

0:56:35.480 --> 0:56:36.799
<v Speaker 2>that op ed. It's like, I want to name what

0:56:36.800 --> 0:56:38.719
<v Speaker 2>they're doing. This is doom trolley. I think you need

0:56:38.719 --> 0:56:41.359
<v Speaker 2>to understand these people are weird and we can't take

0:56:41.400 --> 0:56:43.520
<v Speaker 2>what they're saying at face value. We just can't because

0:56:43.520 --> 0:56:46.719
<v Speaker 2>they're too weird. They've they've proven time and again. But

0:56:46.760 --> 0:56:48.440
<v Speaker 2>the key message I would also want like, if you're

0:56:48.440 --> 0:56:50.880
<v Speaker 2>listening to this pod, if you're listening to this podcast, now,

0:56:51.680 --> 0:56:55.640
<v Speaker 2>do not interpret this conversation as I'm going to categorize

0:56:55.640 --> 0:56:59.280
<v Speaker 2>this in a claim that AI doesn't work. And therefore,

0:57:00.040 --> 0:57:03.239
<v Speaker 2>if I see something cool that AI does, Uh, this

0:57:03.280 --> 0:57:07.239
<v Speaker 2>conversation isn't relevant. You know, cal was wrong. This is

0:57:07.320 --> 0:57:09.560
<v Speaker 2>it's really not about that is about what I'm talking

0:57:09.600 --> 0:57:12.600
<v Speaker 2>about right now. I am talking about the way that

0:57:12.640 --> 0:57:16.360
<v Speaker 2>we actually talk about AI, how we treat it as

0:57:16.400 --> 0:57:19.800
<v Speaker 2>a business, what is morally appropriate and what is morally

0:57:19.880 --> 0:57:24.680
<v Speaker 2>not And these are normal products and I don't I

0:57:24.680 --> 0:57:28.160
<v Speaker 2>don't to me, this isn't this not vaporware. I mean,

0:57:28.200 --> 0:57:30.520
<v Speaker 2>I think the ability to parse natural language like a

0:57:30.560 --> 0:57:34.000
<v Speaker 2>language model can do is like it's fascinating, and I'm

0:57:34.040 --> 0:57:37.760
<v Speaker 2>also surprised by how hard it's been to come up

0:57:37.760 --> 0:57:40.000
<v Speaker 2>with a lot of good, useful, economically viable products off

0:57:40.000 --> 0:57:42.520
<v Speaker 2>of that ability. Like we were still with software development

0:57:42.560 --> 0:57:45.480
<v Speaker 2>tools and and uh, you know, rewriting Google search results.

0:57:45.640 --> 0:57:48.440
<v Speaker 2>I actually thought we would have a lot more interesting

0:57:48.520 --> 0:57:50.640
<v Speaker 2>tools that we were using in our everyday life right now.

0:57:50.640 --> 0:57:52.760
<v Speaker 2>And because there's been some problems that we didn't think about.

0:57:53.120 --> 0:57:57.720
<v Speaker 2>So this whole conversation is about stop treating this like

0:57:57.960 --> 0:58:01.600
<v Speaker 2>a demon that's been summoned. Make fun of doom trolling

0:58:02.240 --> 0:58:05.040
<v Speaker 2>because it's kind of embarrassing. This is the you know,

0:58:05.120 --> 0:58:07.320
<v Speaker 2>this is the magic the gathering club at the high

0:58:07.360 --> 0:58:10.520
<v Speaker 2>school table, and just treat this like a normal product.

0:58:10.520 --> 0:58:12.680
<v Speaker 2>And like I don't want to hear. Let's just what

0:58:12.760 --> 0:58:14.680
<v Speaker 2>are the products you're selling, what are the benefits, how

0:58:14.760 --> 0:58:18.400
<v Speaker 2>much do they cost? Is it worth the money? And uh,

0:58:18.440 --> 0:58:20.120
<v Speaker 2>what are the products are you working on? Like, let's

0:58:20.120 --> 0:58:22.080
<v Speaker 2>just have normal conversation. So so do not as a

0:58:22.120 --> 0:58:24.800
<v Speaker 2>listener categorize this in this in this, in this sort

0:58:24.800 --> 0:58:27.760
<v Speaker 2>of light vers dark is like, okay, this means AI

0:58:27.880 --> 0:58:30.600
<v Speaker 2>doesn't work. And this team says it does work, and

0:58:30.640 --> 0:58:32.280
<v Speaker 2>if if and then I'm going to try to assess

0:58:32.320 --> 0:58:34.160
<v Speaker 2>in my mind is AI useful or not? And if

0:58:34.160 --> 0:58:36.760
<v Speaker 2>it is, then I have to be ready to mate

0:58:36.760 --> 0:58:39.760
<v Speaker 2>with machines. And if it doesn't, then, you know, stop

0:58:39.800 --> 0:58:42.800
<v Speaker 2>all that sort of extreme thinking. I just want us

0:58:42.840 --> 0:58:46.280
<v Speaker 2>to talk about this like a normal, interesting but normal technology,

0:58:47.120 --> 0:58:48.960
<v Speaker 2>and normal people should look at the way that these

0:58:49.000 --> 0:58:53.200
<v Speaker 2>companies are doom trolling us and say that's strange. Please stop.

0:58:53.240 --> 0:58:55.440
<v Speaker 2>And in the meantime, I am going to have to

0:58:55.480 --> 0:58:57.200
<v Speaker 2>really make fun of you, like I think that's where

0:58:57.200 --> 0:58:57.600
<v Speaker 2>we need to be.

0:58:58.960 --> 0:59:01.640
<v Speaker 1>And I think all of the large language models in

0:59:01.680 --> 0:59:05.800
<v Speaker 1>a vacuum in very interesting technology. The problem is they

0:59:05.840 --> 0:59:09.360
<v Speaker 1>tried to scale them like a giant, multi trillion dollar industry,

0:59:09.600 --> 0:59:12.440
<v Speaker 1>which I don't think they fit into the future of

0:59:12.560 --> 0:59:16.280
<v Speaker 1>LMS in whatever form they have is on device. They

0:59:16.440 --> 0:59:18.720
<v Speaker 1>never should have been scaled at this scale, like it

0:59:18.880 --> 0:59:21.920
<v Speaker 1>never made sense, never once, And in the end, it

0:59:22.000 --> 0:59:24.200
<v Speaker 1>is normal technology. And that is the thing to always

0:59:24.240 --> 0:59:27.560
<v Speaker 1>take away from this, which is you wouldn't freak out

0:59:27.560 --> 0:59:32.600
<v Speaker 1>about Kubernetes. You wouldn't Microsoft word doing autocomplete, shouldn't scare you, grammarly,

0:59:32.600 --> 0:59:37.680
<v Speaker 1>shouldn't scare you. This is the actual technology itself is

0:59:37.720 --> 0:59:41.520
<v Speaker 1>not scaring you. It's the people around it and the

0:59:41.560 --> 0:59:47.200
<v Speaker 1>people's extrapolations. And also, like you said, the willingness. We

0:59:47.360 --> 0:59:51.040
<v Speaker 1>have to believe people who are working on something, who

0:59:51.040 --> 0:59:55.800
<v Speaker 1>have supposed expertise, as if expertise guarantees that you're correct,

0:59:55.880 --> 0:59:59.240
<v Speaker 1>or indeed guarantees that you know what you're talking about,

0:59:59.440 --> 1:00:00.840
<v Speaker 1>or indeed that you can see the future.

1:00:02.760 --> 1:00:05.160
<v Speaker 2>Yeah, so there we go. I mean I think that's yeah, yeah,

1:00:05.200 --> 1:00:08.000
<v Speaker 2>I mean I don't know who knows. I'm already seeing

1:00:08.040 --> 1:00:09.640
<v Speaker 2>Like the comments on my New York Times piece is

1:00:09.680 --> 1:00:12.960
<v Speaker 2>just falling into the like uh oh you you you

1:00:13.000 --> 1:00:14.959
<v Speaker 2>said somewhere in there there could be a benefit of AI.

1:00:15.120 --> 1:00:17.840
<v Speaker 2>You apostle AI is terrible, or like the other side

1:00:18.000 --> 1:00:20.760
<v Speaker 2>is is like, well, well, how do you explain the fact,

1:00:20.880 --> 1:00:22.400
<v Speaker 2>you know, I'm oh my god, it's the most is

1:00:22.440 --> 1:00:24.440
<v Speaker 2>this written by AI? These comments like this synth this

1:00:24.560 --> 1:00:27.000
<v Speaker 2>all the time, these these comments that are like I

1:00:27.120 --> 1:00:29.160
<v Speaker 2>work for a such and such company and what used

1:00:29.160 --> 1:00:31.720
<v Speaker 2>to take me three weeks I now do in like

1:00:31.800 --> 1:00:35.200
<v Speaker 2>six seconds. And you know, it's just like falling to

1:00:35.280 --> 1:00:37.760
<v Speaker 2>People are just falling in their camps about like either

1:00:37.800 --> 1:00:41.280
<v Speaker 2>you're one of the priests of the new religion or

1:00:41.760 --> 1:00:44.600
<v Speaker 2>you know you're the you're the the the the Canaanites

1:00:44.640 --> 1:00:46.040
<v Speaker 2>who are who were kicked out of the land or

1:00:46.040 --> 1:00:47.400
<v Speaker 2>something right like, it's like you got to be in

1:00:47.440 --> 1:00:50.440
<v Speaker 2>like one of these two like extremes, and don't do that.

1:00:50.920 --> 1:00:53.120
<v Speaker 2>Just say, like, treat it like a normal technology. Stop

1:00:53.200 --> 1:00:56.360
<v Speaker 2>doom trolling is stupid. Stop it like, treat it like

1:00:56.360 --> 1:00:58.960
<v Speaker 2>a normal technology. Either stop if you really think you're

1:00:59.000 --> 1:01:02.480
<v Speaker 2>destroying the world, or start talking about it like pickup trucks. Right,

1:01:02.600 --> 1:01:06.200
<v Speaker 2>we don't need this in between, So hopefully, hopefully we

1:01:06.240 --> 1:01:07.640
<v Speaker 2>can convince a few more people about that.

1:01:07.920 --> 1:01:10.480
<v Speaker 1>Ope, so cal thank you so much for joining me.

1:01:10.760 --> 1:01:11.600
<v Speaker 2>Oh it's a pleasure.

1:01:20.600 --> 1:01:22.280
<v Speaker 1>Thank you for listening to Better Offline.

1:01:22.440 --> 1:01:24.840
<v Speaker 3>The editor and composer of the Better Offline theme song

1:01:24.920 --> 1:01:27.560
<v Speaker 3>is Matasowski. You can check out more of his music

1:01:27.600 --> 1:01:31.240
<v Speaker 3>and audio projects at Matasowski dot com, M A T

1:01:31.240 --> 1:01:35.720
<v Speaker 3>T O S O W s KI dot com. You

1:01:35.720 --> 1:01:38.240
<v Speaker 3>can email me at easy at Better offline dot com,

1:01:38.320 --> 1:01:40.640
<v Speaker 3>or visit Better Offline dot com to find more podcast

1:01:40.680 --> 1:01:44.000
<v Speaker 3>links and of course my newsletter. I also really recommend

1:01:44.040 --> 1:01:46.000
<v Speaker 3>you go to chat dot Where's youreaed dot at to

1:01:46.080 --> 1:01:48.440
<v Speaker 3>visit the discord and go to our slash.

1:01:48.120 --> 1:01:51.280
<v Speaker 1>Better Offline to check out our reddit. Thank you so

1:01:51.400 --> 1:01:52.120
<v Speaker 1>much for listening.

1:01:52.960 --> 1:01:55.440
<v Speaker 2>Better Offline is a production of cool Zone Media.

1:01:55.560 --> 1:01:58.400
<v Speaker 3>For more from cool Zone Media, visit our website cool

1:01:58.480 --> 1:02:01.800
<v Speaker 3>Zonemedia dot com. Check us out on the iHeartRadio app,

1:02:01.880 --> 1:02:02.160
<v Speaker 3>Apple

1:02:02.200 --> 1:02:17.760
<v Speaker 2>Podcasts, or wherever you get your podcasts