WEBVTT - OpenAI Raises $110B from Amazon, Nvidia, Others 0:00:02.520 --> 0:00:13.480 Bloomberg Audio Studios, podcasts, radio news. Bloomberg Tech is alive 0:00:13.520 --> 0:00:17.319 from coast to coast with Caroline Hide in New York 0:00:17.600 --> 0:00:20.800 and v Lovelow in San Francisco. 0:00:22.280 --> 0:00:23.759 This is Bloomberg Tech coming up. 0:00:23.800 --> 0:00:26.799 Open Ai raises one hundred and ten billion dollars with 0:00:26.960 --> 0:00:29.760 key Amazon backing and a deal that values and startup 0:00:29.800 --> 0:00:34.440 at seven hundred and thirty billion plasanthropics dispute with Appendagon 0:00:34.720 --> 0:00:37.680 and intensifies over safeguards around the use of its AI 0:00:37.720 --> 0:00:41.720 technology and block plans to cut close to half of 0:00:41.800 --> 0:00:45.400 its entire workforce in a move the company describes as 0:00:45.400 --> 0:00:48.159 a bet on AI. But first we check in on 0:00:48.200 --> 0:00:50.639 these markets that have a lot more to digest than 0:00:50.760 --> 0:00:51.160 just AI. 0:00:51.320 --> 0:00:53.559 On the day, we're off by four tens percent, and. 0:00:53.560 --> 0:00:56.640 As that investors trying to navigate what has been a 0:00:56.680 --> 0:01:00.720 pretty brutal month. Yes there's anxiety over AI capex. Yes 0:01:00.760 --> 0:01:03.840 there's worried about where on inflatory pressures are going. The 0:01:03.880 --> 0:01:06.560 PPI data today showing that maybe the FED can't cut 0:01:06.560 --> 0:01:08.479 as much as the market had wanted. But there's also 0:01:08.560 --> 0:01:11.160 geopolitical risk. I ran us front and center. Let's look 0:01:11.160 --> 0:01:12.800 at what we're doing over the course of the month, 0:01:12.840 --> 0:01:15.600 because remember we are winding up February, and it means 0:01:15.600 --> 0:01:18.399 we've had the worst month for the NASDAQ one hundred 0:01:18.400 --> 0:01:21.360 and for the SMP five hundred since March of last year. 0:01:21.440 --> 0:01:23.360 We're off by more than three point eight percent. Move 0:01:23.360 --> 0:01:26.360 on and have individual companies in your line of sight today, 0:01:26.360 --> 0:01:28.759 because the key story is a private and a public 0:01:28.800 --> 0:01:31.280 market one. I'm looking at what happening to Amazon. We're 0:01:31.319 --> 0:01:33.400 up four tens a percent. It is going to be 0:01:33.440 --> 0:01:37.760 slump supplying fifty billion dollars to open ai in terms 0:01:37.800 --> 0:01:39.080 of its latest funding rounds. 0:01:39.319 --> 0:01:40.360 Does that come in compute? 0:01:40.400 --> 0:01:44.039 Does that come in terms of really focused access to 0:01:44.240 --> 0:01:46.200 open AI's own agents. 0:01:46.440 --> 0:01:48.280 We're looking at Video off by one point eight percent? 0:01:48.320 --> 0:01:49.400 What does that mean for their GPUs? 0:01:49.400 --> 0:01:51.600 But they're throwing in thirty billion dollars as well, So 0:01:51.680 --> 0:01:54.960 two is softbak. Let's get more details. Seth Biegerman is 0:01:54.960 --> 0:01:57.400 with us and piece to say. And what's so interesting 0:01:57.600 --> 0:02:01.080 is this deepening intertwined between open ai and Amazon. 0:02:01.520 --> 0:02:02.320 What do you make of it? 0:02:03.680 --> 0:02:05.720 Yeah, I mean it's really telling. I think it's another 0:02:05.760 --> 0:02:09.840 example for us that it takes all parties now to 0:02:09.880 --> 0:02:13.520 meet the needs of AI and cloud computing. Amazon obviously 0:02:13.520 --> 0:02:17.160 has been a long time supporter of OPENINGI rival Anthropic, 0:02:17.520 --> 0:02:21.880 but increasingly opening Ianthropic and Amazon have brokeer cloud computing 0:02:21.919 --> 0:02:23.720 deals as part of this form. In addition to the 0:02:23.760 --> 0:02:26.880 money that Amazon is committing, Opening I will be using 0:02:26.919 --> 0:02:29.760 Amazon's training and chips, which is certainly a vogue of confidence, 0:02:29.800 --> 0:02:32.639 and Amazon's technology, and they'll be working together on a 0:02:32.880 --> 0:02:38.640 models that might be useful for Amazon specifically. So again, Google, Amazon, Microsoft, 0:02:38.639 --> 0:02:41.239 the video, they're all kind of backing everyone right now 0:02:41.280 --> 0:02:43.240 to lift up the wider AI market. 0:02:43.720 --> 0:02:45.679 It's interesting therefore that maybe in video is down a 0:02:45.720 --> 0:02:48.079 little bit on the day throwing money into open ai, 0:02:48.240 --> 0:02:50.800 but maybe people worry if there's more dependence on Trainium 0:02:50.840 --> 0:02:54.160 and Amazon chips, maybe less dependence on GPUs. What's so 0:02:54.240 --> 0:02:57.400 interesting as well is open Ai says the funding round 0:02:57.400 --> 0:02:57.880 isn't over. 0:02:58.080 --> 0:03:00.400 They're also tapping venture capitalists coming forward. 0:03:01.000 --> 0:03:03.240 Yeah, it's our understanding that there's still billions more to 0:03:03.280 --> 0:03:04.840 go here, though I think the majority of it is 0:03:04.880 --> 0:03:08.040 now done and committed, and so that next trench of 0:03:08.080 --> 0:03:11.400 it might come from venture firms and from sovereign funds, 0:03:11.480 --> 0:03:13.720 with our expectation that this will all be wrapped up 0:03:13.840 --> 0:03:14.400 this quarter. 0:03:15.320 --> 0:03:18.880 What's been also notable is how maybe some of the 0:03:19.520 --> 0:03:22.640 pr moon music that have been so all consumed with 0:03:22.720 --> 0:03:25.640 open ai and its prowess had slightly shifted to the 0:03:25.680 --> 0:03:28.360 alternatives and propably been leading a lot of the charge 0:03:28.360 --> 0:03:29.720 in terms of unveils. 0:03:29.280 --> 0:03:32.359 To the enterprise. How much is open Ai now able. 0:03:32.120 --> 0:03:35.560 To refocus to talk up its coding tools, to be 0:03:35.560 --> 0:03:37.760 able to talk up its penetration into the enterprise. 0:03:38.400 --> 0:03:40.440 Well, certainly, first off, this is a vote of confidence 0:03:40.480 --> 0:03:42.320 in OPENINGI. I mean keep in mind that during the 0:03:42.320 --> 0:03:44.040 months they've been working on this deal, I think there's 0:03:44.040 --> 0:03:47.400 been heightened concerns on Wall Street about capex spending. With 0:03:47.440 --> 0:03:49.400 Opening I really at the center of a lot of 0:03:49.400 --> 0:03:53.160 that unprofitable startup committing to spend more than a trillion 0:03:53.160 --> 0:03:55.960 dollars on infrastructure, and yet they still are on the 0:03:56.040 --> 0:03:58.720 cusp of raising the largest funding round in history. Now, 0:03:58.800 --> 0:04:02.400 to your point, Hoping has, like Anthropic, been investing heavily 0:04:02.480 --> 0:04:05.880 in AI coding tools, AI agents and working to help 0:04:06.280 --> 0:04:08.640 enterprises figure out how to make the most of its 0:04:08.680 --> 0:04:11.240 existing AI technology. So I think we should just expect 0:04:11.240 --> 0:04:12.920 them to double down on that in the run up 0:04:12.960 --> 0:04:16.320 to what will likely be in IPO pretty. 0:04:16.080 --> 0:04:18.920 Much, Seth Giegman, we so appreciate the roundup. Thank you. 0:04:19.279 --> 0:04:22.120 Let's get the markets tape chaping Morgan Asset Management Global 0:04:22.160 --> 0:04:24.440 market strategy is Stephanie Adiagas here with us. 0:04:24.680 --> 0:04:25.160 Stephanie. 0:04:25.839 --> 0:04:28.880 This is once again shining a light on the extraordinary 0:04:28.880 --> 0:04:31.640 capital expenditure that is necessary for these frontier labs. 0:04:32.520 --> 0:04:35.080 It absolutely is, and I think it comes on the 0:04:35.120 --> 0:04:39.040 heels of this agentic boom that markets are really beginning 0:04:39.040 --> 0:04:41.640 to take issue with. And the reality is the reason 0:04:41.640 --> 0:04:45.800 why these agents are so phenomenal is because they think longer. 0:04:45.839 --> 0:04:49.200 The amount of inference required in these agents is magnitudes 0:04:49.320 --> 0:04:52.480 larger than the kind of single query just prompting an 0:04:52.560 --> 0:04:55.480 LLLM of the past. So, you know, the economics around 0:04:55.520 --> 0:04:58.040 AI continue to mount. We heard that from Nvidia this week. 0:04:58.240 --> 0:05:00.200 I think for the markets, I mean there's just this 0:05:00.360 --> 0:05:03.120 lingering anxiety of Okay, yes, we need a lot of infrastructure, 0:05:03.120 --> 0:05:04.720 but who's going to make money on all of this, 0:05:05.320 --> 0:05:08.039 and where are the competitive modes actually going to stick 0:05:08.080 --> 0:05:10.359 in this rollout? And then of course, what does this 0:05:10.440 --> 0:05:12.840 mean for some of the companies that may be privy to. 0:05:12.920 --> 0:05:15.200 Disruption well, before we get there. I want to go 0:05:15.240 --> 0:05:17.960 to Nvidia because it was sort of extraordinary. That's dound 0:05:18.040 --> 0:05:21.480 today and yesterday on the back of such strong fundamental 0:05:21.600 --> 0:05:23.960 numbers that it posted. How do you interpret that market 0:05:24.000 --> 0:05:24.680 shake off of it. 0:05:25.080 --> 0:05:26.839 I mean, part of it is we've all become, you know, 0:05:26.920 --> 0:05:31.000 Tiger parents, and we're used to exceptional results. It's not enough. 0:05:31.520 --> 0:05:34.560 We're now more concerned about, okay, moving forward. We got 0:05:34.560 --> 0:05:37.159 to see some kind of moderation in these growth rates, 0:05:37.600 --> 0:05:40.239 given just how remarkable that they've been for so long. 0:05:40.480 --> 0:05:41.960 I think markets are thinking about that. 0:05:42.040 --> 0:05:44.080 I think on the other side, there's also this desire 0:05:44.080 --> 0:05:48.239 to diversify perhaps exposure to some of these AI names, 0:05:48.279 --> 0:05:52.120 and that's really where we're seeing caution really emerging in markets. 0:05:52.560 --> 0:05:55.960 I have to say, I'll produce a made the Tiger Parents. 0:05:55.480 --> 0:05:58.000 An a changer yesterday. I'm having a love it. We 0:05:58.040 --> 0:05:59.960 all expect more and more and more. 0:06:00.120 --> 0:06:02.760 But what's interesting is whether or not we're starting to 0:06:02.800 --> 0:06:05.760 expect more of software now. I spoke to the Intuit 0:06:05.960 --> 0:06:10.040 CEO yesterday after their numbers came out, and Sasan has 0:06:10.080 --> 0:06:12.440 been talking about the disappointment he's had in the sell 0:06:12.480 --> 0:06:15.239 off of his stock, the worries that they've had about 0:06:15.440 --> 0:06:18.320 really their growth rates not supporting some of the decimation 0:06:18.440 --> 0:06:21.040 in the market capitalization of this business. And they feel 0:06:21.080 --> 0:06:24.359 that people are adopting into it, and they are striking 0:06:24.360 --> 0:06:26.520 these partnerships with Anthropic and Open Air, and they're like 0:06:26.720 --> 0:06:29.880 the software concerns, And there we have him saying, it's 0:06:29.960 --> 0:06:32.800 very hard to disprove a negative. How are you thinking 0:06:32.800 --> 0:06:35.880 about the market worrying about software and its ability to 0:06:36.120 --> 0:06:37.479 weather this AI disruption? 0:06:37.640 --> 0:06:40.960 Yeah, I think it's rational to price in a greater 0:06:41.040 --> 0:06:44.200 degree of uncertainty around the sustainability of earnings growth. But 0:06:44.279 --> 0:06:46.520 I think markets maybe getting a little ahead of themselves. 0:06:47.000 --> 0:06:50.640 And there is almost some misunderstanding I think around how 0:06:50.640 --> 0:06:52.560 transformed or how disruptive these tools are. 0:06:52.680 --> 0:06:54.320 Isn't the market meant to get ahead of itself? 0:06:54.320 --> 0:06:55.440 Though that's true. 0:06:55.480 --> 0:06:57.920 That's true, and it makes sense to self first and 0:06:57.960 --> 0:06:58.919 ask questions later. 0:06:59.160 --> 0:07:01.600 But now it's time for nuanced discussion. 0:07:01.480 --> 0:07:04.120 Around what these tools are going to do. Because these 0:07:04.160 --> 0:07:08.480 plugins themselves don't distruct businesses. It's about what you use 0:07:08.560 --> 0:07:12.000 those plugins to do. Who is leaning into AI to 0:07:12.080 --> 0:07:17.400 provide AI empowered services, to bolster competitive motes, to do more. 0:07:17.840 --> 0:07:20.640 And the other thing you know crossing headlines today is 0:07:20.680 --> 0:07:24.360 around labor disruption, and I'll say, you know, companies that 0:07:24.400 --> 0:07:27.880 are thinking about how they use AI to substitute their workers, 0:07:28.120 --> 0:07:31.200 they're missing the forest of the trees here. The real 0:07:31.320 --> 0:07:35.120 transformative impact from AI is what you can do to 0:07:35.200 --> 0:07:38.040 do more, to do faster, to build better, to do 0:07:38.120 --> 0:07:41.280 things that you wouldn't have done before, and ultimately outrun 0:07:41.320 --> 0:07:42.560 your competition in doing so. 0:07:42.640 --> 0:07:45.160 We are going to be discussing BLOCK at length in 0:07:45.200 --> 0:07:47.840 a moment, but I'm assuming that's exactly where you're pointing 0:07:47.880 --> 0:07:51.560 the fact that Jack Dorsey has admitted he over hired 0:07:51.920 --> 0:07:55.080 and in some ways has been rectifying and rebalancing BLOCK 0:07:55.120 --> 0:07:58.560 from a personnel perspective, But he is saying I am late, 0:07:59.000 --> 0:08:01.200 and we are going all in an AI and other 0:08:01.320 --> 0:08:03.440 companies are going to follow all that, other companies to 0:08:03.480 --> 0:08:06.280 follow who are bloated, who do need to RECTI five 0:08:06.320 --> 0:08:08.200 and are going to use AI as a good excuse. 0:08:08.720 --> 0:08:11.320 I think the reality is AI is going to create 0:08:11.440 --> 0:08:15.920 a lot of transformation in work, and that transformation could 0:08:16.000 --> 0:08:19.720 at times look like layoffs in some areas. But I 0:08:19.760 --> 0:08:22.360 think over the medium term, we're going to see all 0:08:22.400 --> 0:08:23.800 sorts of job creation too. 0:08:24.000 --> 0:08:27.320 Maybe they want because we'll keep talking about promot engineering. 0:08:27.360 --> 0:08:28.080 Well that's sort of thing. 0:08:28.680 --> 0:08:29.640 It is a thing of the past. 0:08:29.920 --> 0:08:31.760 Well, one of the things is, look, these agents can 0:08:31.840 --> 0:08:34.800 create a whole bunch of work. They can automate processes, 0:08:35.120 --> 0:08:38.400 but the amount of work that they can reliably automate 0:08:38.480 --> 0:08:40.800 relative to what a human worker does in a services 0:08:40.840 --> 0:08:44.520 oriented economy is still a fraction I should add. And 0:08:44.600 --> 0:08:47.920 all of these you, our listeners should know. Hallucination rates 0:08:47.960 --> 0:08:51.239 for the current frontier models still range between twenty. 0:08:51.000 --> 0:08:52.760 Six percent and eighty percent. 0:08:53.120 --> 0:08:57.280 Whatever frontier model you're using, Claude Gemini, chatchbt. They're still 0:08:57.280 --> 0:08:59.760 getting things wrong. There is still a real layer here 0:09:00.080 --> 0:09:02.240 and oversight. And the thing is, the need for human 0:09:02.280 --> 0:09:05.680 oversight only compounds when now we're talking about a lot 0:09:05.720 --> 0:09:08.280 more product, a lot more output because AI is helping 0:09:08.360 --> 0:09:08.920 us achieve that. 0:09:09.000 --> 0:09:11.320 It's interesting because of late, I think there's been pushed 0:09:11.320 --> 0:09:13.880 back on the hallucinations that are happening at the very 0:09:13.920 --> 0:09:15.000 cutting edge models. 0:09:15.040 --> 0:09:16.400 And yes, that might be a thing of the past. 0:09:16.440 --> 0:09:18.319 So I'd love to dig into that deity that you're 0:09:18.360 --> 0:09:21.560 still seeing maybe up to eighty percent. More broadly, how 0:09:21.600 --> 0:09:23.840 you thinking about the macro impact and whether or not 0:09:23.880 --> 0:09:25.960 we will see a FED that's able to tackle this 0:09:26.040 --> 0:09:26.760 in any way or not. 0:09:26.840 --> 0:09:30.080 Yeah, it's a challenge, and that's an artificial analysis for 0:09:30.120 --> 0:09:32.720 the FED. It's a challenge right now. You see there's 0:09:32.720 --> 0:09:34.920 a little bit warmer inflation. Right now, we kind of 0:09:34.960 --> 0:09:36.320 know what's contributing to that. 0:09:36.360 --> 0:09:38.000 It's tariffs. Maybe we can see through that. 0:09:38.600 --> 0:09:41.120 And I was just in DC earlier this week hearing 0:09:41.160 --> 0:09:43.240 from a bunch of these FED governors, and. 0:09:43.600 --> 0:09:45.120 There is a range of opinions here. 0:09:45.160 --> 0:09:47.000 But I think the reality is the FED is very 0:09:47.080 --> 0:09:50.560 data dependent, and right now, in the hard data, you 0:09:50.679 --> 0:09:55.320 still see very very early impacts of AI labor disruption 0:09:55.960 --> 0:09:57.840 over the medium to long term. There are I think 0:09:57.880 --> 0:09:59.920 some big questions I think we have to rely on 0:10:00.120 --> 0:10:02.240 I guess what his history has told us, and then 0:10:02.320 --> 0:10:05.360 ultimately the fact that yes, we're all amazed at the 0:10:05.400 --> 0:10:08.400 exponential growth of the technology of the frontier, but the 0:10:08.440 --> 0:10:10.679 diffusion the real economy isn't exponential. 0:10:10.960 --> 0:10:11.480 It's linear. 0:10:11.480 --> 0:10:13.480 It's going to be jagged. That's going to give us time. 0:10:13.480 --> 0:10:16.520 It's going to give policy makers time to adjust. Stephanie, 0:10:16.840 --> 0:10:19.880 thank you for helping push us forward. Sevani Aniaga of 0:10:20.120 --> 0:10:24.240 JP Morgan Acid Management. Now coming up, Anthropic rejects the 0:10:24.320 --> 0:10:27.480 pentigons the latest offer in a dispute over AI safeguards. 0:10:27.520 --> 0:10:30.320 All the details coming next is a bretting back tech. 0:10:38.640 --> 0:10:43.000 I offered more talks if so long as they're in 0:10:43.040 --> 0:10:47.320 good faith. We're always open to talks. And we set 0:10:47.480 --> 0:10:50.880 a deadline, and we meant the deadline, and up until 0:10:50.880 --> 0:10:53.000 that deadline, I'm open to more talks. And I told 0:10:53.040 --> 0:10:53.320 them so. 0:10:54.160 --> 0:10:57.520 Emil Michael nan, under Secretary of Defense for Research and Engineering. 0:10:57.520 --> 0:11:01.319 After Anthropic rejected the Pentagon's latest off in a dispute 0:11:01.360 --> 0:11:03.880 over safeguards around the use of its AI technology by 0:11:03.920 --> 0:11:07.560 the US military. Now and Thropic CEO said, quote, we 0:11:07.720 --> 0:11:11.960 cannot in good conscience accede to these their request. And 0:11:12.040 --> 0:11:15.360 now well, the deadline for any further talks is fast approaching. 0:11:15.480 --> 0:11:18.240 Lumos Tech in Industrial Policy reporter Maggie Island joins us 0:11:18.240 --> 0:11:18.679 for more. 0:11:19.000 --> 0:11:19.920 It's fast approaching. 0:11:20.080 --> 0:11:23.800 It's five pm New York time, Washington time. 0:11:23.960 --> 0:11:28.680 Maggie. What can be done until then? Exactly? 0:11:28.760 --> 0:11:31.880 So? Look, the rhetoric here has been rather intense. 0:11:32.040 --> 0:11:32.480 You see. 0:11:32.520 --> 0:11:36.960 Emil Michael the Pentagon under secretary accusing Dario. 0:11:36.679 --> 0:11:40.520 Amidae of having a God complex. Last night, you see. 0:11:40.400 --> 0:11:44.920 Dario continuing to say that we need these two safety guardrails, 0:11:44.920 --> 0:11:48.560 and those guardrails are no autonomous strikes without a human 0:11:48.600 --> 0:11:52.040 in the loop and no surveillance of US citizens, and 0:11:52.120 --> 0:11:54.640 without that he doesn't seem willing to budge. So definitely, 0:11:54.679 --> 0:11:57.600 these two sides are staring each other down, and we're 0:11:57.600 --> 0:11:59.400 going to find out at five pm. 0:11:59.480 --> 0:12:00.520 Who's will to blink. 0:12:01.200 --> 0:12:05.000 We're also seeing coalitions of workers who are employees at Amazon, 0:12:05.040 --> 0:12:07.679 at Google and Microsoft and open Ai also asking their 0:12:07.679 --> 0:12:11.120 companies to join Anthropic in refusing to comply with the 0:12:11.200 --> 0:12:15.200 Defense Department demands. Maggie, what is the argument coming not 0:12:15.240 --> 0:12:17.559 only from amor Michael saying he's got a gold complex 0:12:17.559 --> 0:12:19.880 and actually calling him a liar, but what are the 0:12:20.000 --> 0:12:22.080 legal protections already in place? 0:12:24.280 --> 0:12:24.480 Right? 0:12:24.640 --> 0:12:27.560 So the Pentagon has said, look, according to the Constitution, 0:12:27.559 --> 0:12:30.840 according to US law, the US is not allowed to 0:12:31.160 --> 0:12:34.320 have domestic surveillance over its own citizens now. And Thropic's 0:12:34.360 --> 0:12:37.240 take on that is actually that AI might propose some 0:12:37.400 --> 0:12:40.600 new legal ground where there are situations where just due 0:12:40.640 --> 0:12:44.960 to the sheer power of artificial intelligence to aggregate public data. 0:12:45.080 --> 0:12:48.240 It could lead to new surveillance use cases under which 0:12:48.320 --> 0:12:52.040 the law isn't clear. So there's definitely some semantics going 0:12:52.080 --> 0:12:54.839 on here. At the same time, these are real sort 0:12:54.880 --> 0:12:58.280 of life and death questions about how the US wants 0:12:58.280 --> 0:13:01.160 the AI tools to be used for warfare. 0:13:02.040 --> 0:13:05.240 Mag Eastland, You're reporting throughout has been stella. Thank you 0:13:05.320 --> 0:13:08.000 very much for joining us on it. Let's continue the conversation. 0:13:08.080 --> 0:13:10.200 Sarah creps is with US, director of the Tech Policy 0:13:10.240 --> 0:13:13.720 Institute at rot Cornell University, and you're at the intersection 0:13:13.760 --> 0:13:18.880 of how national security, geopolitics, and tech policy intertwine. This 0:13:19.000 --> 0:13:21.440 is anthropic having to lead the charge here in many 0:13:21.440 --> 0:13:23.080 ways and the only one have got this sort of 0:13:23.080 --> 0:13:26.520 relationship with the Pentagon thus far. Would you make of 0:13:26.640 --> 0:13:27.560 Dario's pushback? 0:13:29.200 --> 0:13:31.120 Yeah, and I think we should take a step back 0:13:31.200 --> 0:13:34.040 and acknowledge why they're at the tip the no pun 0:13:34.120 --> 0:13:36.160 intended the tip of the spear here with the Pentagon, 0:13:36.600 --> 0:13:39.680 which is that a year or so ago, you know, 0:13:39.720 --> 0:13:41.880 they seem to go in the direction of focusing on 0:13:42.000 --> 0:13:44.680 enterprise and what that more than a year ago because 0:13:44.720 --> 0:13:47.240 their first contract with pell Andeer was in twenty twenty four, 0:13:47.880 --> 0:13:50.240 and so that's the market they're going for, and so 0:13:50.280 --> 0:13:53.280 they engaged. They were part of, you know, one two 0:13:53.400 --> 0:13:59.080 hundred million dollar contracts with Pellenteer and the Pentagon in 0:13:59.240 --> 0:14:02.760 part of that enterprise kind of business that they're focusing on, 0:14:03.160 --> 0:14:05.720 and so as part of that, that enters them into 0:14:07.120 --> 0:14:10.560 a you know, this arrangement that I think has then 0:14:10.760 --> 0:14:14.240 led to where we are now, which is this standoff 0:14:14.600 --> 0:14:17.760 in terms of how they balance their kind of ethical 0:14:17.800 --> 0:14:22.800 constitution with the need to be doing enterprise work that 0:14:22.920 --> 0:14:24.480 will bring in the revenue. 0:14:26.000 --> 0:14:29.600 There is a lot at stake because what is It's 0:14:29.640 --> 0:14:32.080 not only just a two hundred million dollar contract of 0:14:32.120 --> 0:14:35.440 work that Anthropic has, but there's almost a threat coming 0:14:35.520 --> 0:14:39.760 from the Pentagon that if you don't abide by our rules, 0:14:40.040 --> 0:14:42.280 we're going to say that you're a supply chain issue 0:14:42.360 --> 0:14:45.720 and lots of other military related companies are not going 0:14:45.760 --> 0:14:47.600 to be able to use your models in the future. 0:14:48.120 --> 0:14:50.000 How much of a problem would that be for Anthropic? 0:14:50.040 --> 0:14:54.080 How much you surprise by the Pentagon's focus there? 0:14:55.080 --> 0:14:58.640 Well, I think they are definitely two sides of this story. 0:14:58.800 --> 0:15:03.080 So we're is what I think makes this generative AI 0:15:03.320 --> 0:15:06.920 so different from bombs and bullets and nuclear weapons is 0:15:06.960 --> 0:15:10.200 that nuclear weapons and missile silo only has one purpose 0:15:10.360 --> 0:15:14.520 and it was built by defense contractors. The cutting edge 0:15:14.640 --> 0:15:18.240 AI is coming out of the civilian world, and it's 0:15:18.280 --> 0:15:21.360 a classic dual use technology problem, which is it's starting 0:15:21.400 --> 0:15:23.760 in a civilian space and now it's getting appropriate and 0:15:23.840 --> 0:15:28.680 used by the Pentagon, and so it really it's a 0:15:28.720 --> 0:15:34.200 civilian technology that now has this critical national security, you know, value. 0:15:34.400 --> 0:15:37.360 And so that's where this tension is. That both sides 0:15:37.360 --> 0:15:40.960 are correct, but the Pentagon does have a lot of 0:15:41.040 --> 0:15:43.280 leverage because it's it's the federal government. 0:15:44.080 --> 0:15:47.160 It's the federal government, and we've been hearing from the 0:15:47.160 --> 0:15:49.640 federal government. I just want to hear a little bit 0:15:49.680 --> 0:15:52.520 more from under Sexuary Michael. That's what he told Blomberg 0:15:52.520 --> 0:15:53.680 earlier today. Just take a listen. 0:15:54.520 --> 0:15:57.800 We've been negotiating in good faith on the Department of 0:15:57.880 --> 0:16:02.280 Warside for about three months and we're working pretty diligently, 0:16:02.600 --> 0:16:05.960 and we sent over a proposal that we thought made 0:16:05.960 --> 0:16:10.600 a lot of concessions to the language that Anthropic wanted, 0:16:10.680 --> 0:16:15.280 and then without any notice, they published an article where 0:16:15.280 --> 0:16:18.000 we thought we were getting close, saying that they were 0:16:18.040 --> 0:16:21.240 breaking off talks well before the deadline, which is generally 0:16:21.280 --> 0:16:24.440 not good partner oriented practice, if you. 0:16:24.440 --> 0:16:27.880 Will and look Panthropic in return, has said that while 0:16:27.880 --> 0:16:31.440 the Pentagonservator's proposal fell short, the company continues to negotiate 0:16:32.040 --> 0:16:35.480 with defense officials and remains committed to working with the military. 0:16:35.560 --> 0:16:38.240 Sarah, So, when you think more broadly. 0:16:37.960 --> 0:16:40.320 And this is an echo of what happened with Google 0:16:40.560 --> 0:16:44.560 years ago, how do you think tech policy can be 0:16:44.640 --> 0:16:49.359 written by the Pentagon, by the government to fit current purposes. 0:16:50.400 --> 0:16:54.400 It is surprising that we're seeing this repeat of twenty eighteen, 0:16:54.440 --> 0:16:58.600 where Google and the CEO seem to have been blindsided 0:16:58.680 --> 0:17:01.960 by the employees reluctance to work with a Pentagon, And 0:17:01.960 --> 0:17:04.560 here we are eight years later, and it seems like 0:17:04.600 --> 0:17:08.159 a repeat of this that could have been avoided. But 0:17:08.240 --> 0:17:11.200 I think if we put ourselves in the position of 0:17:11.240 --> 0:17:13.880 how quickly AI has been moving in the last couple 0:17:13.960 --> 0:17:18.040 of years, you can see how this just becomes kind 0:17:18.040 --> 0:17:20.919 of a somehow new territory, even though it feels like 0:17:20.920 --> 0:17:23.280 we've been here before. And so I think, what was 0:17:23.280 --> 0:17:26.200 happening last you know, the last year or two, AI 0:17:26.320 --> 0:17:29.200 was moving so quickly. The Pentagon is moving quickly, trying 0:17:29.200 --> 0:17:32.159 to do things differently, and you can see then why 0:17:32.760 --> 0:17:34.600 you know? And I worked in the acquisition business in 0:17:34.640 --> 0:17:36.640 the Air Force, and there was always this question, why 0:17:36.680 --> 0:17:38.720 can't we move faster? This is why you can't move 0:17:38.760 --> 0:17:42.719 faster is because things with the federal government and national 0:17:42.720 --> 0:17:47.320 security and classified work and removing leaders from Venezuela are 0:17:47.520 --> 0:17:50.320 just not the same as coming up with a grocery 0:17:50.359 --> 0:17:53.560 list for yourself at home with chatch Ebt or in 0:17:53.600 --> 0:17:56.320 this case, Claude Daria. 0:17:56.400 --> 0:17:59.120 Armide has been very clear that he thinks more about 0:17:59.160 --> 0:18:02.040 the implications of grocery list, and he's written about the 0:18:02.440 --> 0:18:05.440 adolescence of technology, written large at the beginning of this year, 0:18:05.520 --> 0:18:09.200 thinking about the geopolitical implications, the implications for ethics for 0:18:09.720 --> 0:18:13.239 our world of work going forward. Sarah, just tell us 0:18:13.240 --> 0:18:16.120 a little bit about where you think there might be 0:18:17.119 --> 0:18:18.400 any room for agreement. 0:18:18.880 --> 0:18:20.600 Can the Pentagon go as. 0:18:20.480 --> 0:18:24.359 Far as to agree, as with Dariamide's desire to not 0:18:24.640 --> 0:18:27.960 have surveillance of US citizens or not use models for 0:18:28.000 --> 0:18:30.639 autonomous lethal strikes without human in the loop. Is that 0:18:30.760 --> 0:18:33.639 ever something that they could be specific enough about. 0:18:34.920 --> 0:18:36.760 I think it's a great question, and I think that's 0:18:36.760 --> 0:18:40.320 the reluctance of anthropic which is that the US has 0:18:40.320 --> 0:18:43.680 been saying, we are we one is to use AI, 0:18:43.840 --> 0:18:47.280 use cloud for any lawful use, and it's not lawful 0:18:47.560 --> 0:18:50.399 to use autonomous weapons, and it's not lawful to do 0:18:50.520 --> 0:18:53.919 mass surveillance. So what's the problem here. I think the worry, 0:18:54.600 --> 0:18:58.080 probably from the perspective of Anthropic is that slippery slope. 0:18:58.119 --> 0:19:00.280 What does it mean to be fully autonomous? What does 0:19:00.320 --> 0:19:02.400 it mean to do mass surveillance? And so I think 0:19:02.440 --> 0:19:05.399 that's that great. What they think is gray area that 0:19:05.440 --> 0:19:07.640 they want to be very careful about. And I think 0:19:07.680 --> 0:19:11.240 the view has been that they would rather be trying 0:19:11.280 --> 0:19:15.040 to influence safe use of AI from the inside rather 0:19:15.240 --> 0:19:19.560 than take a sanctimonious perspective and be on the outside 0:19:19.600 --> 0:19:23.560 and not be able to influence the implementation and deployment 0:19:23.600 --> 0:19:26.080 of AI. And so I think that's what they're trying 0:19:26.119 --> 0:19:29.040 to do. And so for the Anthropics leverage, I think 0:19:29.160 --> 0:19:31.159 is that their model is really good and that the 0:19:31.200 --> 0:19:34.840 Pentagon wants to use it. So my expectation would be 0:19:34.880 --> 0:19:40.600 that they probably will find some middle ground here. Before 0:19:40.680 --> 0:19:41.400 five pm. 0:19:41.200 --> 0:19:45.920 Today, Sarah Craps of the Tech Policy Institute at Cornell University, 0:19:46.280 --> 0:19:48.879 we thank you very much. Indeed, now we just want 0:19:48.960 --> 0:19:52.359 to bring you some breaking news from a Trump post 0:19:52.920 --> 0:19:56.480 on truth social So President Trump posting about an FTC 0:19:57.000 --> 0:20:00.879 letter to Apple. Now Trump is posting FTC letter to 0:20:00.920 --> 0:20:04.840 Apple on Apple News's political actions. The FTC wrote to 0:20:04.840 --> 0:20:08.200 Apple on Februy the twelfth, regarding Apple News, we're off 0:20:08.200 --> 0:20:10.119 by one point six percent. Will bring you any further 0:20:10.800 --> 0:20:14.679 details regarding that post, It is just simply a post 0:20:14.760 --> 0:20:18.320