1 00:00:01,280 --> 00:00:04,400 Speaker 1: Some of the biggest players in AI reported earnings last week. 2 00:00:04,559 --> 00:00:07,280 Speaker 1: Microsoft reported earnings right after the market closed, and we 3 00:00:07,360 --> 00:00:10,360 Speaker 1: did have a revenue beating analyssessments, but they didn't see 4 00:00:10,360 --> 00:00:13,120 Speaker 1: their shares jump on the good news. But the stock 5 00:00:13,320 --> 00:00:17,320 Speaker 1: is falling in after hours trading. Google's parent company, Alphabet, 6 00:00:17,480 --> 00:00:20,840 Speaker 1: saw its shares fall after it missed revenue expectations. 7 00:00:20,920 --> 00:00:22,920 Speaker 2: Alphabet, parent of Google, is down severely. 8 00:00:23,000 --> 00:00:27,080 Speaker 1: Alphabet less detail on really what contribution AI will have, 9 00:00:27,160 --> 00:00:30,000 Speaker 1: particularly on search. We're in the midst of earning season 10 00:00:30,080 --> 00:00:32,839 Speaker 1: for some of the biggest tech stocks, and falling shares 11 00:00:32,880 --> 00:00:35,680 Speaker 1: in some of these companies suggest that investors may be 12 00:00:35,760 --> 00:00:38,400 Speaker 1: disappointed by what these companies can deliver with AI. 13 00:00:38,800 --> 00:00:42,080 Speaker 3: There are real questions about the business about how these 14 00:00:42,120 --> 00:00:45,199 Speaker 3: new technologies, exciting as they are, cool as they are, 15 00:00:45,280 --> 00:00:47,520 Speaker 3: are going to make these companies money. 16 00:00:47,960 --> 00:00:50,919 Speaker 1: That's my colleague Max Chafkin, who reports on these companies. 17 00:00:51,120 --> 00:00:54,320 Speaker 3: What I think people are starting to realize is that sure, 18 00:00:54,480 --> 00:00:56,920 Speaker 3: you can be very bought in into the promise of AI, 19 00:00:57,880 --> 00:00:59,760 Speaker 3: but even so, it's going to be a slog. 20 00:01:02,920 --> 00:01:06,000 Speaker 1: On today's episode, we dig into the expectations for AI 21 00:01:06,319 --> 00:01:10,440 Speaker 1: versus the reality why these companies are failing to meet 22 00:01:10,520 --> 00:01:13,520 Speaker 1: some investors' hopes, and what it would take for these 23 00:01:13,560 --> 00:01:17,039 Speaker 1: technologies to catch up to their promise. I'm your host, 24 00:01:17,120 --> 00:01:26,200 Speaker 1: Sarah Holder, and this is big take from Bloomberg News. 25 00:01:27,319 --> 00:01:28,560 Speaker 2: My name's Max Chafkin. 26 00:01:28,800 --> 00:01:31,959 Speaker 3: I am a senior reporter with Bloomberg BusinessWeek and I 27 00:01:32,000 --> 00:01:37,600 Speaker 3: cover technology, particularly kind of the intersection of technology and power, 28 00:01:37,800 --> 00:01:39,800 Speaker 3: and I also co host the Elonink podcast. 29 00:01:40,400 --> 00:01:42,120 Speaker 1: I know you've written a lot about sort of hype 30 00:01:42,120 --> 00:01:47,319 Speaker 1: cycles in bitcoin and cryptocurrency, and so interested to talk 31 00:01:47,360 --> 00:01:51,520 Speaker 1: about a new hype cycle today. Artificial intelligence has been 32 00:01:51,720 --> 00:01:55,360 Speaker 1: making headlines for the supposedly vast potential of the technology. 33 00:01:55,760 --> 00:01:59,920 Speaker 1: What kinds of expectations have technology leaders set about what 34 00:02:00,120 --> 00:02:03,040 Speaker 1: they're going to do with this AI capacity. 35 00:02:03,400 --> 00:02:06,000 Speaker 3: It's so funny you say the expectations. I mean, because 36 00:02:06,000 --> 00:02:10,520 Speaker 3: the expectations are wild, like the craziest things you can 37 00:02:10,560 --> 00:02:15,360 Speaker 3: possibly imagine. So Sam Altman, who is the founder a 38 00:02:15,400 --> 00:02:19,840 Speaker 3: CEO of Open Ai, has talked about what his company 39 00:02:19,880 --> 00:02:23,840 Speaker 3: is doing as you know, basically modern day you know, Oppenheimer, 40 00:02:24,000 --> 00:02:28,080 Speaker 3: This is like as significant as potentially destructive as the 41 00:02:28,080 --> 00:02:28,799 Speaker 3: atom bomb. 42 00:02:28,880 --> 00:02:31,280 Speaker 2: It's a big part of why I'm here today and 43 00:02:31,320 --> 00:02:32,440 Speaker 2: why we've been here in the past. 44 00:02:32,680 --> 00:02:36,280 Speaker 1: Here's open AI CEO Sam Altman speaking before Congress in 45 00:02:36,360 --> 00:02:37,200 Speaker 1: May of last year. 46 00:02:37,440 --> 00:02:40,040 Speaker 3: I think if this technology goes wrong, it can go 47 00:02:40,160 --> 00:02:43,640 Speaker 3: quite wrong, and we want to be vocal about that. 48 00:02:43,680 --> 00:02:46,080 Speaker 2: We want to work with the government to prevent that 49 00:02:46,120 --> 00:02:46,680 Speaker 2: from happening. 50 00:02:47,040 --> 00:02:50,680 Speaker 3: What's interesting to me about these warnings, and I think 51 00:02:50,680 --> 00:02:53,320 Speaker 3: it's worth taking, you know, any warning about technology. 52 00:02:53,320 --> 00:02:56,280 Speaker 2: Seriously. Technologies can have unintended consequences, but. 53 00:02:56,200 --> 00:02:59,120 Speaker 3: They also serve, of course as a sales pitch, because 54 00:02:59,160 --> 00:03:02,160 Speaker 3: if you're going around saying that, hey, this thing I'm 55 00:03:02,200 --> 00:03:02,960 Speaker 3: building is. 56 00:03:02,960 --> 00:03:05,200 Speaker 2: So effective it could take. 57 00:03:05,000 --> 00:03:07,960 Speaker 3: Over the world, it could render human beings irrelevant, that 58 00:03:08,080 --> 00:03:10,960 Speaker 3: is another way of saying, like the thing I'm building works. 59 00:03:11,280 --> 00:03:14,040 Speaker 3: And in technology, you know, over my career, I feel 60 00:03:14,040 --> 00:03:16,519 Speaker 3: like a question like nobody asks enough, is like does 61 00:03:16,560 --> 00:03:19,919 Speaker 3: it work? And one of the reasons is because technologists 62 00:03:19,919 --> 00:03:21,679 Speaker 3: are very good at kind of obscuring that question and 63 00:03:21,760 --> 00:03:25,560 Speaker 3: raising other questions, including does this technology potentially destroy the world. 64 00:03:25,600 --> 00:03:31,320 Speaker 1: So even the terrifying expectations are inherently potentially overly positive 65 00:03:31,919 --> 00:03:33,200 Speaker 1: about the technology itself. 66 00:03:33,720 --> 00:03:34,680 Speaker 2: Absolutely, you know. 67 00:03:34,800 --> 00:03:38,240 Speaker 3: On the Tesla earnings call, my favorite moment was when 68 00:03:38,480 --> 00:03:40,960 Speaker 3: Elon Musk is sort of spinning this story about Tesla's 69 00:03:40,960 --> 00:03:44,960 Speaker 3: AI investments. He's saying, optimists, their robot is going to 70 00:03:45,760 --> 00:03:48,880 Speaker 3: be the greatest product of all time, better than any 71 00:03:48,920 --> 00:03:51,320 Speaker 3: product in human history, and it's actually going to like 72 00:03:51,400 --> 00:03:55,200 Speaker 3: usher in a world of unlimited productivity. Essentially, it's going 73 00:03:55,240 --> 00:03:57,040 Speaker 3: to be like in the movie Wally, we can all 74 00:03:57,080 --> 00:03:59,120 Speaker 3: just like sit back and let the robots. 75 00:03:58,880 --> 00:03:59,720 Speaker 2: Do all our work. 76 00:04:00,160 --> 00:04:03,320 Speaker 3: And while he's saying this, another executive on the Tesla 77 00:04:03,360 --> 00:04:07,280 Speaker 3: call cuts in and says, the only issue is we 78 00:04:07,360 --> 00:04:10,800 Speaker 3: have not found actual use for these robots to do 79 00:04:10,840 --> 00:04:13,640 Speaker 3: where that it works. And you got to remember, like 80 00:04:14,160 --> 00:04:16,400 Speaker 3: a lot of the people who buy software are not 81 00:04:17,080 --> 00:04:20,200 Speaker 3: worried about apocalyptic scenarios. They're just trying to figure out 82 00:04:20,240 --> 00:04:21,960 Speaker 3: is the return on investment for this going to be 83 00:04:22,120 --> 00:04:25,360 Speaker 3: worth it? If I'm a chief technology offerer, chief information 84 00:04:25,400 --> 00:04:28,159 Speaker 3: officer for a large company, I don't think I'm mostly 85 00:04:28,200 --> 00:04:31,120 Speaker 3: thinking about like what's the long term societal impact of 86 00:04:31,160 --> 00:04:33,440 Speaker 3: this giant check? I'm going to write Microsoft or open AI. 87 00:04:33,520 --> 00:04:35,960 Speaker 3: I'm thinking, is it going to return? Am I going 88 00:04:35,960 --> 00:04:38,719 Speaker 3: to look like a moron to my CEO for writing 89 00:04:38,760 --> 00:04:39,240 Speaker 3: this check? 90 00:04:39,760 --> 00:04:39,960 Speaker 2: Right? 91 00:04:40,680 --> 00:04:45,200 Speaker 1: So Microsoft alphabet Apple, Meta have all been kind of 92 00:04:45,279 --> 00:04:49,239 Speaker 1: riding high on these expectations. But can you talk about 93 00:04:49,240 --> 00:04:52,119 Speaker 1: what we saw in earnings this past week? 94 00:04:52,360 --> 00:04:54,800 Speaker 3: All right, So just take a step back for a second. 95 00:04:55,240 --> 00:04:57,680 Speaker 3: There was kind of a hangover after the pandemic, right 96 00:04:57,800 --> 00:05:02,400 Speaker 3: where a lot of technology company had invested huge amounts 97 00:05:02,400 --> 00:05:04,560 Speaker 3: of money, sort of assuming that the growth that was 98 00:05:04,560 --> 00:05:07,200 Speaker 3: happening because we're all sitting at home doing nothing, but 99 00:05:07,320 --> 00:05:09,360 Speaker 3: like you know, scrolling our phones was going to go 100 00:05:09,400 --> 00:05:11,880 Speaker 3: on forever, and they all kind of, in one way 101 00:05:11,920 --> 00:05:15,320 Speaker 3: or another, went into a dip. And when open ai 102 00:05:16,200 --> 00:05:20,039 Speaker 3: introduced chatch Ept, it essentially set the world on fire 103 00:05:20,440 --> 00:05:25,560 Speaker 3: because it promised a new platform, a new thing to 104 00:05:25,600 --> 00:05:28,200 Speaker 3: get people excited about. And all the companies that you 105 00:05:28,279 --> 00:05:31,680 Speaker 3: mentioned in one way or another have investments in AI. 106 00:05:32,320 --> 00:05:36,200 Speaker 3: And what's happened is, i'd say, just you know, a 107 00:05:36,240 --> 00:05:41,560 Speaker 3: reality check. Essentially, Google reported earnings. It was not even 108 00:05:41,600 --> 00:05:43,840 Speaker 3: like it was that bad. It just seems like the 109 00:05:44,360 --> 00:05:49,160 Speaker 3: view from investors is like, WHOA, this stuff is going 110 00:05:49,520 --> 00:05:53,360 Speaker 3: to take longer and be harder than maybe we had hoped. 111 00:05:53,720 --> 00:05:58,040 Speaker 3: And you know, with Google, they were until very recently 112 00:05:58,279 --> 00:06:02,080 Speaker 3: the market leader in AI, and then open Ai comes 113 00:06:02,120 --> 00:06:03,960 Speaker 3: out kind of steals a step on them. There's this 114 00:06:04,080 --> 00:06:07,200 Speaker 3: kind of sudden like, oh no, like maybe there's another player. 115 00:06:08,240 --> 00:06:12,839 Speaker 3: But Google still has like all of these very impressive 116 00:06:13,560 --> 00:06:14,919 Speaker 3: pieces of AI software. 117 00:06:15,440 --> 00:06:18,200 Speaker 2: But the thing is that as impressive as. 118 00:06:18,080 --> 00:06:20,600 Speaker 3: They are, they're all in one way or another kind 119 00:06:20,600 --> 00:06:24,800 Speaker 3: of research projects. They're not really commercial projects yet. And 120 00:06:25,000 --> 00:06:29,280 Speaker 3: what is a commercial product is you know, search advertising, 121 00:06:29,320 --> 00:06:32,680 Speaker 3: this business that is getting mature. And this is kind 122 00:06:32,720 --> 00:06:35,320 Speaker 3: of a dynamic with most of the big tech companies, 123 00:06:35,360 --> 00:06:40,159 Speaker 3: where Google, Facebook, and Apple kind of dependent on a 124 00:06:40,200 --> 00:06:43,760 Speaker 3: single mature or maybe a little long in the tooth product. 125 00:06:44,520 --> 00:06:48,359 Speaker 3: Microsoft actually, you know, Microsoft reported earnings and did not 126 00:06:48,480 --> 00:06:51,080 Speaker 3: see its stock crashed down and the same with the 127 00:06:51,080 --> 00:06:54,039 Speaker 3: same kind of velocity as Google's. And that's because I 128 00:06:54,040 --> 00:06:57,000 Speaker 3: think Microsoft is actually in the best position here because 129 00:06:57,200 --> 00:07:01,000 Speaker 3: Microsoft has a way to turn in the AI into 130 00:07:01,040 --> 00:07:04,599 Speaker 3: actual dollars, both because it's the market leader in selling 131 00:07:05,200 --> 00:07:08,520 Speaker 3: cloud computing for AI, and Microsoft also has been very 132 00:07:08,520 --> 00:07:13,280 Speaker 3: aggressively incorporating AI software in the form of open Ai, 133 00:07:13,320 --> 00:07:16,600 Speaker 3: which has a partnership with into its office suite and 134 00:07:16,680 --> 00:07:18,320 Speaker 3: charging actual money for it. 135 00:07:18,840 --> 00:07:21,160 Speaker 1: This is the question at the heart of the disconnect 136 00:07:21,240 --> 00:07:25,400 Speaker 1: between tech companies and investors. Can these companies make money 137 00:07:25,480 --> 00:07:28,320 Speaker 1: off of the technology, and can they do it before 138 00:07:28,360 --> 00:07:30,560 Speaker 1: investors sour on the project entirely. 139 00:07:32,080 --> 00:07:35,320 Speaker 3: I mean, I think it's tough to read into any 140 00:07:36,200 --> 00:07:39,640 Speaker 3: small stock movement and try to see something big about 141 00:07:39,640 --> 00:07:42,840 Speaker 3: the future because in a lot of cases it's not 142 00:07:42,880 --> 00:07:46,440 Speaker 3: necessarily that, oh, like Google stock went down. That doesn't 143 00:07:46,480 --> 00:07:49,280 Speaker 3: mean that their AI doesn't work. It means two things. 144 00:07:49,320 --> 00:07:53,720 Speaker 3: It means one is that investors maybe overestimated how quickly 145 00:07:53,920 --> 00:07:55,200 Speaker 3: they were going to be able. 146 00:07:54,960 --> 00:07:56,880 Speaker 2: To turn this technology into business. 147 00:07:57,080 --> 00:07:59,440 Speaker 3: And also it says something about their legacy businesses right 148 00:08:00,120 --> 00:08:03,040 Speaker 3: search ads, and their ability to continue to make money 149 00:08:03,040 --> 00:08:05,080 Speaker 3: and their ability to cut costs. I do think the 150 00:08:05,160 --> 00:08:07,960 Speaker 3: company that's worth paying the most attention to in this 151 00:08:08,040 --> 00:08:12,040 Speaker 3: world is Microsoft. Microsoft actually has a plan, like there's 152 00:08:12,080 --> 00:08:15,160 Speaker 3: a real clear sort of way that they are going 153 00:08:15,200 --> 00:08:18,120 Speaker 3: to make money off of this, in contrast right to 154 00:08:18,200 --> 00:08:23,160 Speaker 3: Facebook and Google, and what we're seeing with their business 155 00:08:23,440 --> 00:08:24,600 Speaker 3: is that it's growing. 156 00:08:24,760 --> 00:08:25,920 Speaker 2: It's very good. 157 00:08:26,360 --> 00:08:29,960 Speaker 3: It's just not totally clear how fast it's going to 158 00:08:30,040 --> 00:08:33,680 Speaker 3: grow or how big it's going to be. Large language models, 159 00:08:33,720 --> 00:08:35,800 Speaker 3: which is what most of the stuff that everyone's excited 160 00:08:35,800 --> 00:08:38,920 Speaker 3: about is they cost a lot of money to make, 161 00:08:39,280 --> 00:08:41,199 Speaker 3: and every time you update them, they cost a lot 162 00:08:41,240 --> 00:08:41,640 Speaker 3: of money. 163 00:08:41,880 --> 00:08:45,160 Speaker 2: Every time you ask chat GPT, write an. 164 00:08:45,040 --> 00:08:48,120 Speaker 3: Angry letter to a friend in the style of Chaucer, 165 00:08:48,440 --> 00:08:52,520 Speaker 3: or like some fun thing you're costing Microsoft money and 166 00:08:52,600 --> 00:08:54,439 Speaker 3: way more money, or if you ask them, if you 167 00:08:54,480 --> 00:08:56,959 Speaker 3: ask them like a Google query, Hey, what's the best 168 00:08:56,960 --> 00:08:59,640 Speaker 3: pizza restaurant in New York? Or something that costs a 169 00:08:59,679 --> 00:09:02,680 Speaker 3: lot more money to answer than it does when you 170 00:09:02,800 --> 00:09:06,760 Speaker 3: Google it, and there's not like an easy way to 171 00:09:07,040 --> 00:09:08,760 Speaker 3: monetize it, Like they're not going to be as silly 172 00:09:08,800 --> 00:09:11,719 Speaker 3: and little ads for pizza next to it. So I 173 00:09:11,760 --> 00:09:16,400 Speaker 3: think there's like real questions, hard questions about the business. 174 00:09:16,960 --> 00:09:19,959 Speaker 1: When we return one company in the AI space, and 175 00:09:20,120 --> 00:09:23,440 Speaker 1: Vidia has bucked this trend, we dig into what makes 176 00:09:23,480 --> 00:09:26,199 Speaker 1: Nvidia different and we'll get into what this all says 177 00:09:26,360 --> 00:09:36,600 Speaker 1: about the maturity of the AI industry. Welcome back. I'm 178 00:09:36,600 --> 00:09:39,800 Speaker 1: discussing the AI industry with my colleague BusinessWeek reporter and 179 00:09:39,920 --> 00:09:44,720 Speaker 1: editor Max Chafkin. So Microsoft has this potential. We're going 180 00:09:44,760 --> 00:09:48,400 Speaker 1: to see how it realizes it. Other companies are also 181 00:09:48,760 --> 00:09:53,320 Speaker 1: kind of trying to make strides in the AI space, 182 00:09:53,679 --> 00:09:56,280 Speaker 1: but Navidia seems to be one of the few that's 183 00:09:56,320 --> 00:10:00,360 Speaker 1: actually making money on AI right now, what's going right 184 00:10:00,480 --> 00:10:01,319 Speaker 1: for Navidia. 185 00:10:01,400 --> 00:10:04,240 Speaker 3: First of all, this is a company that all these 186 00:10:04,280 --> 00:10:08,400 Speaker 3: companies need. They make these GPUs that the entire industry 187 00:10:09,000 --> 00:10:09,640 Speaker 3: depends on. 188 00:10:10,080 --> 00:10:12,640 Speaker 2: A GPU is a graphical processing unit. 189 00:10:13,080 --> 00:10:16,600 Speaker 3: You can use a CPU, a central processing unit, that's 190 00:10:16,600 --> 00:10:19,000 Speaker 3: the kind of chip that your computer is running on, 191 00:10:19,920 --> 00:10:21,800 Speaker 3: but it goes a lot slower and it uses a 192 00:10:21,800 --> 00:10:24,080 Speaker 3: lot more power. So if you're using GPUs, you can 193 00:10:24,400 --> 00:10:27,920 Speaker 3: do AI training much more efficiently and you can sort 194 00:10:27,960 --> 00:10:30,719 Speaker 3: of answer AI queries more efficiently. When you look at 195 00:10:30,840 --> 00:10:34,840 Speaker 3: how the other big like the big companies that are 196 00:10:34,840 --> 00:10:37,920 Speaker 3: trying to develop AI, all they think about is how 197 00:10:37,960 --> 00:10:41,319 Speaker 3: do we how do we have enough chip capacity. Elon 198 00:10:41,400 --> 00:10:44,160 Speaker 3: Musk is developing his sort of like his own like 199 00:10:44,280 --> 00:10:45,840 Speaker 3: AI computing solution. 200 00:10:46,800 --> 00:10:51,880 Speaker 2: It's called Dojo, and on Tesla's earnings said, well, Dojo 201 00:10:52,000 --> 00:10:54,760 Speaker 2: is a real long shot. It may not pay off. 202 00:10:54,800 --> 00:10:56,520 Speaker 3: Which again, if Elon Muskay is somebody's not going to 203 00:10:56,600 --> 00:10:59,080 Speaker 3: pay off, then like that that means that there's a 204 00:10:59,080 --> 00:11:00,720 Speaker 3: real chance I don't want pay Because he's a very 205 00:11:00,720 --> 00:11:03,679 Speaker 3: optimistic guy, and he said, you know, he's still betting 206 00:11:03,720 --> 00:11:08,120 Speaker 3: on Nvidia, everybody is, and so in certain ways with Nvidia, 207 00:11:08,480 --> 00:11:11,280 Speaker 3: it doesn't really matter at least for now, like how 208 00:11:11,320 --> 00:11:13,720 Speaker 3: big a business AI is, because all these companies have 209 00:11:13,800 --> 00:11:16,560 Speaker 3: decided they're in on the mania. There's this thought, you know, 210 00:11:16,600 --> 00:11:19,160 Speaker 3: in a gold rush and this better shell sell shovels. 211 00:11:19,200 --> 00:11:21,439 Speaker 3: Then you know, try to like pan for gold they 212 00:11:21,480 --> 00:11:24,760 Speaker 3: need in Nvidia's shovels, and Nvidia is like right, selling 213 00:11:24,840 --> 00:11:27,679 Speaker 3: equipment for this mania, and so that's always going to 214 00:11:27,720 --> 00:11:28,680 Speaker 3: be a good business. 215 00:11:29,000 --> 00:11:31,160 Speaker 1: I love how you put it that Navidia is selling 216 00:11:31,200 --> 00:11:32,800 Speaker 1: shovels in a gold rush. 217 00:11:32,840 --> 00:11:33,000 Speaker 3: You know. 218 00:11:33,040 --> 00:11:35,760 Speaker 2: Funnily enough, you need GPUs to mind crypto. 219 00:11:36,120 --> 00:11:38,360 Speaker 3: So like they rote that boom and now they've they've 220 00:11:38,440 --> 00:11:39,839 Speaker 3: kind of moved on to this next one. 221 00:11:40,040 --> 00:11:42,800 Speaker 1: I'm interested in what this conversation we're having about kind 222 00:11:42,840 --> 00:11:46,000 Speaker 1: of the winners and the losers and the people in 223 00:11:46,040 --> 00:11:48,840 Speaker 1: the middle of the race in the AI sphere says 224 00:11:48,960 --> 00:11:51,839 Speaker 1: about where we are in the life cycle of the 225 00:11:51,880 --> 00:11:52,880 Speaker 1: AI industry. 226 00:11:53,240 --> 00:11:55,120 Speaker 2: Yeah, you know, it's funny. I've been thinking. 227 00:11:55,160 --> 00:11:59,400 Speaker 3: I've written a lot about self driving cars and and 228 00:11:59,520 --> 00:12:04,079 Speaker 3: because of that, I'm probably more skeptical or at least 229 00:12:04,080 --> 00:12:07,200 Speaker 3: a little bit more cautious about these large language models. 230 00:12:07,559 --> 00:12:11,160 Speaker 3: If you cast your mind back to twenty twelve or 231 00:12:11,160 --> 00:12:14,960 Speaker 3: twenty thirteen or twenty fourteen, the conversation around. 232 00:12:14,640 --> 00:12:15,640 Speaker 2: Cars was very similar. 233 00:12:15,679 --> 00:12:17,600 Speaker 3: It was like, look at these things, they're like almost 234 00:12:17,640 --> 00:12:19,320 Speaker 3: as good as a human and we kind of see 235 00:12:19,320 --> 00:12:22,839 Speaker 3: that a similar conversation happening with large language models, where 236 00:12:22,840 --> 00:12:25,360 Speaker 3: you ask them a question. You ask it to say. 237 00:12:25,400 --> 00:12:27,280 Speaker 3: You can ask it to describe yourself. Right, you ask 238 00:12:27,320 --> 00:12:29,560 Speaker 3: it like who is Max Chafkin? And and I haven't 239 00:12:29,559 --> 00:12:31,560 Speaker 3: done this in a while, but right, it makes up 240 00:12:31,559 --> 00:12:33,960 Speaker 3: a lot of stuff. Right, it's like pretty good. It's close, 241 00:12:34,040 --> 00:12:34,480 Speaker 3: but it's. 242 00:12:34,320 --> 00:12:35,000 Speaker 2: Not quite right. 243 00:12:35,080 --> 00:12:36,520 Speaker 3: And it's got you know, it's got me going to 244 00:12:36,559 --> 00:12:39,680 Speaker 3: a different college, and I watch with some on or 245 00:12:39,720 --> 00:12:45,600 Speaker 3: these like beautiful photorealistic images and then you look closely, 246 00:12:45,640 --> 00:12:47,959 Speaker 3: it's got six fingers. If you were talking to a human, 247 00:12:48,040 --> 00:12:52,320 Speaker 3: you could just be like, hey, artist, actually human beings 248 00:12:52,360 --> 00:12:55,560 Speaker 3: have five fingers, so in the future, just make the 249 00:12:55,640 --> 00:12:56,400 Speaker 3: hands with five. 250 00:12:56,679 --> 00:12:58,480 Speaker 2: But AI, these AI. 251 00:12:58,320 --> 00:13:00,960 Speaker 3: Models, you cannot have that conversation with them. They don't 252 00:13:01,040 --> 00:13:02,920 Speaker 3: understand what a finger is. They don't understand what a 253 00:13:03,000 --> 00:13:05,960 Speaker 3: hand is, and no one knows. Just like with the cars, 254 00:13:06,280 --> 00:13:08,840 Speaker 3: no one knows what it's going to take to cross 255 00:13:08,880 --> 00:13:12,840 Speaker 3: that bridge. There are people on smart people think to 256 00:13:12,960 --> 00:13:16,880 Speaker 3: really cross that bridge, you would need an entirely new technology, 257 00:13:17,120 --> 00:13:19,560 Speaker 3: and that could mean new winners and new losers and 258 00:13:19,640 --> 00:13:23,560 Speaker 3: you know and so on, new chips. And the other 259 00:13:23,600 --> 00:13:28,480 Speaker 3: thing is, you know, cost really matters. Like weimo right now, 260 00:13:28,880 --> 00:13:32,840 Speaker 3: the Google Driver's car company is doing all these tests 261 00:13:32,840 --> 00:13:36,160 Speaker 3: in Phoenix and in other localities, and in a lot 262 00:13:36,200 --> 00:13:38,840 Speaker 3: of ways like their cars are matching up to human drivers, 263 00:13:39,360 --> 00:13:42,960 Speaker 3: but except in a very important way they're not, which 264 00:13:43,000 --> 00:13:46,320 Speaker 3: is that these cars have you know, potentially hundreds of 265 00:13:46,360 --> 00:13:49,080 Speaker 3: thousands of dollars of censors on them, and they're powered 266 00:13:49,080 --> 00:13:50,800 Speaker 3: by armies of PhDs. 267 00:13:51,320 --> 00:13:54,880 Speaker 2: And your Uber is a you know, twenty. 268 00:13:54,720 --> 00:13:57,640 Speaker 3: Five thousand dollars Corolla and a gig worker, and you 269 00:13:57,720 --> 00:14:00,400 Speaker 3: have some of the same dynamics with these large language 270 00:14:00,440 --> 00:14:04,080 Speaker 3: models where like wow, like this sound this is almost 271 00:14:04,160 --> 00:14:06,480 Speaker 3: as good as somebody in a call center could do 272 00:14:06,559 --> 00:14:09,920 Speaker 3: on customer service, But we're not actually sure that it 273 00:14:10,040 --> 00:14:14,960 Speaker 3: costs less than hiring a human being to do customer service. 274 00:14:15,520 --> 00:14:18,160 Speaker 1: We saw a lot of AI companies pop up begin 275 00:14:18,200 --> 00:14:20,600 Speaker 1: to open their products to the public. Are we on 276 00:14:20,600 --> 00:14:23,880 Speaker 1: the brink of seeing contraction in the AI industry. 277 00:14:23,960 --> 00:14:27,280 Speaker 3: I think, you know, it's it's very hard to like 278 00:14:27,400 --> 00:14:29,440 Speaker 3: call a bubble, and like if you go back to 279 00:14:29,920 --> 00:14:31,800 Speaker 3: if you like you go back to other example like 280 00:14:31,840 --> 00:14:34,040 Speaker 3: Crypto right like, people were saying it was a bubble 281 00:14:34,600 --> 00:14:39,920 Speaker 3: like for years before it became apparent. So it's it's 282 00:14:39,960 --> 00:14:43,320 Speaker 3: a little hard to know how valuable these things are 283 00:14:43,360 --> 00:14:45,680 Speaker 3: because it seems like they may be everywhere, you know. 284 00:14:46,160 --> 00:14:49,600 Speaker 3: So the sort of like worst case scenario here is 285 00:14:49,880 --> 00:14:52,320 Speaker 3: consumers just get kind of used to this, but they're 286 00:14:52,320 --> 00:14:54,480 Speaker 3: not willing to pay for it because they're versions of 287 00:14:54,520 --> 00:14:55,280 Speaker 3: it everywhere. 288 00:14:55,360 --> 00:14:57,440 Speaker 2: They're not that different from one another. 289 00:14:57,440 --> 00:15:00,680 Speaker 3: And no one thinks like, oh, you know know, like 290 00:15:01,320 --> 00:15:04,320 Speaker 3: have you seen like Microsoft's latest spell check, Like just 291 00:15:04,360 --> 00:15:06,360 Speaker 3: no one cares because it's because all spell checks are 292 00:15:06,400 --> 00:15:08,480 Speaker 3: the same. And that would be the that would be 293 00:15:08,560 --> 00:15:12,560 Speaker 3: the kind of worst case scenario, although again not necessarily 294 00:15:12,560 --> 00:15:14,760 Speaker 3: a bad scenario for Nvidia because it's still gonna they're 295 00:15:14,760 --> 00:15:18,120 Speaker 3: all gonna be still running these these these models and. 296 00:15:18,080 --> 00:15:19,000 Speaker 2: Buying chips and so on. 297 00:15:19,040 --> 00:15:21,960 Speaker 3: It's just that maybe, as it turns out, you just 298 00:15:22,080 --> 00:15:26,160 Speaker 3: need to make a bunch of investments in AI to 299 00:15:26,240 --> 00:15:28,920 Speaker 3: continue selling the same software that you've been selling. 300 00:15:30,240 --> 00:15:36,000 Speaker 1: Well. Thank you so much, Max. Thanks Sarah, thanks for 301 00:15:36,040 --> 00:15:38,760 Speaker 1: listening to Big Take. I'm Sarah Holder. If you have 302 00:15:38,760 --> 00:15:41,440 Speaker 1: a moment to rate and review the show, we'd appreciate it. 303 00:15:41,440 --> 00:15:45,000 Speaker 1: It helps other listeners find us. This episode was produced 304 00:15:45,000 --> 00:15:48,320 Speaker 1: by Alex Suguira and Naomi Shaven. Our senior producers are 305 00:15:48,400 --> 00:15:51,640 Speaker 1: Naomi Shaven and Jill de Decari. It was edited by 306 00:15:51,760 --> 00:15:55,520 Speaker 1: Caitlin Kenny. It was fact checked by Adriana Tapia. It 307 00:15:55,600 --> 00:15:59,120 Speaker 1: was mixed by Alex Uguira. We get editorial direction from 308 00:15:59,160 --> 00:16:03,480 Speaker 1: Elizabeth Ponso. Nicole Beemster Bort is our executive producer. Saint 309 00:16:03,520 --> 00:16:07,200 Speaker 1: Bauman is Bloomberg's head of podcasts. We'll be back tomorrow.