1 00:00:02,440 --> 00:00:13,400 Speaker 1: Bloomberg Audio Studios, Podcasts, Radio News. 2 00:00:17,960 --> 00:00:21,120 Speaker 2: Hello and welcome to another episode of the Odd Lots Podcast. 3 00:00:21,200 --> 00:00:22,560 Speaker 2: I'm Tracy Alloway. 4 00:00:22,280 --> 00:00:23,400 Speaker 3: And I'm Joe Whysenthal. 5 00:00:23,680 --> 00:00:25,639 Speaker 2: Joe, have you watched The Three Body Problem? 6 00:00:25,960 --> 00:00:28,200 Speaker 3: No, but I really want to, and I didn't read 7 00:00:28,200 --> 00:00:29,720 Speaker 3: the book. So in case you're going to ask that 8 00:00:29,800 --> 00:00:32,000 Speaker 3: I didn't, I want to do that too, but I 9 00:00:32,080 --> 00:00:33,319 Speaker 3: intend to at some point. 10 00:00:33,760 --> 00:00:37,200 Speaker 2: There goes my carefully crafted intro where we talk about 11 00:00:37,200 --> 00:00:40,000 Speaker 2: the Three Body Problem. Okay, well, this will work well. 12 00:00:40,400 --> 00:00:43,640 Speaker 2: As everyone knows except for Joe, there's sort of two 13 00:00:43,760 --> 00:00:46,080 Speaker 2: types of people in the world when it comes to 14 00:00:46,080 --> 00:00:48,440 Speaker 2: the Three Body Problem. There are those who see it 15 00:00:48,520 --> 00:00:52,760 Speaker 2: as an allegory for climate change, so humans coming together 16 00:00:52,920 --> 00:00:56,120 Speaker 2: to unite against a common threat, which, in this case, 17 00:00:56,200 --> 00:00:58,760 Speaker 2: since you haven't read the book, is an alien. 18 00:01:00,240 --> 00:01:00,440 Speaker 4: Yeah. 19 00:01:00,440 --> 00:01:02,800 Speaker 3: A friend of mine this weekend told me like two 20 00:01:02,880 --> 00:01:03,440 Speaker 3: plot points. 21 00:01:03,440 --> 00:01:06,600 Speaker 2: Okay, good, good, good, yes, okay. And then there are 22 00:01:06,640 --> 00:01:09,520 Speaker 2: also those who see it as sort of an allegory 23 00:01:09,800 --> 00:01:13,600 Speaker 2: for the trade or tech war between the US and China, 24 00:01:13,800 --> 00:01:16,319 Speaker 2: So the idea that humans are going up against a 25 00:01:16,400 --> 00:01:20,760 Speaker 2: much more technologically advanced opponent, and in this scenario, I 26 00:01:20,760 --> 00:01:24,840 Speaker 2: guess Earth is China and the aliens are the US. Well, 27 00:01:25,000 --> 00:01:28,280 Speaker 2: today we are firmly in that second camp. We're going 28 00:01:28,360 --> 00:01:33,080 Speaker 2: to talk about US China rivalry in tech, and in 29 00:01:33,120 --> 00:01:35,559 Speaker 2: particular one area of tech AI. 30 00:01:35,920 --> 00:01:39,920 Speaker 3: Right, so obviously AI AI AI, everyone talks about it 31 00:01:40,040 --> 00:01:42,560 Speaker 3: all the time. We don't really know where it's going 32 00:01:42,600 --> 00:01:44,560 Speaker 3: to go, but we know a few things in the meantime, 33 00:01:44,600 --> 00:01:47,000 Speaker 3: which is that people are spending money like crazy on chips, 34 00:01:47,160 --> 00:01:50,520 Speaker 3: but they're also spending money like crazy on talent. And 35 00:01:50,600 --> 00:01:53,880 Speaker 3: anyone who is capable of doing sort of cutting edge 36 00:01:54,000 --> 00:01:57,440 Speaker 3: research in AI, from what I can tell based on articles, 37 00:01:57,840 --> 00:01:59,760 Speaker 3: like they basically just get to pick where they want 38 00:01:59,800 --> 00:02:02,480 Speaker 3: to work, can basically pick their salary. There's a great 39 00:02:02,640 --> 00:02:05,160 Speaker 3: article in the Information a couple of weeks ago about 40 00:02:05,360 --> 00:02:08,680 Speaker 3: Facebook hiring top researchers without even doing an interview. It's like, 41 00:02:08,680 --> 00:02:10,800 Speaker 3: if you know this stuff, someone will hire you and 42 00:02:10,840 --> 00:02:11,639 Speaker 3: pay you a lot of money. 43 00:02:11,800 --> 00:02:14,160 Speaker 2: Yeah, And I have so many questions in this space. 44 00:02:14,200 --> 00:02:17,280 Speaker 2: So first of all, like who is an AI talent 45 00:02:17,960 --> 00:02:20,519 Speaker 2: or what is an AI talent? Where do they come from? 46 00:02:20,639 --> 00:02:22,840 Speaker 2: Is it the same as being a software engineer, but 47 00:02:22,880 --> 00:02:26,679 Speaker 2: you have a slightly different area of expertise. I really 48 00:02:26,720 --> 00:02:29,320 Speaker 2: don't know. And then secondly, I'm kind of curious how 49 00:02:29,360 --> 00:02:32,240 Speaker 2: fungible the jobs are. From what you just said and 50 00:02:32,280 --> 00:02:35,880 Speaker 2: the fact that companies are hiring without interviews and things 51 00:02:35,919 --> 00:02:38,880 Speaker 2: like that, and that demand is so strong, it seems 52 00:02:38,960 --> 00:02:42,400 Speaker 2: like you can just do AI anywhere, whether it's China 53 00:02:42,520 --> 00:02:44,320 Speaker 2: or the US or somewhere else in the world, or 54 00:02:44,320 --> 00:02:47,520 Speaker 2: whether it's a specific company versus another one. But so 55 00:02:47,600 --> 00:02:51,280 Speaker 2: many questions on this AI talent war. I guess you 56 00:02:51,280 --> 00:02:52,280 Speaker 2: could say totally. 57 00:02:52,320 --> 00:02:54,760 Speaker 3: And there's two things. So I sort of consider myself 58 00:02:54,760 --> 00:02:56,600 Speaker 3: a bit of an AI talent because I think I'm 59 00:02:56,600 --> 00:02:58,400 Speaker 3: pretty good at coming up with chet GPT. 60 00:02:58,639 --> 00:03:01,720 Speaker 2: You are, actually I listeners. I have learned a lot 61 00:03:01,760 --> 00:03:04,679 Speaker 2: from watching Joe enter his prompts, and I still find 62 00:03:04,720 --> 00:03:07,120 Speaker 2: it incredibly endearing that you say please and thank you. 63 00:03:07,720 --> 00:03:10,760 Speaker 3: Well, it's important for when AI becomes sentient that they're 64 00:03:10,760 --> 00:03:13,560 Speaker 3: going to remember who said please and thank you. But 65 00:03:13,760 --> 00:03:16,600 Speaker 3: beyond that, you know, there's this other element, and you 66 00:03:16,639 --> 00:03:19,360 Speaker 3: already sort of alluded to it. But it's clear that 67 00:03:19,840 --> 00:03:24,600 Speaker 3: for whatever reason, countries feel like AI, almost as if 68 00:03:24,639 --> 00:03:28,040 Speaker 3: it's a commodity there it must be some every country, 69 00:03:28,160 --> 00:03:30,840 Speaker 3: or there's this narrative being pushed by the industry, and 70 00:03:30,880 --> 00:03:34,240 Speaker 3: maybe it's just a narrative to sell chips or subscriptions 71 00:03:34,280 --> 00:03:36,720 Speaker 3: to the open AI APIs. 72 00:03:36,200 --> 00:03:36,600 Speaker 2: Et cetera. 73 00:03:36,880 --> 00:03:40,080 Speaker 3: But there seems to be this narrative that every country 74 00:03:40,160 --> 00:03:45,360 Speaker 3: must have some sort of homegrown AI strategy data center 75 00:03:45,480 --> 00:03:49,480 Speaker 3: or something like. Something about this technology seems to engender 76 00:03:50,080 --> 00:03:52,480 Speaker 3: political and nationalistic anxieties. 77 00:03:52,960 --> 00:03:55,480 Speaker 2: Yes, I think that's absolutely true, and we're back to 78 00:03:55,600 --> 00:04:00,160 Speaker 2: sort of the three yady geopolitical tension point. But I 79 00:04:00,200 --> 00:04:02,480 Speaker 2: am very pleased to say that we in fact have 80 00:04:02,680 --> 00:04:05,000 Speaker 2: the perfect guest to talk about all of this. We're 81 00:04:05,040 --> 00:04:07,160 Speaker 2: going to be speaking with Damien ma He is the 82 00:04:07,280 --> 00:04:10,760 Speaker 2: managing director at macro Polo, which is the think tank 83 00:04:10,840 --> 00:04:14,800 Speaker 2: at the Paulson Institute, and they publish something called the 84 00:04:14,880 --> 00:04:19,960 Speaker 2: Global AI Talent Tracker, so actually keeping track of where 85 00:04:20,000 --> 00:04:22,520 Speaker 2: AI talent is coming from, how much there is, and 86 00:04:22,560 --> 00:04:25,359 Speaker 2: where it's going. So Damian, thank you so much for 87 00:04:25,400 --> 00:04:26,000 Speaker 2: coming on all. 88 00:04:25,920 --> 00:04:28,120 Speaker 4: Thoughts, Thank you so much, it's great to be here. 89 00:04:28,440 --> 00:04:31,320 Speaker 2: How long have you guys been doing this talent tracker? 90 00:04:31,400 --> 00:04:34,640 Speaker 2: And what was the genesis because for me chat, GPT 91 00:04:35,000 --> 00:04:37,200 Speaker 2: and all the chatbots seem to have come out of 92 00:04:37,200 --> 00:04:40,960 Speaker 2: nowhere almost basically a year ago. So how did you 93 00:04:41,080 --> 00:04:43,440 Speaker 2: get an early start on tracking AI? 94 00:04:44,040 --> 00:04:46,280 Speaker 4: Well, the original conception is that we thought a little 95 00:04:46,279 --> 00:04:48,320 Speaker 4: bit hard about, you know, what would you need to 96 00:04:48,400 --> 00:04:53,000 Speaker 4: have a robust AI ecosystem or an AI industry, And 97 00:04:53,040 --> 00:04:56,080 Speaker 4: we thought there are three key pieces. You need. Obviously, 98 00:04:56,440 --> 00:05:00,320 Speaker 4: compute power, so things like chips and the infrastructure need. 99 00:05:00,320 --> 00:05:03,440 Speaker 4: Obviously a lot of training data. Data is obviously everywhere now. 100 00:05:03,480 --> 00:05:05,480 Speaker 4: And we thought the last piece that people haven't thought 101 00:05:05,520 --> 00:05:08,360 Speaker 4: too much about is human capital because it is a 102 00:05:08,480 --> 00:05:12,320 Speaker 4: very human capital intensive area and discipline because it's highly 103 00:05:12,360 --> 00:05:15,200 Speaker 4: complex and complicated and you need to highly trained people 104 00:05:15,240 --> 00:05:16,919 Speaker 4: to be able to do it. So we thought, nobody's 105 00:05:16,960 --> 00:05:19,560 Speaker 4: really looked at the human capital side of things. Is 106 00:05:19,600 --> 00:05:21,320 Speaker 4: there a way to do that? And then so we 107 00:05:21,440 --> 00:05:24,960 Speaker 4: sort of found this one conference that's widely known in 108 00:05:25,000 --> 00:05:28,320 Speaker 4: the AI community as one of the most prestigious, and 109 00:05:28,400 --> 00:05:31,080 Speaker 4: so we looked at papers and researchers that went to 110 00:05:31,120 --> 00:05:34,120 Speaker 4: that conference. This was back in twenty twenty during the pandemic. 111 00:05:34,279 --> 00:05:37,400 Speaker 4: Was when we first launched the initial tracker. That gave 112 00:05:37,480 --> 00:05:39,880 Speaker 4: us our idea that's a proxy for sort of the 113 00:05:39,920 --> 00:05:44,240 Speaker 4: top twenty percent of global AI talent. So this is 114 00:05:44,279 --> 00:05:47,360 Speaker 4: not all AI talent, This is not everybody in the world, 115 00:05:47,680 --> 00:05:49,560 Speaker 4: but this is really sort of what we might call 116 00:05:49,560 --> 00:05:51,919 Speaker 4: it the cream of the crop, top twenty percent, and 117 00:05:52,000 --> 00:05:55,120 Speaker 4: within that there's also the top two percent. So we're 118 00:05:55,160 --> 00:05:57,200 Speaker 4: looking at really kind of the elite people, which is 119 00:05:57,320 --> 00:06:00,440 Speaker 4: probably the type of people that's being thought of, you know, 120 00:06:00,480 --> 00:06:03,799 Speaker 4: most fiercely, because people want the top talent real quickly. 121 00:06:03,839 --> 00:06:04,560 Speaker 3: What's the conference. 122 00:06:04,800 --> 00:06:07,960 Speaker 4: It's called the Newer IPS. It's a conference that's held 123 00:06:08,160 --> 00:06:10,360 Speaker 4: I think, I think every year, but we didn't track 124 00:06:10,400 --> 00:06:12,440 Speaker 4: it every year. We tracked in twenty twenty and then 125 00:06:12,480 --> 00:06:14,239 Speaker 4: we did it again and we looked at the twenty 126 00:06:14,320 --> 00:06:17,440 Speaker 4: twenty two. We were trying to see, you know, had 127 00:06:17,480 --> 00:06:20,440 Speaker 4: there been any changes after the three year pandemic, to 128 00:06:20,440 --> 00:06:23,600 Speaker 4: see if there were different mobility patterns. This is a 129 00:06:23,600 --> 00:06:28,960 Speaker 4: conference that's mainly focused on neuro networks, large language models, 130 00:06:28,960 --> 00:06:31,160 Speaker 4: so a lot of things that are currently really pushing 131 00:06:31,240 --> 00:06:34,360 Speaker 4: up frontiers of a generative AI. So we thought that 132 00:06:34,400 --> 00:06:37,120 Speaker 4: those are the kinds of people that would probably want 133 00:06:37,160 --> 00:06:39,279 Speaker 4: to work for the Googles that open AIS and you know, 134 00:06:39,360 --> 00:06:41,760 Speaker 4: buy dues of the world, and so that seemed like 135 00:06:41,800 --> 00:06:44,960 Speaker 4: a good sampling. Again, we don't pretend that this is comprehensive, 136 00:06:45,200 --> 00:06:47,839 Speaker 4: but it is sort of the elite twenty percent sample. 137 00:06:48,080 --> 00:06:51,440 Speaker 3: Just real quickly. Since you say you're able to distinguish 138 00:06:51,560 --> 00:06:54,360 Speaker 3: between the top twenty percent and the top two percent, 139 00:06:54,839 --> 00:06:56,640 Speaker 3: how do you do that part? I mean, it can't 140 00:06:56,640 --> 00:06:58,760 Speaker 3: just be people who attend the conference, Like how do 141 00:06:58,800 --> 00:07:01,120 Speaker 3: you sort of grade or figure out like who is 142 00:07:01,160 --> 00:07:04,559 Speaker 3: this specific ultra elite AI engineering talent? 143 00:07:04,960 --> 00:07:07,600 Speaker 4: So we looked at authors whose papers got accepted, and 144 00:07:07,600 --> 00:07:11,680 Speaker 4: within that acceptance there's a oral presentation. You don't get 145 00:07:11,720 --> 00:07:14,520 Speaker 4: accepted to oral presentation unless you're really really good. So 146 00:07:14,520 --> 00:07:16,280 Speaker 4: they are only about two percent of people that got 147 00:07:16,320 --> 00:07:19,240 Speaker 4: accepted at oral presentation, So that to us was sort 148 00:07:19,280 --> 00:07:20,800 Speaker 4: of the proxy for the two percent. 149 00:07:21,240 --> 00:07:23,760 Speaker 2: This kind of leads into what I was wondering, which 150 00:07:23,800 --> 00:07:28,040 Speaker 2: is what makes a really good AI engineer? Like what 151 00:07:28,160 --> 00:07:30,640 Speaker 2: is it that would lead them to be someone who 152 00:07:30,680 --> 00:07:32,320 Speaker 2: presents at a conference like this. 153 00:07:32,920 --> 00:07:35,560 Speaker 4: I mean, Joe just said, you know, he's a really 154 00:07:35,560 --> 00:07:37,080 Speaker 4: good prompt engineer. 155 00:07:36,640 --> 00:07:36,720 Speaker 1: So. 156 00:07:38,520 --> 00:07:39,600 Speaker 3: They would let me present. 157 00:07:39,760 --> 00:07:41,240 Speaker 2: Joe, I'm sure your invites. 158 00:07:40,920 --> 00:07:44,000 Speaker 4: In the mail, you know, like really curate the questions. Well, 159 00:07:44,040 --> 00:07:45,240 Speaker 4: but I think that's really good. 160 00:07:46,000 --> 00:07:48,640 Speaker 2: It's not just curating the questions, right, it's like actually 161 00:07:48,680 --> 00:07:50,880 Speaker 2: coming up with the natural language models and things like that. 162 00:07:50,960 --> 00:07:53,200 Speaker 4: Okay, yeah, so so I think it's a really good question, 163 00:07:53,240 --> 00:07:55,720 Speaker 4: and I'm not sure the distinction is huge. I think 164 00:07:55,760 --> 00:07:59,160 Speaker 4: the foundation of AI is all computer science. Most AI 165 00:07:59,200 --> 00:08:02,280 Speaker 4: people would call them those computer scientists first and foremost, 166 00:08:02,680 --> 00:08:05,240 Speaker 4: or people that have a lot of mathematical training. And 167 00:08:05,280 --> 00:08:07,680 Speaker 4: in fact, I think some of those people I think 168 00:08:07,720 --> 00:08:09,560 Speaker 4: back into two thousands and two thousand tens, we're the 169 00:08:09,560 --> 00:08:12,040 Speaker 4: same people that got attracted to big finance, right and 170 00:08:12,120 --> 00:08:15,240 Speaker 4: went to build algorithms for you know, trading desks. Those 171 00:08:15,280 --> 00:08:17,400 Speaker 4: are probably a similar type of people now they're just 172 00:08:17,400 --> 00:08:20,680 Speaker 4: doing AI. And the AI specific apply part is being 173 00:08:20,680 --> 00:08:23,640 Speaker 4: able to train large amounts of data and be able 174 00:08:23,680 --> 00:08:26,320 Speaker 4: to write out algorithms. But those are the things that 175 00:08:26,360 --> 00:08:28,960 Speaker 4: you would get from computer science training with a bit 176 00:08:28,960 --> 00:08:31,720 Speaker 4: of sort of a you know, added AI specific component 177 00:08:31,760 --> 00:08:34,240 Speaker 4: to it. And I think the neuro networks thing is 178 00:08:34,320 --> 00:08:36,920 Speaker 4: probably you know, one distinguishing characteristic is trying to really 179 00:08:36,960 --> 00:08:39,160 Speaker 4: figure out how do you make the computer mimic the 180 00:08:39,240 --> 00:08:44,040 Speaker 4: human brain in a way. But fundamentally it's just mathematics, 181 00:08:44,120 --> 00:08:47,679 Speaker 4: quantitative computer science. All those things you know eventually can 182 00:08:47,760 --> 00:08:49,000 Speaker 4: become AI scientists. 183 00:08:49,600 --> 00:08:52,280 Speaker 3: So there's a certain type of person who is seeking 184 00:08:52,320 --> 00:08:55,199 Speaker 3: out the hardest or maybe most lucrative sort of real 185 00:08:55,240 --> 00:08:58,000 Speaker 3: world math problem or computer science problem. At any time. 186 00:08:58,559 --> 00:09:01,320 Speaker 3: Maybe in the two thousand they were going to Wall 187 00:09:01,360 --> 00:09:03,640 Speaker 3: Street to figure out the best way to create new 188 00:09:03,679 --> 00:09:07,040 Speaker 3: securitized products and derivatives. In the twenty tens, they went 189 00:09:07,080 --> 00:09:10,560 Speaker 3: to Facebook and Google to figure out the ways to 190 00:09:10,640 --> 00:09:12,920 Speaker 3: pack the most number of ads on a smartphone or 191 00:09:13,160 --> 00:09:15,840 Speaker 3: get you to click on them. And now apparently they're 192 00:09:15,880 --> 00:09:19,120 Speaker 3: going into AI research. So let's start with what the 193 00:09:19,240 --> 00:09:23,199 Speaker 3: data shows. Big picture. When you started first started collecting 194 00:09:23,240 --> 00:09:26,120 Speaker 3: the data in twenty twenty, where were they coming from 195 00:09:26,160 --> 00:09:26,839 Speaker 3: and where were they going? 196 00:09:27,240 --> 00:09:29,040 Speaker 4: A lot of them came out of China in the 197 00:09:29,120 --> 00:09:32,440 Speaker 4: United States in twenty twenty, that was pretty clear. Most 198 00:09:32,440 --> 00:09:35,160 Speaker 4: of them ended up in the United States by far, 199 00:09:35,559 --> 00:09:39,240 Speaker 4: and we're still seeing that in our latest update in 200 00:09:39,440 --> 00:09:42,400 Speaker 4: twenty twenty three. Although I would say the big surprise 201 00:09:42,600 --> 00:09:44,760 Speaker 4: was that China has done a really good job really 202 00:09:44,880 --> 00:09:48,840 Speaker 4: ramping up its domestic supply of top AS scientists, so 203 00:09:48,880 --> 00:09:52,600 Speaker 4: they're producing nearly half of the world's top tier AI 204 00:09:52,640 --> 00:09:55,840 Speaker 4: scientists now, and many of them are actually also staying 205 00:09:55,800 --> 00:09:58,800 Speaker 4: in China. And the reason is, I think it's pretty simple, 206 00:09:58,920 --> 00:10:01,720 Speaker 4: is that China is obviously been focusing on its own 207 00:10:02,040 --> 00:10:05,640 Speaker 4: AI industry, and as we already said, you know, people 208 00:10:05,679 --> 00:10:08,400 Speaker 4: go where the jobs are, and if you look at 209 00:10:08,760 --> 00:10:11,720 Speaker 4: the major economies where they're focused on building out AI 210 00:10:11,760 --> 00:10:15,960 Speaker 4: industry opportunities, it's probably the United States and China. And 211 00:10:16,000 --> 00:10:18,560 Speaker 4: if you look at Europe, actually I think punches way 212 00:10:18,559 --> 00:10:20,840 Speaker 4: below its way in terms of having an AI industry, 213 00:10:21,440 --> 00:10:24,280 Speaker 4: and so they you know, they don't tend to attract 214 00:10:24,720 --> 00:10:27,600 Speaker 4: as many top tier AI talent as China or the US. 215 00:10:28,160 --> 00:10:33,920 Speaker 4: And if you look within top US institutions where top 216 00:10:33,960 --> 00:10:38,240 Speaker 4: A italent work, it really is almost a Chinese American doopoly. 217 00:10:38,800 --> 00:10:42,960 Speaker 4: Chinese origin and American AI scientists are seventy five percent 218 00:10:43,400 --> 00:10:46,000 Speaker 4: of the top aalent within US institutions. 219 00:10:46,840 --> 00:10:50,640 Speaker 2: What are the factors that would go into say a 220 00:10:50,679 --> 00:10:54,560 Speaker 2: computer scientist who has been educated in China and they're 221 00:10:54,600 --> 00:10:58,800 Speaker 2: surveying the different opportunities available to them, what are the 222 00:10:58,840 --> 00:11:02,040 Speaker 2: factors that would go into them making a decision, like 223 00:11:02,120 --> 00:11:07,600 Speaker 2: are there immigration considerations? I imagine pay and renewneration would 224 00:11:07,679 --> 00:11:10,000 Speaker 2: have to factor into that how easy is it for 225 00:11:10,040 --> 00:11:13,720 Speaker 2: them to switch from China to the US. 226 00:11:14,120 --> 00:11:16,640 Speaker 4: I think the skills and you know, and the training 227 00:11:16,760 --> 00:11:18,560 Speaker 4: is fairly similar if if you come out of a 228 00:11:18,600 --> 00:11:22,400 Speaker 4: top program, whether it's Chinhua in China or or you know, 229 00:11:22,480 --> 00:11:26,520 Speaker 4: Stanford in California. I think the key from what we're seeing, 230 00:11:26,559 --> 00:11:28,800 Speaker 4: you know, one key indicator of where people end up 231 00:11:29,120 --> 00:11:31,320 Speaker 4: for work, you know, is really where they go to 232 00:11:31,360 --> 00:11:34,160 Speaker 4: graduate school. That's probably not a surprise. If you're going 233 00:11:34,200 --> 00:11:37,360 Speaker 4: to do your master's or PhD somewhere, you generally start 234 00:11:37,360 --> 00:11:40,000 Speaker 4: to search for job opportunities you know, near you, around you, 235 00:11:40,600 --> 00:11:43,360 Speaker 4: unless you happen to be in a country in an 236 00:11:43,400 --> 00:11:46,599 Speaker 4: area where there's not a lot of opportunities post graduation. 237 00:11:47,080 --> 00:11:50,240 Speaker 4: And of course, when you're considered an elite AI talent, 238 00:11:50,280 --> 00:11:53,480 Speaker 4: you generally have a terminal degree, usually a PhD, but 239 00:11:53,559 --> 00:11:56,240 Speaker 4: at least a master's, so, you know, I think where 240 00:11:56,280 --> 00:11:58,560 Speaker 4: you choose to go to graduate school is really important, 241 00:11:58,559 --> 00:12:00,760 Speaker 4: and we see that in the data. You know, those 242 00:12:00,760 --> 00:12:03,200 Speaker 4: who come to the United States or graduate school, by 243 00:12:03,240 --> 00:12:05,199 Speaker 4: and large tend to stay in the US to work 244 00:12:05,559 --> 00:12:08,560 Speaker 4: unless there's some very lucrative opportunity that attracts them back 245 00:12:08,600 --> 00:12:11,320 Speaker 4: home or somewhere else, But generally there's a bit of 246 00:12:11,320 --> 00:12:14,080 Speaker 4: a path dependence between you know, graduate school and staying 247 00:12:14,120 --> 00:12:15,160 Speaker 4: in that country to work. 248 00:12:15,600 --> 00:12:18,280 Speaker 3: There has been a lot of anxiety for years in 249 00:12:18,320 --> 00:12:21,600 Speaker 3: the tech industry where you see CEOs and leaders complaining 250 00:12:22,040 --> 00:12:24,880 Speaker 3: that the US immigration policy has made it too hard 251 00:12:25,160 --> 00:12:28,120 Speaker 3: to keep talent who has graduated in the United States, 252 00:12:28,160 --> 00:12:30,000 Speaker 3: and there's this idea of like, hey, if they're going 253 00:12:30,040 --> 00:12:32,080 Speaker 3: to come here for education, why are we not reaping 254 00:12:32,120 --> 00:12:36,520 Speaker 3: the benefits of the US educated talent. It does seem 255 00:12:36,559 --> 00:12:39,200 Speaker 3: like from your data that still many are staying in 256 00:12:39,200 --> 00:12:42,920 Speaker 3: the United States, But the numbers have changed since twenty twenty. 257 00:12:43,040 --> 00:12:45,719 Speaker 4: Yes, yes they have, you know, gone down a little bit. 258 00:12:46,000 --> 00:12:49,839 Speaker 4: We didn't go into really exploring exactly what happened over 259 00:12:49,840 --> 00:12:52,680 Speaker 4: the last three years, in part because I think many 260 00:12:52,679 --> 00:12:55,480 Speaker 4: people realize the pandemic years have been a little strange, 261 00:12:56,120 --> 00:12:59,000 Speaker 4: whether it's for economic data or just general ability for people, 262 00:12:59,000 --> 00:13:01,640 Speaker 4: where people work, how people work. So there's going to 263 00:13:01,679 --> 00:13:04,000 Speaker 4: be a lot of distortions in those last three years. 264 00:13:04,160 --> 00:13:07,240 Speaker 4: But there has been a relative decline, especially among the 265 00:13:07,280 --> 00:13:10,280 Speaker 4: Asian talent. It's not just China. India has also done 266 00:13:10,280 --> 00:13:12,800 Speaker 4: a better job retaining its own top tier AI talent 267 00:13:13,200 --> 00:13:16,640 Speaker 4: South Korea. Interestingly, that's not on our data set yet, 268 00:13:16,640 --> 00:13:20,920 Speaker 4: but we're about to publish Regional South Korea. They've retained 269 00:13:21,000 --> 00:13:24,040 Speaker 4: ninety percent of their talent, they've not let anybody leave, 270 00:13:24,240 --> 00:13:26,080 Speaker 4: and they've been really good at doing that. And places 271 00:13:26,120 --> 00:13:28,440 Speaker 4: like France have actually done a very good job on 272 00:13:28,520 --> 00:13:31,680 Speaker 4: retaining their talent. So I can't say definitively what the 273 00:13:31,720 --> 00:13:34,680 Speaker 4: reason is. Whether countries have stepped up their gain to 274 00:13:34,800 --> 00:13:37,600 Speaker 4: retain domestic talent, or there's been other things that happen 275 00:13:37,640 --> 00:13:39,960 Speaker 4: in the pandemic that's triggered it, or there could be 276 00:13:40,240 --> 00:13:44,320 Speaker 4: immigration challenges and so on. I think maybe in the future, 277 00:13:44,320 --> 00:13:46,760 Speaker 4: when we do the next iteration, we will have more 278 00:13:46,840 --> 00:13:49,160 Speaker 4: clarity to see the pattern. So I'd be a little 279 00:13:49,200 --> 00:13:51,880 Speaker 4: hesitant to give definitive conclusions at this point. 280 00:14:07,080 --> 00:14:10,400 Speaker 3: Tracy, If France does a really good job keeping their talent, 281 00:14:10,640 --> 00:14:13,400 Speaker 3: who will fill the niche of blowing up trading desks 282 00:14:13,400 --> 00:14:17,360 Speaker 3: with exotic derivatives? If all those it called polytechnique in 283 00:14:17,480 --> 00:14:19,840 Speaker 3: sciences PO graduates going to AI. 284 00:14:19,680 --> 00:14:23,800 Speaker 2: Onstay, Yeah, Yes, it is always a French person working 285 00:14:23,800 --> 00:14:27,520 Speaker 2: in equity derivatives with a mathematics degree. You're absolutely correct, 286 00:14:27,720 --> 00:14:31,600 Speaker 2: but on the degree topic. So I hadn't realized that 287 00:14:31,680 --> 00:14:35,600 Speaker 2: in China and Damian, I think this factoid was in 288 00:14:35,680 --> 00:14:38,280 Speaker 2: one of the reading materials that you sent. But Chinese 289 00:14:38,360 --> 00:14:42,480 Speaker 2: universities have launched more than two thy three hundred undergraduate 290 00:14:42,560 --> 00:14:47,120 Speaker 2: programs since twenty eighteen, when the Ministry of Education designated 291 00:14:47,200 --> 00:14:52,520 Speaker 2: AI as a separate major that's distinct from computer science. So, 292 00:14:52,920 --> 00:14:55,520 Speaker 2: first of all, how common is that that you would 293 00:14:55,560 --> 00:14:58,640 Speaker 2: get a separation between computer science versus AI? Is that 294 00:14:58,760 --> 00:15:00,840 Speaker 2: the standard in other parts arts of the world or 295 00:15:00,920 --> 00:15:04,680 Speaker 2: is it still relatively new? And then secondly, presumably this 296 00:15:04,720 --> 00:15:08,400 Speaker 2: is part of China trying to build up its domestic 297 00:15:08,520 --> 00:15:13,080 Speaker 2: AI talent pool and eventually its capabilities in this area. 298 00:15:13,200 --> 00:15:15,080 Speaker 2: What else is it doing on that front? 299 00:15:15,200 --> 00:15:17,520 Speaker 4: Yeah, So that's why one of the reasons we think 300 00:15:17,560 --> 00:15:20,480 Speaker 4: that China has really seen this boom on top AI 301 00:15:20,560 --> 00:15:22,800 Speaker 4: talent is you have just kind of a graduating class 302 00:15:22,800 --> 00:15:25,320 Speaker 4: in twenty twenty two. If you start in twenty eighteen, 303 00:15:25,560 --> 00:15:27,320 Speaker 4: some of them are graduate students, some of them are 304 00:15:27,840 --> 00:15:31,040 Speaker 4: undergrad so they've really pushed really hard to grow at 305 00:15:31,040 --> 00:15:33,160 Speaker 4: the AI talent when now not all of them are 306 00:15:33,200 --> 00:15:35,920 Speaker 4: the top twenty percent, But I think China looks at 307 00:15:35,960 --> 00:15:37,240 Speaker 4: it as a way that they're going to need a 308 00:15:37,280 --> 00:15:41,400 Speaker 4: lot of AI specific technicians. China's not really thinking about 309 00:15:41,400 --> 00:15:44,200 Speaker 4: AI in the generative AI sense. I think there are 310 00:15:44,200 --> 00:15:48,680 Speaker 4: definitely some startups and folks pursuing things like chatbt Chatboss, 311 00:15:48,680 --> 00:15:51,280 Speaker 4: but my understanding is that China's probably going to focus 312 00:15:51,360 --> 00:15:58,479 Speaker 4: much more on industrial applications of AI, manufacturing, robotics, probably healthcare, biotech. 313 00:15:58,600 --> 00:16:00,000 Speaker 4: I'm going to bet that's going to be a huge 314 00:16:00,120 --> 00:16:03,400 Speaker 4: application for China. And I think for obvious reasons, generative 315 00:16:03,400 --> 00:16:07,360 Speaker 4: AI is probably not as copesthetic with the governance system 316 00:16:07,760 --> 00:16:10,560 Speaker 4: in China ultimately, and I think that's a pretty clear 317 00:16:10,600 --> 00:16:13,160 Speaker 4: thing that I think everyone knows. But I think they're 318 00:16:13,200 --> 00:16:17,120 Speaker 4: really looking at how to apply artificial intelligence to energy, 319 00:16:17,600 --> 00:16:21,480 Speaker 4: to industry, to advanced manufacturing, or things like climate. That's 320 00:16:21,520 --> 00:16:24,040 Speaker 4: where China's really focused on, and I think they feel 321 00:16:24,040 --> 00:16:26,480 Speaker 4: like they need a lot more people, not just the 322 00:16:26,520 --> 00:16:28,560 Speaker 4: cream of the crop, but sort of you know, middle 323 00:16:28,640 --> 00:16:32,160 Speaker 4: level technicians, people that are just familiar with being able 324 00:16:32,240 --> 00:16:35,320 Speaker 4: to like run data or to run Python, or to 325 00:16:35,480 --> 00:16:38,280 Speaker 4: just check all the data. So I think they're viewing 326 00:16:38,320 --> 00:16:42,119 Speaker 4: AI as a very wide, expansive way of creating certain jobs. 327 00:16:42,560 --> 00:16:45,480 Speaker 2: Yeah, I can't imagine China's ambition here is to have 328 00:16:45,560 --> 00:16:49,520 Speaker 2: like five thousand different chatbots. Like there is clearly a 329 00:16:49,680 --> 00:16:54,880 Speaker 2: tendency towards industrial sort of real world applications of this technology. 330 00:16:55,320 --> 00:16:58,880 Speaker 2: On which note, do you think there's currently enough places 331 00:16:59,160 --> 00:17:04,400 Speaker 2: for AI graduates or specialists to actually go within China, 332 00:17:04,480 --> 00:17:06,760 Speaker 2: Because in some respects it feels like this might be 333 00:17:06,760 --> 00:17:09,640 Speaker 2: a very hot degree. People are being encouraged to do it, 334 00:17:09,880 --> 00:17:14,240 Speaker 2: but at the moment companies aren't necessarily at the same 335 00:17:14,280 --> 00:17:17,000 Speaker 2: sort of level. It feels like there's sort of a 336 00:17:17,040 --> 00:17:18,960 Speaker 2: mismatch in the evolution of this. 337 00:17:19,160 --> 00:17:21,560 Speaker 4: At the moment, I think you're absolutely right. So we've 338 00:17:21,560 --> 00:17:23,879 Speaker 4: seen these kinds of bubbles before that you know, the 339 00:17:23,960 --> 00:17:26,840 Speaker 4: new hottest sector in China, everyone goes there because they 340 00:17:26,840 --> 00:17:29,800 Speaker 4: think that's where the opportunities are. And then you know, 341 00:17:29,880 --> 00:17:32,359 Speaker 4: China already had what we would call a college bubble 342 00:17:32,720 --> 00:17:34,520 Speaker 4: for the last ten years, and that's why you have, 343 00:17:34,600 --> 00:17:37,919 Speaker 4: you know, really high youth joblessness in China. Though. The 344 00:17:37,960 --> 00:17:40,359 Speaker 4: way I think about how China works in that respect 345 00:17:40,400 --> 00:17:44,000 Speaker 4: specifically is that they're basically two different cycles. In China, 346 00:17:44,000 --> 00:17:46,919 Speaker 4: there is a policy induced cycle, and then there's an 347 00:17:46,920 --> 00:17:49,760 Speaker 4: actual Marcus cycle that comes after that. So right now 348 00:17:49,760 --> 00:17:51,840 Speaker 4: we're in sort of this policy driven like you know, 349 00:17:51,920 --> 00:17:54,399 Speaker 4: you guys got to come in and we really like AI. 350 00:17:54,520 --> 00:17:57,280 Speaker 4: We're going to create all these programs and you should 351 00:17:57,280 --> 00:17:59,960 Speaker 4: just get AI. And then you know, parents are like, well, 352 00:18:00,080 --> 00:18:02,480 Speaker 4: well that seems like a good new thing, and that's 353 00:18:02,520 --> 00:18:05,239 Speaker 4: what the government's promoting. So all my kids that are 354 00:18:05,240 --> 00:18:06,840 Speaker 4: going to you know, do computer science, they're going to 355 00:18:06,880 --> 00:18:08,960 Speaker 4: add the AI component to it. So that's sort of 356 00:18:09,040 --> 00:18:12,159 Speaker 4: the policy induced cycle. And then after that, once the 357 00:18:12,200 --> 00:18:14,560 Speaker 4: bubble happens, it will kind of eventually get into a 358 00:18:14,600 --> 00:18:17,360 Speaker 4: market cycle where it'll correct a little bit. And then 359 00:18:17,400 --> 00:18:19,480 Speaker 4: and then people will be like, oh, well, actually we 360 00:18:19,840 --> 00:18:22,480 Speaker 4: probably not have an oversupply of a lot of these 361 00:18:22,520 --> 00:18:25,080 Speaker 4: you know, you know middle AI technicians that will have 362 00:18:25,119 --> 00:18:27,360 Speaker 4: no jobs. What are we going to do with them? 363 00:18:28,160 --> 00:18:28,679 Speaker 5: We don't know. 364 00:18:29,000 --> 00:18:31,560 Speaker 4: So I think this is a pattern that happens in 365 00:18:31,640 --> 00:18:33,680 Speaker 4: China a lot, and I wouldn't be surprised if that 366 00:18:33,760 --> 00:18:35,640 Speaker 4: happens with the AI talent pool as well. 367 00:18:36,119 --> 00:18:39,480 Speaker 3: So there's a lot of interesting threads to pull on 368 00:18:39,640 --> 00:18:42,600 Speaker 3: already in this conversation, and I want to return to 369 00:18:43,280 --> 00:18:46,199 Speaker 3: the non chatbot applications of AI, like how can we 370 00:18:46,240 --> 00:18:49,680 Speaker 3: make better robots and factories and drug discovery, et cetera. 371 00:18:49,840 --> 00:18:51,800 Speaker 3: But I want to ask another question. So, okay, all 372 00:18:51,920 --> 00:18:56,159 Speaker 3: these new institutions or graduate programs have been launched in 373 00:18:56,359 --> 00:19:00,320 Speaker 3: China and more and more universities offering degrees and or 374 00:19:00,320 --> 00:19:03,880 Speaker 3: computer science or related fields. In my mind's eye, if 375 00:19:03,920 --> 00:19:07,120 Speaker 3: I imagine what a top AI researcher, I imagine maybe 376 00:19:07,119 --> 00:19:11,360 Speaker 3: they have a PhD from MIT or Stanford or something 377 00:19:11,440 --> 00:19:14,080 Speaker 3: like that. When you look at the institutions in China, 378 00:19:14,200 --> 00:19:17,199 Speaker 3: has there been any sort of broadening out of the 379 00:19:17,320 --> 00:19:21,720 Speaker 3: number of schools that are capable of producing either those 380 00:19:21,760 --> 00:19:25,679 Speaker 3: top twenty percent or top two percent talent beyond just 381 00:19:25,720 --> 00:19:27,800 Speaker 3: the sort of like handful of schools that we've for 382 00:19:27,840 --> 00:19:30,520 Speaker 3: a long time understood as the elite schools. 383 00:19:31,119 --> 00:19:33,240 Speaker 4: There have been a little bit. And when it comes 384 00:19:33,240 --> 00:19:35,600 Speaker 4: to Asia specifically and China, I think they have the 385 00:19:35,600 --> 00:19:39,480 Speaker 4: eleven of the fourteen top AI institutions in Asia. But 386 00:19:39,640 --> 00:19:42,440 Speaker 4: in terms sort of you know, just top in general, 387 00:19:42,960 --> 00:19:46,800 Speaker 4: China has climbed quite a bit. Places like Due John University, 388 00:19:47,359 --> 00:19:50,080 Speaker 4: shanghaijel Tone, which are not your traditional names that you 389 00:19:50,080 --> 00:19:50,480 Speaker 4: would hear. 390 00:19:50,920 --> 00:19:52,000 Speaker 3: Yeah, I've never heard of either. 391 00:19:52,240 --> 00:19:56,119 Speaker 4: It's not Pku, it's not chin Hua. And interestingly, this 392 00:19:56,280 --> 00:19:58,880 Speaker 4: is an interesting you enter into into twenty twenty two. 393 00:19:58,960 --> 00:20:02,600 Speaker 4: Huawei is actually one of the top twenty five institutions 394 00:20:02,600 --> 00:20:05,960 Speaker 4: for AI researcher globally, so they've invested a lot in 395 00:20:06,040 --> 00:20:08,520 Speaker 4: hiring top AI talent for obvious reasons. 396 00:20:09,280 --> 00:20:12,080 Speaker 2: This is actually exactly what I wanted to ask you next, 397 00:20:12,119 --> 00:20:15,240 Speaker 2: which is you mentioned I do as well earlier in 398 00:20:15,280 --> 00:20:20,800 Speaker 2: the conversation. But in terms of domestic destinations for AI specialists, 399 00:20:21,280 --> 00:20:24,280 Speaker 2: is the idea here that a lot of the existing 400 00:20:24,680 --> 00:20:28,320 Speaker 2: internet companies in China that they're going to devote more 401 00:20:28,359 --> 00:20:32,680 Speaker 2: development and more resources to this particular technology as we've 402 00:20:32,720 --> 00:20:35,479 Speaker 2: seen here in the US. But also that maybe some 403 00:20:35,520 --> 00:20:40,520 Speaker 2: of those big like consumer internet companies, the ones that 404 00:20:40,840 --> 00:20:44,600 Speaker 2: had a very rough few years during Shi Shinping's big 405 00:20:44,680 --> 00:20:48,879 Speaker 2: crackdown on disorderly capital expansion, that they're going to pivot 406 00:20:48,920 --> 00:20:49,720 Speaker 2: as well. 407 00:20:49,840 --> 00:20:52,600 Speaker 4: So I think that's basically correct. I do, as far 408 00:20:52,640 --> 00:20:55,320 Speaker 4: as I'm concerned, has basically become an AI company, and 409 00:20:55,359 --> 00:20:58,679 Speaker 4: I think they made that strategic change many many years ago, 410 00:20:59,080 --> 00:21:01,439 Speaker 4: and one of their big focuses is I think like 411 00:21:01,480 --> 00:21:04,760 Speaker 4: Tesla autonomous driving, and no one has really been able 412 00:21:04,800 --> 00:21:07,280 Speaker 4: to crack that. I think that's sort of the AI 413 00:21:07,359 --> 00:21:11,440 Speaker 4: frontier that everyone's really focused on is how to solve vision, right, 414 00:21:11,440 --> 00:21:14,400 Speaker 4: because everyone's now focused on how to solve language, which 415 00:21:14,440 --> 00:21:16,520 Speaker 4: is what generative AI, and a lot of the products 416 00:21:16,560 --> 00:21:19,560 Speaker 4: we see today is kind of language based. But vision 417 00:21:19,640 --> 00:21:22,119 Speaker 4: is a really tough not to crack, and Baidu is 418 00:21:22,160 --> 00:21:24,679 Speaker 4: the one in China that's really been trying to solve it, 419 00:21:25,040 --> 00:21:27,199 Speaker 4: and I'm not sure their progress is any better than 420 00:21:27,280 --> 00:21:29,800 Speaker 4: Google or anybody else. But in terms of some of 421 00:21:29,840 --> 00:21:32,880 Speaker 4: the software companies like alibabacent has been doing a lot 422 00:21:32,880 --> 00:21:36,320 Speaker 4: of AI investments and obviously by Dance, so there's been 423 00:21:36,359 --> 00:21:39,240 Speaker 4: a lot of that. But what we're also seeing. We 424 00:21:39,320 --> 00:21:41,720 Speaker 4: did a recent piece where we looked at where Chinese 425 00:21:41,800 --> 00:21:45,280 Speaker 4: VC money has been going venture capital, whether venture capital 426 00:21:45,359 --> 00:21:47,680 Speaker 4: is going to a lot of these places, but in fact, 427 00:21:47,800 --> 00:21:51,560 Speaker 4: venture capital actually has been invested less in software in 428 00:21:51,600 --> 00:21:53,679 Speaker 4: the last few years, but actually you've invested more in 429 00:21:53,840 --> 00:21:57,800 Speaker 4: sort of hard tech hardware, so similar things like the 430 00:21:57,840 --> 00:22:00,879 Speaker 4: advanced manufacturing side. So I really think, you know, in 431 00:22:00,880 --> 00:22:02,520 Speaker 4: the next few years We're going to see a lot 432 00:22:02,520 --> 00:22:05,239 Speaker 4: of money, private and public going into sort of these 433 00:22:05,280 --> 00:22:08,320 Speaker 4: advanced manufacturing hard tech side of things that will have 434 00:22:08,359 --> 00:22:11,359 Speaker 4: AI applications. And I think there will be some startups 435 00:22:11,359 --> 00:22:13,960 Speaker 4: in China that probably we haven't heard of today that's 436 00:22:14,000 --> 00:22:15,679 Speaker 4: going to put a lot of money into AI. But 437 00:22:15,880 --> 00:22:17,600 Speaker 4: the big eyes are doing it. But by Do is 438 00:22:17,600 --> 00:22:20,840 Speaker 4: probably the one that's the most prominent in trying to 439 00:22:20,880 --> 00:22:24,560 Speaker 4: solve the sort of autonomous vision problem, and they will 440 00:22:24,560 --> 00:22:27,320 Speaker 4: be a big employer in China for sure for AI talent. 441 00:22:27,880 --> 00:22:31,560 Speaker 3: So going back to the other industrial applications of AI, 442 00:22:31,760 --> 00:22:36,000 Speaker 3: like already there's this just tremendous anxiety in the US 443 00:22:36,119 --> 00:22:39,000 Speaker 3: and Europe about whether there's any way to catch up 444 00:22:39,040 --> 00:22:42,800 Speaker 3: with China's sort of advanced manufacturing prowess, whether we're talking 445 00:22:42,800 --> 00:22:46,000 Speaker 3: about cars, whether we're talking about batteries, whether certainly whether 446 00:22:46,040 --> 00:22:48,720 Speaker 3: we're talking about certain types of chips. Should the US 447 00:22:48,760 --> 00:22:52,720 Speaker 3: be concerned perhaps that here chatbots are the shiny new thing, 448 00:22:52,760 --> 00:22:55,879 Speaker 3: and everyone wants to work on a better chatbot, And 449 00:22:55,960 --> 00:22:59,119 Speaker 3: in the meantime, China gets even better at sort of 450 00:22:59,200 --> 00:23:03,960 Speaker 3: automated factories. Particularly imagine with better vision technology that factory 451 00:23:04,000 --> 00:23:07,360 Speaker 3: floor robots could be safer, or could be more agile, etc. 452 00:23:08,200 --> 00:23:11,320 Speaker 3: Do you see a sort of like further widening of 453 00:23:11,359 --> 00:23:15,520 Speaker 3: the nature of the US China competition as a function 454 00:23:15,720 --> 00:23:17,679 Speaker 3: of where the AI talent has gone. 455 00:23:17,800 --> 00:23:20,280 Speaker 4: I'm not sure I can give you a very satisfying answer. 456 00:23:20,520 --> 00:23:23,800 Speaker 4: I guess the way I would think about that something 457 00:23:23,800 --> 00:23:26,520 Speaker 4: that would be emblematic of sort of both advanced manufacturing 458 00:23:26,520 --> 00:23:30,439 Speaker 4: and AI applications, sery software and hardware. I think the 459 00:23:30,520 --> 00:23:33,919 Speaker 4: key for both countries, and I think all countries is 460 00:23:33,960 --> 00:23:36,520 Speaker 4: probably going to be in robotics. That's sort of the 461 00:23:36,720 --> 00:23:40,240 Speaker 4: new frontier of whether it's the optimist humanoid robot China's 462 00:23:40,280 --> 00:23:44,160 Speaker 4: got I'm guessing like half a dozen robotics startups already. 463 00:23:44,520 --> 00:23:48,919 Speaker 4: So if one country, one company succeeds in that arena 464 00:23:49,280 --> 00:23:51,960 Speaker 4: and is able to really blend that hardware and software 465 00:23:51,960 --> 00:23:54,840 Speaker 4: and make it work and commercially viable, I think that 466 00:23:54,880 --> 00:23:58,040 Speaker 4: could send a lot of strong signals about the relative 467 00:23:58,080 --> 00:23:59,680 Speaker 4: capabilities of each country. 468 00:24:00,160 --> 00:24:03,119 Speaker 2: Are you going to start a robotics talent tracker? 469 00:24:03,560 --> 00:24:06,960 Speaker 4: Robots is that's going to involve a lot of supply chains, 470 00:24:07,000 --> 00:24:09,000 Speaker 4: So it's a little tougher than just looking at the people. 471 00:24:09,280 --> 00:24:10,639 Speaker 4: You got to bring in the chips. You got to 472 00:24:10,680 --> 00:24:14,480 Speaker 4: bring in the engineers, the mechanics. So it's more than 473 00:24:14,600 --> 00:24:18,080 Speaker 4: just EI scientists when it comes to robots. But interesting 474 00:24:18,080 --> 00:24:18,440 Speaker 4: for sure. 475 00:24:18,960 --> 00:24:22,280 Speaker 2: So one thing I wanted to ask, because you're looking 476 00:24:22,440 --> 00:24:25,679 Speaker 2: at this world very carefully and sort of watching what 477 00:24:25,720 --> 00:24:28,960 Speaker 2: people are doing and saying, But what is the language 478 00:24:29,040 --> 00:24:32,840 Speaker 2: that I guess policy makers in China are using around 479 00:24:33,000 --> 00:24:36,720 Speaker 2: AI talent, Like what sort of statements do you tend 480 00:24:36,760 --> 00:24:39,840 Speaker 2: to hear? And I'm thinking back again to that famous 481 00:24:39,880 --> 00:24:45,119 Speaker 2: disorderly capital expansion phrase that She Shinping deployed when he 482 00:24:45,240 --> 00:24:47,840 Speaker 2: was cracking down on things like the education sector and 483 00:24:48,000 --> 00:24:51,080 Speaker 2: consumer internet companies and stuff like that. But like, how 484 00:24:51,160 --> 00:24:56,480 Speaker 2: is this whole dynamic, this talent war couched in among 485 00:24:56,560 --> 00:24:57,400 Speaker 2: policy makers. 486 00:24:58,280 --> 00:25:00,399 Speaker 4: I think it's natural and it's given the you know, 487 00:25:00,440 --> 00:25:03,760 Speaker 4: no country generally likes brain drain. Everybody wants to have 488 00:25:03,800 --> 00:25:07,360 Speaker 4: brain gains, and I think you know that rhetoric aside 489 00:25:07,400 --> 00:25:09,840 Speaker 4: the actualization of that, And how do you set up 490 00:25:09,840 --> 00:25:11,680 Speaker 4: your own country, How do you set up the environment, 491 00:25:11,880 --> 00:25:15,919 Speaker 4: and you know, incentives, you know, compensation, all sorts of things. 492 00:25:16,359 --> 00:25:19,959 Speaker 4: The thing about top tier talent in any arena, but 493 00:25:20,000 --> 00:25:23,000 Speaker 4: particularly in computer science and these sort of frontier technologies. 494 00:25:23,680 --> 00:25:26,639 Speaker 4: Most of that talent, I would imagine would want to 495 00:25:26,680 --> 00:25:30,160 Speaker 4: be in the most competitive and dynamic industries. That's where 496 00:25:30,200 --> 00:25:32,520 Speaker 4: they probably feel the most comfortable. That's where they want 497 00:25:32,560 --> 00:25:34,800 Speaker 4: to make a difference with that's where they want to 498 00:25:34,800 --> 00:25:38,200 Speaker 4: make an impact and obviously the compensation all that stuff 499 00:25:38,240 --> 00:25:40,080 Speaker 4: follows that. But I think they want to have the 500 00:25:40,080 --> 00:25:43,359 Speaker 4: freedom to do the best cutting edge work possible. So 501 00:25:43,520 --> 00:25:46,800 Speaker 4: I think having dynamic industry is really important. And so 502 00:25:47,560 --> 00:25:50,399 Speaker 4: I'll bring the Europe example again. Europe doesn't seem to 503 00:25:50,440 --> 00:25:53,560 Speaker 4: have that, which is why they've consistently been sort of 504 00:25:53,640 --> 00:25:56,960 Speaker 4: underweighted when it comes to tracking top tier talent. And 505 00:25:57,160 --> 00:25:59,359 Speaker 4: if you look at the UK, which has been the 506 00:25:59,440 --> 00:26:02,440 Speaker 4: main place in Europe where most top tier AI talent work, 507 00:26:03,000 --> 00:26:05,840 Speaker 4: but in UK most of them work for Google DeepMind, 508 00:26:05,880 --> 00:26:09,560 Speaker 4: which is a US company. Right, having that industry is 509 00:26:09,600 --> 00:26:13,200 Speaker 4: I think really really important. And so in our current 510 00:26:13,200 --> 00:26:16,840 Speaker 4: debate about regulating AI and industry, I think it's going 511 00:26:16,840 --> 00:26:18,840 Speaker 4: to get controversial, it's going to get testy. We all 512 00:26:18,840 --> 00:26:20,679 Speaker 4: have known that, we all can see that, but I 513 00:26:20,720 --> 00:26:23,840 Speaker 4: think we have to think about, you know, if countries 514 00:26:23,920 --> 00:26:27,600 Speaker 4: want to attract the top tier talent. They want to 515 00:26:27,640 --> 00:26:29,919 Speaker 4: work in the most cutting edge, dynamic thing where they 516 00:26:29,920 --> 00:26:33,959 Speaker 4: can do the coolest, the most transformative stuff possible. And 517 00:26:34,000 --> 00:26:36,560 Speaker 4: if that's in America, great, But if China does that, 518 00:26:36,800 --> 00:26:39,840 Speaker 4: maybe it's China. But you know, right now, China still 519 00:26:39,960 --> 00:26:43,200 Speaker 4: mainly relies on its own domestic talent. They're not really 520 00:26:43,200 --> 00:26:46,360 Speaker 4: importing much foreign talent either. So to me, I think 521 00:26:46,400 --> 00:26:49,359 Speaker 4: having that industry is really really vital. 522 00:27:04,600 --> 00:27:07,600 Speaker 3: What are US universities doing. I imagine a high schooler 523 00:27:07,800 --> 00:27:11,600 Speaker 3: graduating in twenty twenty four, probably way more than four 524 00:27:11,680 --> 00:27:14,359 Speaker 3: years ago or even one year ago, are saying like, oh, yeah, well, 525 00:27:14,400 --> 00:27:15,520 Speaker 3: this is what I want to do. I want to 526 00:27:15,560 --> 00:27:18,800 Speaker 3: work in AI or something in this realm. Have we 527 00:27:18,960 --> 00:27:23,840 Speaker 3: seen an expansion of what US universities are offering or 528 00:27:24,040 --> 00:27:27,000 Speaker 3: capable of offering. Has there been that sort of supply 529 00:27:27,160 --> 00:27:30,639 Speaker 3: side capacity increase here to take advantage of what is 530 00:27:30,680 --> 00:27:33,080 Speaker 3: almost certain an increased interest in this industry. 531 00:27:33,240 --> 00:27:36,439 Speaker 4: Well, did you see the wsjpiece yesterday where all the 532 00:27:36,480 --> 00:27:39,040 Speaker 4: gen zs are becoming plumbers and electricians? Oh? 533 00:27:39,040 --> 00:27:42,000 Speaker 2: I did, Yeah, a return to trades. 534 00:27:42,080 --> 00:27:44,560 Speaker 4: Yeah. I mean, frankly, if I were anything, I might 535 00:27:44,560 --> 00:27:47,600 Speaker 4: consider that routes. But my understanding is that a lot 536 00:27:47,600 --> 00:27:50,280 Speaker 4: of the top tier technical schools or things that have 537 00:27:50,320 --> 00:27:55,960 Speaker 4: a technical school reputation, whether Stanford, cal Tech, Mit, Carnegie Mellon. 538 00:27:56,080 --> 00:27:59,280 Speaker 4: I mean, they definitely have AI programs. I don't know 539 00:27:59,280 --> 00:28:03,080 Speaker 4: if it's to d you know, extreme volume that China 540 00:28:03,119 --> 00:28:05,199 Speaker 4: has offered in a span of two or three years, 541 00:28:05,520 --> 00:28:09,160 Speaker 4: but they've definitely added those. But again, the foundation really 542 00:28:09,200 --> 00:28:11,520 Speaker 4: is computer science. So I think if you go in 543 00:28:11,560 --> 00:28:13,600 Speaker 4: and study computer science or some sort of you know, 544 00:28:13,840 --> 00:28:17,920 Speaker 4: you know, mathematics foundation, that's going to get you into 545 00:28:17,920 --> 00:28:20,960 Speaker 4: AI warming or another much easier than if you just 546 00:28:21,000 --> 00:28:23,720 Speaker 4: go straight into sort of you know AI, because you 547 00:28:23,800 --> 00:28:27,280 Speaker 4: can't really think about AI without having any foundational knowledge 548 00:28:27,320 --> 00:28:28,879 Speaker 4: from CS or mathematics. 549 00:28:29,480 --> 00:28:32,439 Speaker 2: This might be a weird question, but it's related to 550 00:28:32,480 --> 00:28:35,920 Speaker 2: the idea of people choosing to become plumbers or plasterers 551 00:28:36,280 --> 00:28:39,440 Speaker 2: or whatever it might be. Do you sense a sort 552 00:28:39,480 --> 00:28:43,960 Speaker 2: of like note of caution among potential graduates in the 553 00:28:44,040 --> 00:28:47,640 Speaker 2: sense that a lot of people in recent decades were 554 00:28:47,760 --> 00:28:53,120 Speaker 2: encouraged to go into coding and become fluent in Python 555 00:28:53,280 --> 00:28:57,000 Speaker 2: or Rust or whatever it might be. And now we've 556 00:28:57,040 --> 00:29:00,680 Speaker 2: seen the rise of AI, We've seen model that can 557 00:29:00,800 --> 00:29:04,080 Speaker 2: actually write your code for you pretty much, and a 558 00:29:04,080 --> 00:29:07,240 Speaker 2: lot of software engineers are currently a little bit worried 559 00:29:07,360 --> 00:29:11,240 Speaker 2: about their job security and the outlook for their skills. 560 00:29:12,080 --> 00:29:15,640 Speaker 2: Does that impact the potential AI talent pool at all? Like, 561 00:29:15,760 --> 00:29:17,680 Speaker 2: is there a sense that, okay, I can get into this, 562 00:29:17,880 --> 00:29:20,600 Speaker 2: but then maybe in ten or twenty years the AI 563 00:29:20,800 --> 00:29:23,680 Speaker 2: is just going to be developing itself. Right? Self learning 564 00:29:23,720 --> 00:29:26,600 Speaker 2: models are already a thing, So why get into it 565 00:29:26,640 --> 00:29:26,920 Speaker 2: at all? 566 00:29:27,240 --> 00:29:29,680 Speaker 4: Oh? Yeah, that's a tough question. Can AI be so 567 00:29:29,760 --> 00:29:32,080 Speaker 4: good that it doesn't need any human input anymore? 568 00:29:32,440 --> 00:29:34,920 Speaker 2: Again, I've been watching the three body problems, so a 569 00:29:34,920 --> 00:29:36,080 Speaker 2: little bit of a side pipe. 570 00:29:37,080 --> 00:29:39,760 Speaker 4: I don't know. I can't see that far into the future, 571 00:29:39,880 --> 00:29:41,880 Speaker 4: but what I will say, I guess kind of a 572 00:29:41,880 --> 00:29:45,240 Speaker 4: more realistic near term feature. I think we said earlier 573 00:29:45,280 --> 00:29:48,760 Speaker 4: that if AI is able to really solve human language, 574 00:29:48,760 --> 00:29:51,600 Speaker 4: which is obviously a big indicator of human intelligence, and 575 00:29:51,640 --> 00:29:53,400 Speaker 4: that seems to be a lot of the word the 576 00:29:53,400 --> 00:29:55,920 Speaker 4: efforts are large language models and you know, trying to 577 00:29:55,920 --> 00:30:00,080 Speaker 4: figure out how to mimic human language, human thought through language. 578 00:30:00,200 --> 00:30:02,680 Speaker 4: I would say one of the areas that's probably going 579 00:30:02,720 --> 00:30:05,560 Speaker 4: to be in trouble a lot is translators, that whole area, 580 00:30:05,600 --> 00:30:08,520 Speaker 4: it seems like it's going to be probably for lack 581 00:30:08,560 --> 00:30:10,760 Speaker 4: of a better term, disrupted quite a bit. Or if 582 00:30:10,800 --> 00:30:13,000 Speaker 4: you think about somebody that needs to do research in 583 00:30:13,040 --> 00:30:16,000 Speaker 4: different languages, maybe in two or three years, I can 584 00:30:16,040 --> 00:30:18,800 Speaker 4: read Japanese as easily as anyone else. Just get it 585 00:30:18,920 --> 00:30:21,840 Speaker 4: quickly translated on some AI software, and I can be 586 00:30:21,840 --> 00:30:25,680 Speaker 4: pretty fluent in reading Japanese. That doesn't mean you shouldn't 587 00:30:25,720 --> 00:30:27,440 Speaker 4: be studying foreign languages, so that there are a lot 588 00:30:27,440 --> 00:30:30,040 Speaker 4: of intellectual benefits to that, but I think as a 589 00:30:30,080 --> 00:30:34,080 Speaker 4: research tool and as the ability to kind of use 590 00:30:34,120 --> 00:30:37,560 Speaker 4: it as a way to understand the world. Once AI 591 00:30:37,640 --> 00:30:39,320 Speaker 4: really gets to that point, there are going to be 592 00:30:39,360 --> 00:30:42,600 Speaker 4: a lot of I think disciplines like translation, interpretation, those 593 00:30:42,640 --> 00:30:45,400 Speaker 4: kinds of things. It doesn't seem like there's going to 594 00:30:45,720 --> 00:30:47,640 Speaker 4: maybe be a huge need for that sort of stuff. 595 00:30:48,000 --> 00:30:50,360 Speaker 3: So in the earlier part of the conversation, you know, 596 00:30:50,400 --> 00:30:54,160 Speaker 3: we talked about three necessary components to have a domestic 597 00:30:54,240 --> 00:30:58,360 Speaker 3: AI industry. One is talent, one is sort of infrastructure, 598 00:30:58,360 --> 00:31:00,600 Speaker 3: and then the other one is just the pure compute. 599 00:31:00,640 --> 00:31:05,520 Speaker 3: And we see companies like Facebook, like they tout as 600 00:31:05,560 --> 00:31:09,520 Speaker 3: an advantage we just acquired so and so many h 601 00:31:09,640 --> 00:31:12,560 Speaker 3: one hundreds from in Nvidia, and we're spending ten billion dollars, 602 00:31:12,960 --> 00:31:16,640 Speaker 3: And I kind of get the impression that having a 603 00:31:16,640 --> 00:31:19,960 Speaker 3: lot of computing power is a recruiting tactic, and that 604 00:31:20,000 --> 00:31:22,920 Speaker 3: if you're a top AI researcher, you want to be 605 00:31:23,200 --> 00:31:26,320 Speaker 3: at the place that has the most advantage just sort 606 00:31:26,320 --> 00:31:30,120 Speaker 3: of raw computing capacity. We know that there's a lot 607 00:31:30,160 --> 00:31:33,680 Speaker 3: of restrictions on some of the cutting edge semiconductors going 608 00:31:33,720 --> 00:31:36,960 Speaker 3: into China, and Jensen Wong of Nvidia has talked about 609 00:31:36,960 --> 00:31:41,160 Speaker 3: this and the constraints there for a potential talented AI 610 00:31:41,240 --> 00:31:44,920 Speaker 3: researcher maybe from China or studied in China. Does that 611 00:31:45,040 --> 00:31:47,680 Speaker 3: factor into it the fact that, at least for now, 612 00:31:47,760 --> 00:31:52,160 Speaker 3: it looks like, still without question, that the US institutions, 613 00:31:52,200 --> 00:31:55,000 Speaker 3: whether we're talking about Meta, whether we're talking about Amazon, 614 00:31:55,440 --> 00:31:59,720 Speaker 3: Microsoft with OpenAI, have the most computing power to play with. 615 00:31:59,720 --> 00:32:01,200 Speaker 3: For lack of a better term. 616 00:32:01,520 --> 00:32:05,440 Speaker 4: That could certainly be one attractive factor. But I can't 617 00:32:05,480 --> 00:32:07,920 Speaker 4: remember where I read it, but I was shown like 618 00:32:07,960 --> 00:32:11,400 Speaker 4: an interesting survey on one of those Chinese social media 619 00:32:11,440 --> 00:32:14,120 Speaker 4: sites where apparently our a talent tractor got some traction 620 00:32:14,560 --> 00:32:17,080 Speaker 4: in Chinese and so a bunch of AI people in 621 00:32:17,160 --> 00:32:19,640 Speaker 4: China wade in and if I remember correctly, don't quote 622 00:32:19,680 --> 00:32:21,520 Speaker 4: me on it, but I think one of the main 623 00:32:21,560 --> 00:32:23,600 Speaker 4: things that stood out was that one of the things 624 00:32:23,640 --> 00:32:26,680 Speaker 4: that really attract that kind of talent is the research 625 00:32:26,800 --> 00:32:29,520 Speaker 4: environment where they're able to have the freedom and the 626 00:32:29,560 --> 00:32:32,000 Speaker 4: ability to have free thought and be able to, you know, 627 00:32:32,240 --> 00:32:34,320 Speaker 4: kind of pursue things that they think are really interesting, 628 00:32:34,360 --> 00:32:36,640 Speaker 4: that are really worthwhile. So that stood out to me 629 00:32:36,680 --> 00:32:40,440 Speaker 4: as a really important factor. And beyond the compute you know, 630 00:32:40,520 --> 00:32:44,520 Speaker 4: of prowess and beyond beyond compensation obviously, but I think 631 00:32:45,000 --> 00:32:47,280 Speaker 4: it seems like, you know, at least I think the 632 00:32:47,400 --> 00:32:50,000 Speaker 4: United States still seems to really have that you know, 633 00:32:50,080 --> 00:32:52,880 Speaker 4: culture like default, and I think that's a really important 634 00:32:53,040 --> 00:32:56,080 Speaker 4: ingredient that people shouldn't forget about. Again. I just think 635 00:32:56,120 --> 00:33:00,640 Speaker 4: top tier talent tend to want to be unencumbered, restricted 636 00:33:00,920 --> 00:33:02,880 Speaker 4: because they want to pursue things that they think are 637 00:33:02,920 --> 00:33:06,320 Speaker 4: really really, really interesting and groundbreaking, and that's just the 638 00:33:06,360 --> 00:33:08,640 Speaker 4: way they work, and so you got to give them 639 00:33:08,640 --> 00:33:09,920 Speaker 4: that environment to work in. 640 00:33:10,680 --> 00:33:14,080 Speaker 2: All right, Damien, that was such an interesting conversation. Thank 641 00:33:14,120 --> 00:33:16,440 Speaker 2: you so much for coming on odd lots and it 642 00:33:16,520 --> 00:33:19,600 Speaker 2: is the Global AI Talent Tracker, and you can look 643 00:33:19,600 --> 00:33:22,520 Speaker 2: it up online. It's got some really good charts and 644 00:33:22,600 --> 00:33:25,280 Speaker 2: sort of interactive elements that you can play around with. 645 00:33:25,400 --> 00:33:28,400 Speaker 2: So thanks Damian for coming on and walking us through 646 00:33:28,440 --> 00:33:29,760 Speaker 2: the latest work that you've been doing. 647 00:33:30,040 --> 00:33:30,760 Speaker 4: Thank you so much. 648 00:33:30,840 --> 00:33:44,240 Speaker 5: Great talking to you, Joe. 649 00:33:44,240 --> 00:33:47,680 Speaker 2: That conversation answered a lot of questions for me. It 650 00:33:47,760 --> 00:33:50,440 Speaker 2: was just interesting to talk about the patterns that we're 651 00:33:50,480 --> 00:33:53,120 Speaker 2: seeing play out. I think it's kind of funny that 652 00:33:53,240 --> 00:33:56,120 Speaker 2: in many ways, like this is a new technology that 653 00:33:56,200 --> 00:33:58,960 Speaker 2: everyone is excited about, but it's kind of playing out 654 00:33:59,040 --> 00:34:01,480 Speaker 2: the way a lot of stuff has played out historically, 655 00:34:01,480 --> 00:34:04,880 Speaker 2: where the US has a lead at the moment, and 656 00:34:04,920 --> 00:34:08,640 Speaker 2: then China is like rapidly on its heels and trying 657 00:34:08,640 --> 00:34:11,560 Speaker 2: to build out its own capacity, and then Europe is 658 00:34:11,640 --> 00:34:16,239 Speaker 2: like in the background, publishing like thought pieces and new 659 00:34:16,320 --> 00:34:18,000 Speaker 2: pieces of regulation about it. 660 00:34:18,000 --> 00:34:21,000 Speaker 3: It's kind of funny, It's it's exactly it's exactly right. 661 00:34:21,040 --> 00:34:24,759 Speaker 3: I'm really interested in this idea that you know. I 662 00:34:24,800 --> 00:34:26,799 Speaker 3: do think that in the US, if you say AI 663 00:34:26,920 --> 00:34:29,560 Speaker 3: at this point, either people think about the text generator, yes, 664 00:34:29,719 --> 00:34:33,160 Speaker 3: or the image generators, which are amazing, But this idea 665 00:34:33,200 --> 00:34:35,319 Speaker 3: and we've been and I think we're doing some more 666 00:34:35,360 --> 00:34:38,319 Speaker 3: episodes coming up on it, but like there's also a 667 00:34:38,320 --> 00:34:41,080 Speaker 3: lot of excitement that like there's more to AI than 668 00:34:41,239 --> 00:34:44,080 Speaker 3: just human language, And we talked about it a little 669 00:34:44,120 --> 00:34:46,719 Speaker 3: bit on the food Automation episode. The idea that like 670 00:34:47,000 --> 00:34:50,680 Speaker 3: if robots could sort of have the same framework where 671 00:34:50,680 --> 00:34:53,600 Speaker 3: they've had tons of data and then make better decisions 672 00:34:53,600 --> 00:34:56,239 Speaker 3: so the arms aren't swinging or slight deviation and on 673 00:34:56,320 --> 00:35:00,279 Speaker 3: the assembly line doesn't disrupt them, then you know, that 674 00:35:00,320 --> 00:35:03,439 Speaker 3: could be incredibly powerful if they had enough training data 675 00:35:03,480 --> 00:35:06,640 Speaker 3: about all of these different scenarios that they face. And 676 00:35:06,719 --> 00:35:09,680 Speaker 3: so it's interesting to see that China, which seems to 677 00:35:09,719 --> 00:35:11,920 Speaker 3: be you know, leading the world in many ways in 678 00:35:12,000 --> 00:35:15,920 Speaker 3: terms of sort of electrical engineering capacity, that's also in 679 00:35:16,000 --> 00:35:18,799 Speaker 3: alignment with where a lot of the AI researchers are going. 680 00:35:19,040 --> 00:35:21,040 Speaker 2: Yes, absolutely, and I know I brought it up a 681 00:35:21,120 --> 00:35:24,800 Speaker 2: number of times now, but that's why the consumer Internet 682 00:35:24,960 --> 00:35:28,920 Speaker 2: crackdown was so interesting to me, because China explicitly said, like, 683 00:35:29,200 --> 00:35:32,839 Speaker 2: we don't want all this money pouring into another new 684 00:35:33,000 --> 00:35:36,120 Speaker 2: online retailer. We have enough of those. Why don't you 685 00:35:36,120 --> 00:35:39,800 Speaker 2: take that money and invest it in chips or something 686 00:35:39,960 --> 00:35:43,399 Speaker 2: tangible like that, and so I do think we are 687 00:35:43,600 --> 00:35:47,760 Speaker 2: seeing that tendency right now that focus on like real 688 00:35:47,840 --> 00:35:53,399 Speaker 2: world applications, industrial applications, manufacturing that you don't necessarily see 689 00:35:53,440 --> 00:35:56,000 Speaker 2: in the US and other places in the West, because 690 00:35:56,480 --> 00:35:58,879 Speaker 2: as you know very well, Jo, it's fun to play 691 00:35:58,920 --> 00:36:01,840 Speaker 2: around with the chat bots and have become the public 692 00:36:01,880 --> 00:36:06,120 Speaker 2: face of this entire new technology. So that's probably one 693 00:36:06,200 --> 00:36:09,080 Speaker 2: area where China does have an advantage. But the other 694 00:36:09,120 --> 00:36:11,920 Speaker 2: thing I think so first of all, Damien talked about 695 00:36:11,960 --> 00:36:14,719 Speaker 2: the brain drain aspect of it and the idea that well, 696 00:36:14,800 --> 00:36:17,960 Speaker 2: a lot of China AI talent does end up in 697 00:36:18,000 --> 00:36:20,479 Speaker 2: the US because they go to university in the US 698 00:36:20,520 --> 00:36:23,320 Speaker 2: and then they stay there and there's demand for their services, 699 00:36:23,360 --> 00:36:26,040 Speaker 2: et cetera, et cetera, although maybe that will change soon. 700 00:36:26,560 --> 00:36:29,920 Speaker 2: But then the other thing I was thinking is you 701 00:36:30,000 --> 00:36:32,880 Speaker 2: brought up that question of compute power and whether or 702 00:36:32,960 --> 00:36:36,840 Speaker 2: not that's sort of a carrot for AI developers. I 703 00:36:36,880 --> 00:36:40,799 Speaker 2: also wonder about data and data restrictions in China and 704 00:36:40,840 --> 00:36:44,600 Speaker 2: what data sets they're playing around with, you know, specifically 705 00:36:44,600 --> 00:36:47,120 Speaker 2: for the large language models, but maybe for other things 706 00:36:47,400 --> 00:36:50,640 Speaker 2: as well. That could maybe be a competitive advantage if 707 00:36:50,719 --> 00:36:53,319 Speaker 2: you're really interested in this area, maybe you want to 708 00:36:53,360 --> 00:36:57,040 Speaker 2: go to a place that has bigger and more wide 709 00:36:57,160 --> 00:36:59,640 Speaker 2: ranging data sets like came in was kind of alluding 710 00:36:59,640 --> 00:37:00,560 Speaker 2: to totally. 711 00:37:01,040 --> 00:37:02,879 Speaker 3: The other thing I think is really important to watch. 712 00:37:03,000 --> 00:37:06,319 Speaker 3: I remember like twenty twenty five years ago, you know, 713 00:37:06,360 --> 00:37:08,320 Speaker 3: when the number of if you just looked at the 714 00:37:08,400 --> 00:37:12,880 Speaker 3: raw number of people graduating with an engineering degree, it 715 00:37:13,000 --> 00:37:15,120 Speaker 3: was like exploding in China, and there was a lot 716 00:37:15,120 --> 00:37:18,600 Speaker 3: of sneering and sort of Western publications it's like, oh, 717 00:37:18,600 --> 00:37:22,000 Speaker 3: these are trash degrees, Like, yeah, people graduate with a 718 00:37:22,000 --> 00:37:25,640 Speaker 3: degree in engineering, but it's like pretty mediocre talent and 719 00:37:25,719 --> 00:37:27,480 Speaker 3: you know, not really that good, and we sort of 720 00:37:27,480 --> 00:37:29,000 Speaker 3: have to take some of these numbers with a grain 721 00:37:29,040 --> 00:37:33,160 Speaker 3: of salt. I get the impression that's changed dramatically a 722 00:37:33,160 --> 00:37:35,279 Speaker 3: lot of these schools, and so the fact that you 723 00:37:35,320 --> 00:37:37,040 Speaker 3: know that there is you can sort of come up 724 00:37:37,080 --> 00:37:39,920 Speaker 3: with this subjective measure of talent, which is who gets 725 00:37:39,960 --> 00:37:42,879 Speaker 3: to speak at these big conferences, And if there is 726 00:37:43,080 --> 00:37:47,120 Speaker 3: a broadening out of the number of degree granting institutions 727 00:37:47,560 --> 00:37:50,319 Speaker 3: that are represented in that top two percent or top 728 00:37:50,360 --> 00:37:53,680 Speaker 3: twenty percent, that strikes me as like a very important 729 00:37:54,360 --> 00:37:57,279 Speaker 3: trend to watch and So these universities in China that 730 00:37:57,440 --> 00:37:59,160 Speaker 3: you know, I'm not familiar with any of them, but 731 00:37:59,200 --> 00:38:01,360 Speaker 3: if there's like, you know, beyond just the sort of 732 00:38:01,360 --> 00:38:05,799 Speaker 3: the equivalents of the MIT or Stanford are also contributing 733 00:38:06,040 --> 00:38:07,920 Speaker 3: to that elite, that strikes me as like a very 734 00:38:08,160 --> 00:38:09,280 Speaker 3: key indicator to watch. 735 00:38:09,400 --> 00:38:15,360 Speaker 2: Absolutely and neural information processing systems conference organizers, if you're listening, 736 00:38:15,760 --> 00:38:19,200 Speaker 2: Joe's interested in going, so send him an invite please. 737 00:38:19,320 --> 00:38:23,120 Speaker 3: Yeah, I'll demonstrate some of the great poems and songs. No, 738 00:38:23,200 --> 00:38:25,000 Speaker 3: I've done something I don't know and like I had to. 739 00:38:25,080 --> 00:38:27,040 Speaker 3: You know, AI come up with a new verb tense 740 00:38:27,080 --> 00:38:29,279 Speaker 3: for me is very impressive. So I come up with 741 00:38:29,280 --> 00:38:29,879 Speaker 3: creative stuff. 742 00:38:29,920 --> 00:38:32,200 Speaker 2: Oh that's interesting. You didn't tell me about that one. 743 00:38:32,360 --> 00:38:34,480 Speaker 3: I didn't want to bore you with all my it's 744 00:38:34,480 --> 00:38:36,239 Speaker 3: not boring, all right, all right, I'll show you. I'll 745 00:38:36,239 --> 00:38:36,799 Speaker 3: show you that one. 746 00:38:37,120 --> 00:38:38,640 Speaker 2: Have you started using Claude? 747 00:38:38,880 --> 00:38:39,080 Speaker 1: Yeah? 748 00:38:39,120 --> 00:38:39,960 Speaker 3: I love Claude. 749 00:38:40,000 --> 00:38:42,160 Speaker 2: It's better, right, there's something about it. 750 00:38:42,239 --> 00:38:45,040 Speaker 3: I don't know objectively about it. But this is also 751 00:38:45,040 --> 00:38:47,839 Speaker 3: another interesting question. So while we're talking about this, this 752 00:38:47,880 --> 00:38:50,560 Speaker 3: is like another interesting thing I'm wondering about, which is, 753 00:38:51,040 --> 00:38:53,120 Speaker 3: what if it turns out that some of the sort 754 00:38:53,160 --> 00:38:57,080 Speaker 3: of motes that we associate with software do not end 755 00:38:57,200 --> 00:39:00,600 Speaker 3: up applying as well to AI. Absolutely. Yeah, it's like 756 00:39:00,640 --> 00:39:02,920 Speaker 3: I like, for whatever reason, because I like the interface, 757 00:39:03,000 --> 00:39:06,000 Speaker 3: I like the way the nature of the language it speaks. 758 00:39:06,160 --> 00:39:08,640 Speaker 3: I started using Claude a lot more in a way 759 00:39:08,640 --> 00:39:10,640 Speaker 3: that I could never just imagine and say, like going 760 00:39:10,719 --> 00:39:13,400 Speaker 3: back and forth between Like once I used Google in 761 00:39:13,480 --> 00:39:15,680 Speaker 3: two thousand, I never like went back to Yahoo after that, 762 00:39:15,719 --> 00:39:17,680 Speaker 3: you know, or something like that. I've been using Google 763 00:39:17,719 --> 00:39:20,880 Speaker 3: ever since. It does make me wonder whether, like it'll 764 00:39:20,920 --> 00:39:24,360 Speaker 3: turn out that a lot of institutions with sufficient talent, 765 00:39:24,600 --> 00:39:27,399 Speaker 3: with sufficient compute can kind of do the same thing, 766 00:39:27,600 --> 00:39:29,319 Speaker 3: and switching costs aren't that high. 767 00:39:29,520 --> 00:39:31,839 Speaker 2: Yeah, I was wondering about this as well, because the 768 00:39:31,840 --> 00:39:35,200 Speaker 2: premise of this entire conversation was there's like a war 769 00:39:35,320 --> 00:39:38,440 Speaker 2: going on. People are trying to develop their AI capabilities 770 00:39:38,840 --> 00:39:42,879 Speaker 2: really fast because first one wins kind of. But it 771 00:39:42,960 --> 00:39:46,160 Speaker 2: does seem like some of these programs, like the motes 772 00:39:46,239 --> 00:39:49,040 Speaker 2: might not actually be that high, and once you crack 773 00:39:49,160 --> 00:39:52,960 Speaker 2: like one level, it might be kind of fungible in 774 00:39:53,040 --> 00:39:55,480 Speaker 2: other ways. I don't know. I guess it'll it'll be 775 00:39:55,520 --> 00:39:58,359 Speaker 2: interesting to see definitely all right, shall we leave it there? 776 00:39:58,400 --> 00:39:59,120 Speaker 3: Let's leave it there. 777 00:39:59,360 --> 00:40:02,000 Speaker 2: This has been an another episode of the aud Thoughts podcast. 778 00:40:02,040 --> 00:40:04,920 Speaker 2: I'm Tracy Alloway. You can follow me at Tracy Alloway. 779 00:40:05,239 --> 00:40:08,160 Speaker 3: And I'm Joe Wisenthal. You can follow me at the Stalwart. 780 00:40:08,360 --> 00:40:11,839 Speaker 3: Follow our guest Damien ma He's at Damian Nicks and 781 00:40:11,920 --> 00:40:15,880 Speaker 3: I'll also check out his AI talent tracker at macro Polo. 782 00:40:16,239 --> 00:40:19,319 Speaker 3: Follow our producers Carmen Rodriguez at Carmen armand dash Ol 783 00:40:19,360 --> 00:40:22,759 Speaker 3: Bennett at Dashbot and Kilbrooks at Kilbrooks. Thank you to 784 00:40:22,800 --> 00:40:25,759 Speaker 3: our producer Moses Onam. For more Oddlots content, go to 785 00:40:25,760 --> 00:40:28,400 Speaker 3: Bloomberg dot com slash odd Lots, where we have transcripts, 786 00:40:28,400 --> 00:40:30,759 Speaker 3: a blog in the newsletter and you can chat about 787 00:40:30,760 --> 00:40:32,600 Speaker 3: all of these topics twenty four to seven in the 788 00:40:32,680 --> 00:40:36,320 Speaker 3: discord Discord dot gg slash odlots. 789 00:40:36,080 --> 00:40:38,480 Speaker 2: And if you enjoy odlots, if you like it when 790 00:40:38,520 --> 00:40:42,200 Speaker 2: we have these conversations over artificial intelligence, if you want 791 00:40:42,320 --> 00:40:46,760 Speaker 2: a live demonstration of Joe's prompting of chat GPT or claude, 792 00:40:46,840 --> 00:40:49,440 Speaker 2: then please leave us a positive review on your favorite 793 00:40:49,480 --> 00:40:53,160 Speaker 2: podcast platform. And remember, if you are a Bloomberg subscriber, 794 00:40:53,239 --> 00:40:56,320 Speaker 2: you can listen to all of our episodes absolutely ad free. 795 00:40:56,440 --> 00:40:58,880 Speaker 2: All you need to do is connect your Bloomberg account 796 00:40:59,040 --> 00:41:01,400 Speaker 2: with Apple podcast Us. Thanks for listening.