WEBVTT - Bitcoin Hits Record High, Update on Huawei's Chip Breakthrough, Racial Bias in ChatGPT 0:00:02.600 --> 0:00:09.559 Bloomberg Audio Studios, podcasts, radio news. 0:00:08.520 --> 0:00:13.720 From Mahard were Innovation, Money and Power Collie in Silicon Valley, Nbon. 0:00:14.120 --> 0:00:18.120 This is Bloomberg Technology with Caroline Hyde and Ed Ludlow. 0:00:32.000 --> 0:00:34.280 I'm Caroline had a Bloomberg's World headquarters in New York. 0:00:34.800 --> 0:00:37.800 I'm Ed Dudlow in San Francisco. This is BlueBag Technology. 0:00:37.840 --> 0:00:38.279 Coming up. 0:00:38.440 --> 0:00:42.520 Full market coverage ahead. Stocks Crypto pushed a record highs. 0:00:42.680 --> 0:00:45.239 We'll discuss Bitcoin at seventy thousand, and the outlook for 0:00:45.560 --> 0:00:48.840 the AI rally as a video adds a trillion dollars 0:00:48.840 --> 0:00:50.560 to its market cap this year alone. 0:00:51.200 --> 0:00:54.800 Plus, we'll bring you exclusive reporting on Huawei's chip breakthrough, 0:00:54.880 --> 0:00:58.920 which used technology from two US suppliers. 0:00:58.960 --> 0:00:59.680 Details ahead. 0:01:00.000 --> 0:01:03.720 Microsoft says it's seeing signs hacker group Midnight Blizzard it's 0:01:03.760 --> 0:01:06.000 been attempting to gain access to its systems. 0:01:06.200 --> 0:01:07.319 Bloomberg Exclusive. 0:01:07.680 --> 0:01:11.920 Huawei and its partner Smick relied on US technology to 0:01:12.040 --> 0:01:16.280 produce an advanced chip in China last year, according to sources. 0:01:16.319 --> 0:01:20.880 The previously unreported information suggests that China still cannot rely 0:01:21.120 --> 0:01:25.040 entirely on its domestic supply chain and needs foreign components 0:01:25.080 --> 0:01:29.080 and equipment that's required for cutting edge products like semiconductor. 0:01:29.120 --> 0:01:32.640 The reporter that broke that news Bloomberg's Mackenzie Hawkins. We 0:01:32.760 --> 0:01:35.720 just showed the teardown video that Bloomberg did of the 0:01:35.800 --> 0:01:39.959 Huawei Mate sixty. It has this seven animeter processor from Smick. 0:01:40.360 --> 0:01:44.000 Smick used LAMB and applied Materials chip making equipment to 0:01:44.080 --> 0:01:44.440 do it. 0:01:45.000 --> 0:01:46.600 Tell me the rest. That's right. 0:01:46.959 --> 0:01:50.880 So this seventh animeter chip was lauded as a massive 0:01:50.960 --> 0:01:54.680 breakthrough in China. The US was trying to keep Bijing 0:01:54.800 --> 0:01:58.640 from getting seven animeter technology on the fear that it 0:01:58.680 --> 0:02:02.320 could give the country advanced AI capabilities that could lend 0:02:02.360 --> 0:02:04.720 it a military edge. And so we've seen the US 0:02:04.800 --> 0:02:07.640 implement this sweeping set of controls on the types of 0:02:07.720 --> 0:02:11.400 chips and semiconductor manufacturing equipment that can be shipped to 0:02:11.480 --> 0:02:15.639 China starting in October twenty twenty two, and the machines 0:02:15.680 --> 0:02:18.520 that were reporting were used to create this Huawei chip 0:02:18.800 --> 0:02:20.440 were shipped prior to that ban. 0:02:21.200 --> 0:02:22.880 But the through line on. 0:02:22.880 --> 0:02:26.200 This story is that despite Beijing's efforts to indigenize the 0:02:26.240 --> 0:02:30.600 full semiconductor supply chain and catch up to the United States, 0:02:30.720 --> 0:02:33.960 to the Netherlands, to Japan, these countries that are really 0:02:34.000 --> 0:02:36.919 dominant in the global chip industry to get to their 0:02:36.919 --> 0:02:42.120 most advanced ship, yet they still relied on foreign technology McKenzie. 0:02:42.400 --> 0:02:45.919 We know that the US is still applying pressure too well. 0:02:46.000 --> 0:02:49.800 Ultimately those other countries that they want to ensure that 0:02:49.840 --> 0:02:52.560 these sort of ways of ma navigating through the blocks 0:02:52.560 --> 0:02:54.799 don't still happen. They are turning to the Netherlands to 0:02:54.880 --> 0:02:58.600 Japan to squeeze even tighter. What then, of these indigenous 0:02:58.639 --> 0:03:01.840 make is over in China, won't of advanced microfabrication equipment 0:03:02.000 --> 0:03:04.040 then of Nara, Are they actually getting to a place 0:03:04.040 --> 0:03:05.080 where they could do it alone. 0:03:06.880 --> 0:03:08.680 It's an excellent question I had. 0:03:08.760 --> 0:03:12.799 Certainly the hope from the the Netherlands and Japan, which 0:03:12.800 --> 0:03:15.560 are part of this tripartite agreement to squeeze China on 0:03:15.600 --> 0:03:20.400 semicdector technology. They hope that the controls that they've already 0:03:20.440 --> 0:03:23.960 imposed will keep China from ever catching up. But the 0:03:24.040 --> 0:03:27.840 reality is there are significant differences between the controls imposed 0:03:27.840 --> 0:03:31.160 by those three countries. US firms are not allowed to 0:03:31.320 --> 0:03:34.440 send their employees to service equipment that's already in China, 0:03:34.520 --> 0:03:38.040 so these tools from LAM and Applied Materials, those employees 0:03:38.360 --> 0:03:42.800 can't go do repairs but their Dutch companies peers are 0:03:42.880 --> 0:03:46.000 able to in many cases, so there are still sort 0:03:46.040 --> 0:03:49.240 of big gaps that the US seeks to close between 0:03:49.520 --> 0:03:53.000 those three countries regimes. And then you have the US 0:03:53.080 --> 0:03:55.880 reaching out to South Korea and Germany, which are major 0:03:56.400 --> 0:03:58.960 or home to major producers of spare parts that go 0:03:59.000 --> 0:04:02.000 in chip making tools, to get them to join in 0:04:02.080 --> 0:04:04.440 on this accord and try to squeeze China further. 0:04:06.600 --> 0:04:07.440 So just really quick. 0:04:07.480 --> 0:04:10.880 At a point of clarification, that seven nanometers two generation 0:04:11.000 --> 0:04:13.640 removes from the cutting edge right three nanometer. But as 0:04:13.680 --> 0:04:16.839 you point out in your story, Mackenzie Smick obtained the 0:04:16.880 --> 0:04:20.719 lithography machines prior to that October twenty twenty two ban 0:04:21.080 --> 0:04:22.679 So I think there's more to go on this story, 0:04:22.680 --> 0:04:25.000 which I'm sure you'll look into, on how they get 0:04:25.000 --> 0:04:28.000 that domestic industry up to speed. 0:04:28.000 --> 0:04:31.560 Bluebogs. Mackenzie Hawkins terrific reporting. Thank you. Let's keep the 0:04:31.560 --> 0:04:32.400 conversation going. 0:04:32.560 --> 0:04:35.440 China is in the process of raising more than twenty 0:04:35.520 --> 0:04:38.559 seven billion dollars for its largest chip fund to date, 0:04:39.080 --> 0:04:42.840 accelerating the development of cutting edge technologies as it faces 0:04:42.920 --> 0:04:46.600 those restrictions on US tech known as the Big Fund 0:04:46.800 --> 0:04:50.279 the state back firm is expanding as the US prepares 0:04:50.320 --> 0:04:54.600 to sharply escalate technology curves designed to curtail Chinese chip 0:04:54.760 --> 0:04:59.040 and artificial intelligence progress. A tit for tat going on here. 0:04:58.920 --> 0:05:02.960 Caroline, and that investors have to navigate around. Are we 0:05:03.120 --> 0:05:05.680 still looking to put more money to work in companies 0:05:05.800 --> 0:05:09.159 that have significant exposure to China? Vida being one of them, 0:05:09.360 --> 0:05:11.360 has talked to someone at the cutting edge. Nacy Tanglin, 0:05:11.400 --> 0:05:14.279 piece of a joins our CEO CIO of Laffatanglo Investments 0:05:14.600 --> 0:05:17.200 for more. Also the author, of course, The Women's Guide 0:05:17.240 --> 0:05:20.600 to Successful Investing, You dive into where women must invest 0:05:20.680 --> 0:05:23.080 for their future. But most notably today you're joining from 0:05:23.080 --> 0:05:26.440 the CBOE a day after ringing the bell to celebrate 0:05:26.920 --> 0:05:28.840 not only International Women's Day but also the launch of 0:05:28.880 --> 0:05:32.000 your new ETF TGLR. We want to dig into the 0:05:32.040 --> 0:05:34.600 ETF in a moment, Nancy, but first your take on 0:05:34.720 --> 0:05:37.800 China US and how it factors into your investing thesis. 0:05:39.720 --> 0:05:43.400 Yeah, so it's a tricky one to navigate, clearly, Caroline. 0:05:43.800 --> 0:05:47.719 Our portfolios have about three to five percent exposure in 0:05:47.800 --> 0:05:51.000 terms of revenues to China, so we've been very careful 0:05:51.040 --> 0:05:53.279 about it, and of course the companies that have tended 0:05:53.279 --> 0:05:55.839 to underperform in the last year have been the companies 0:05:55.880 --> 0:05:58.760 that do actually have exposure to China, not just in 0:05:58.839 --> 0:06:03.080 technology but across the board. So I'm not as concerned 0:06:03.120 --> 0:06:06.600 about the twenty seven billion dollar chip fund that China's starting. 0:06:06.800 --> 0:06:09.239 I mean, I think that's pretty much what Intel spent 0:06:09.560 --> 0:06:11.640 in terms of R and D and cap X last 0:06:11.720 --> 0:06:15.400 year alone. So it's it's it's a drop in the bucket. 0:06:15.560 --> 0:06:17.760 But you do have to be aware and conscious of 0:06:17.800 --> 0:06:23.120 the fact that there are going to be continued hostilities, 0:06:23.160 --> 0:06:25.720 if you will, trade hostilities, which. 0:06:25.560 --> 0:06:27.720 Might factor into whether or not we pull back from 0:06:27.760 --> 0:06:30.120 these record highs. Nancy, we are talking to you in 0:06:30.160 --> 0:06:31.839 the nasdacs at are record high. The S and P 0:06:31.880 --> 0:06:34.720 five hundreds that a record high. Global World Index set 0:06:34.720 --> 0:06:36.960 a record high thanks to the likes of Nvidia. 0:06:37.040 --> 0:06:38.160 It's interesting that we have. 0:06:38.200 --> 0:06:40.200 Bruelcom for example. I know you've sort of called it 0:06:40.240 --> 0:06:43.320 the pullman's version of an Nvidia, but there is some 0:06:43.360 --> 0:06:45.520 weakness in that name. After its earnings, how are you 0:06:45.640 --> 0:06:48.760 ranking the AI Frenzy and the chip names to be owning? 0:06:50.640 --> 0:06:53.800 So if I think I've shared with your investing theme 0:06:53.920 --> 0:06:57.040 is old economy companies that are embracing AI and then 0:06:57.080 --> 0:07:00.200 the suppliers of the pix and shovels. So we have 0:07:00.279 --> 0:07:03.200 been expanding our exposure. We trimmed back on many of 0:07:03.240 --> 0:07:06.679 the names, including Broadcom and palowout to Networks a number 0:07:06.680 --> 0:07:10.600 of weeks ago and then again yesterday, and so I 0:07:10.640 --> 0:07:13.480 think you have to be disciplined in this environment. We 0:07:13.560 --> 0:07:17.280 still own plenty, but we've been adding to names like AXP. 0:07:17.520 --> 0:07:21.280 I have a triple A or a cube stock pick 0:07:21.360 --> 0:07:24.679 for you, and it's American Express, which is using AI 0:07:25.280 --> 0:07:29.880 to improve fraud detection, and then Amazon and Adobe the 0:07:29.960 --> 0:07:33.080 unsung hero we think of AI. So you have to 0:07:33.160 --> 0:07:37.320 pick and choose and navigate around and really focus on 0:07:37.400 --> 0:07:39.240 domestic revenues if you can. 0:07:40.480 --> 0:07:43.520 The thing about Broadcom is it's not a maker of 0:07:43.600 --> 0:07:46.680 AI accelerators, right Nancy. It kind of It has a 0:07:46.720 --> 0:07:50.440 custom silicon business, makes TPUs, but it's really in the 0:07:50.480 --> 0:07:54.000 serve a design build out. You know, it provides networking tech. 0:07:54.880 --> 0:07:58.360 But they had this specific forecast that AI will be 0:07:58.440 --> 0:08:02.600 thirty five percent contribution revenue previously from twenty five percent. 0:08:02.800 --> 0:08:05.880 Does that signal anything bigger to you? About how how 0:08:05.960 --> 0:08:09.840 continued or sustained the AI buildout is going to be. 0:08:12.200 --> 0:08:14.000 I think, I mean, you have to piece together from 0:08:14.000 --> 0:08:17.040 a lot of different companies. But I think the security 0:08:17.080 --> 0:08:19.800 I have in owning a name like Broadcom is that 0:08:19.920 --> 0:08:24.360 it is hoktan is consistently underestimated. We were buying the 0:08:24.400 --> 0:08:27.680 stock at two hundred dollars a share when he did 0:08:27.680 --> 0:08:30.440 the Computer Associates deal, and everyone hated it. It was 0:08:30.480 --> 0:08:33.400 a creative in a year. So there's more to Broadcom 0:08:33.400 --> 0:08:35.840 than just AI. But if you look at the use 0:08:35.880 --> 0:08:39.640 cases almost you know, we're sitting on earning season, everybody's 0:08:39.679 --> 0:08:44.480 talking about it across sectors, but they're really using AI. 0:08:44.679 --> 0:08:46.959 And I think that's why I think this market is 0:08:47.000 --> 0:08:49.480 analogous to the nineties, because we are going to continue 0:08:49.520 --> 0:08:53.000 to see productivity improvements and that's going to drive stock 0:08:53.040 --> 0:08:55.000 prices higher. I would just point out I didn't answer 0:08:55.080 --> 0:08:57.240 Caroline's question. I mean, if you look at the all 0:08:57.280 --> 0:09:00.720 times highs, yes, that's true, Caroline, But over the last 0:09:00.800 --> 0:09:02.880 two years, if you went twenty two to twenty three, 0:09:02.920 --> 0:09:05.840 the Nasdaq was actually down something like three percent, So 0:09:06.120 --> 0:09:08.880 we're not in nosebleed territory yet. 0:09:09.600 --> 0:09:13.440 Well, the difference is the contribution from Nvidia and Videos 0:09:13.480 --> 0:09:17.160 at nine hundred and twenty seven dollars a share stock split, 0:09:17.600 --> 0:09:21.640 stock split in video, please. 0:09:23.040 --> 0:09:25.000 So you would favor that, you think you'll go for it. 0:09:27.040 --> 0:09:30.199 I mean, they've done it before, so I think it's likely. 0:09:30.600 --> 0:09:32.760 I think it would be great if Broadcom split as well, 0:09:32.800 --> 0:09:35.200 and some of these other stocks that traded these, you know, 0:09:35.240 --> 0:09:38.959 really high levels. It makes it harder to manage portfolios 0:09:38.960 --> 0:09:41.520 with them in the portfolio, harder for the retail investor. 0:09:41.559 --> 0:09:43.520 But you know, we'll take what we get. We'll take 0:09:43.520 --> 0:09:46.640 dividend increases, we'll take share buybacks, whatever it takes to 0:09:46.720 --> 0:09:47.960 drive the stock price higher. 0:09:48.360 --> 0:09:51.000 And let's talk about the retail investor. You clearly care 0:09:51.040 --> 0:09:52.880 about them. You've been writing books for them, and you've 0:09:52.920 --> 0:09:55.520 also been looking at well, now we've got an actively 0:09:55.520 --> 0:09:59.040 managed ETF coming from you at the moment, Lafeateangla equity 0:09:59.080 --> 0:10:03.000 income ETF quality large cap stocks have strong earnings and 0:10:03.040 --> 0:10:06.800 dividend growth potential. What is your edge here, Nancy? Where 0:10:06.800 --> 0:10:10.280 do you want to be showing an offering to this clientele? 0:10:12.080 --> 0:10:14.480 Yeah, so this strategy is very different from what you 0:10:14.480 --> 0:10:17.560 would normally think of as an equity income portfolio. There's 0:10:17.559 --> 0:10:20.440 no electric utilities. We own one rate in the portfolio, 0:10:21.200 --> 0:10:24.400 but we buy what I call fallen angel growth stocks, 0:10:24.480 --> 0:10:26.800 and we do a lot of research because everybody knows 0:10:26.800 --> 0:10:29.200 the problems when the stocks are out of favor. So 0:10:29.240 --> 0:10:31.640 we have our own proprietary research model and our own 0:10:31.679 --> 0:10:35.280 proprietary valuation metrics, and that is the one we use 0:10:35.320 --> 0:10:38.480 in that portfolio. Is relative dividend yield. In these companies, 0:10:38.600 --> 0:10:41.079 management sets the dividend based on what they think long 0:10:41.160 --> 0:10:44.120 term sustainable earnings power is. So it's a total return 0:10:44.160 --> 0:10:47.520 strategy with dividend growth as the focus, and it's a 0:10:47.520 --> 0:10:50.160 great workhorse strategy. I have you the majority of my 0:10:50.200 --> 0:10:52.800 assets in this strategy, so you don't get to own 0:10:52.840 --> 0:10:55.640 some of the sexier names, but you do get really 0:10:55.679 --> 0:10:59.600 consistent risk adjusted returns that have been able to generate 0:10:59.640 --> 0:11:02.440 excess return above the benchmarks over time. 0:11:03.640 --> 0:11:07.120 Nancy Tegler, CEO CEO of Lapputengler Investments. Great to have 0:11:07.160 --> 0:11:09.400 you on the show on this Friday. In creats on 0:11:09.440 --> 0:11:12.680 the ETF All right, coming up, Microsoft says Russian hacking 0:11:12.679 --> 0:11:16.520 group Midnight Blizzard is making attempts to access its systems. 0:11:16.520 --> 0:11:18.640 We're gonna bring you all the updates on that story. Next, 0:11:18.800 --> 0:11:38.240 this is Bloomberg Technology. 0:11:35.360 --> 0:11:36.280 Time for talking tech. 0:11:36.320 --> 0:11:39.760 First up, Microsoft says there's evidence that information from an 0:11:39.760 --> 0:11:43.360 account compromised by a Russia linked hacker known as Midnight 0:11:43.400 --> 0:11:46.440 Blizzard was used in recent weeks to try and gain 0:11:46.480 --> 0:11:50.080 access to corporate data. In a blog post, Microsoft revealed 0:11:50.080 --> 0:11:52.400 that the hackers made attempts for some of the company's 0:11:52.400 --> 0:11:56.840 source code, repositories, and internal systems, but found no evidence 0:11:56.920 --> 0:12:01.599 that any customer facing systems were compromised. Plus semiconductor connectivity 0:12:01.600 --> 0:12:04.720 company Astera Labs is said to be planning to raise 0:12:04.880 --> 0:12:07.280 as much as five hundred and thirty four million dollars 0:12:07.480 --> 0:12:10.480 in an IPO. The Intel backed company hopes to tap 0:12:10.520 --> 0:12:14.280 into investor demand four artificial intelligence with a stock offering 0:12:14.360 --> 0:12:17.360 of twenty seven dollars to thirty dollars a share. That's 0:12:17.400 --> 0:12:19.679 according to a filing. The stock will trade on the 0:12:19.760 --> 0:12:24.400 Nasdaq and will debut under the ticker ALAB and Rivian 0:12:24.480 --> 0:12:27.120 hits the brakes, halting plans to build out a new 0:12:27.200 --> 0:12:29.440 multi billion dollar factory in Georgia. 0:12:29.520 --> 0:12:30.520 In a company. 0:12:30.160 --> 0:12:32.920 Filing, Rivian said the move will save the automaker as 0:12:32.960 --> 0:12:36.400 much as two point twenty five billion in capital expenditures 0:12:36.600 --> 0:12:39.960 this coming is Evmaker debuted its first mass market. 0:12:39.840 --> 0:12:41.719 R two model, which will now be. 0:12:41.720 --> 0:12:45.400 Made at first in its existing plant in Illinois. This 0:12:45.679 --> 0:12:49.920 was the Rivian CEO Carros sort of channeling his inner 0:12:49.960 --> 0:12:53.320 Steve Jobs. But the news was a change of plans. 0:12:54.080 --> 0:12:56.920 Who is charming the inner Steve Jobs with a surprise? 0:12:57.160 --> 0:12:58.120 Right the R two? 0:12:58.200 --> 0:13:00.439 Yes, we kind of had the leak on about the 0:13:00.480 --> 0:13:00.760 R three. 0:13:00.760 --> 0:13:01.560 What do you make of it? 0:13:02.120 --> 0:13:03.800 Yes, the R three is a crossover. 0:13:03.920 --> 0:13:05.040 They did keep it a secret. 0:13:05.080 --> 0:13:08.040 I'm so surprised it wasn't leaked frankly ahead of time. 0:13:08.120 --> 0:13:09.920 So it was kind of like the one more thing 0:13:10.120 --> 0:13:14.719 Steve Jobs moment. It's a prototype. It's a prototype right 0:13:14.760 --> 0:13:16.679 far in the future. But it did boost the stock. 0:13:17.120 --> 0:13:19.160 We'll see what the price point is. But what forty 0:13:19.160 --> 0:13:21.640 five thousand for an R two. We're starting to get 0:13:21.679 --> 0:13:23.600 a little bit more economical. 0:13:30.360 --> 0:13:31.240 So AI is. 0:13:31.160 --> 0:13:34.080 Helping drive stocks to new highs today, but there remains 0:13:34.120 --> 0:13:37.800 the ongoing concerns about its limitations, particularly when it comes 0:13:37.880 --> 0:13:40.720 to generator of AI and ethical standards safety. 0:13:41.080 --> 0:13:41.760 This is something I. 0:13:41.720 --> 0:13:45.240 Discussed with the CEO of Salesforce AI it's Clara Shee. 0:13:45.280 --> 0:13:45.800 Take a listen. 0:13:46.400 --> 0:13:49.320 There are so many ways of thinking about trust. It 0:13:49.400 --> 0:13:51.360 is something that I think about all the time. It's 0:13:51.800 --> 0:13:54.040 the number one company value at Salesforce. 0:13:54.200 --> 0:13:56.000 But when it comes to AI, I think we've all. 0:13:55.840 --> 0:13:59.959 Heard the stories of chat GPT and the various equivalent 0:14:00.160 --> 0:14:03.880 out there just making things up or spewing out toxic 0:14:03.960 --> 0:14:08.800 outputs or biased outputs, right because consumer chatbots are trained 0:14:08.920 --> 0:14:11.600 on the corpus of data on the Internet and so honestly, 0:14:11.679 --> 0:14:15.360 unfortunately in humanity, and there's a lot of incorrect, biased 0:14:15.360 --> 0:14:17.240 information out there, and so I think that's really the 0:14:17.280 --> 0:14:20.040 difference between consumer AI and business AI. 0:14:20.600 --> 0:14:21.360 Business AI is. 0:14:21.360 --> 0:14:25.320 Just unacceptable to make up an answer to what your 0:14:25.360 --> 0:14:28.720 sales forecast is or responding to a customer service request. 0:14:28.760 --> 0:14:30.160 And that's why our. 0:14:30.080 --> 0:14:32.680 Team, my team has spent so much time building out 0:14:32.720 --> 0:14:35.800 what we call our Einstein trust layer, So everything from 0:14:35.880 --> 0:14:40.320 data masking to citations, to zero retention prompt so that 0:14:40.480 --> 0:14:43.040 there's no data leakage, to keeping an audit trail, just 0:14:43.040 --> 0:14:45.000 to make it really trusted for companies. 0:14:46.120 --> 0:14:48.640 What about therefore, the developers you're then going out and 0:14:48.640 --> 0:14:52.760 talking to on a daylight today, why are they seeing 0:14:53.800 --> 0:14:56.640 worries about what they look like, the bias within what 0:14:56.720 --> 0:14:59.240 they're producing, or are you seeing a more diverse castic 0:14:59.280 --> 0:15:02.720 characters coming in building, particularly as you make AI easier 0:15:03.240 --> 0:15:05.680 with co pilots. The fact is that ultimately maybe we're 0:15:05.680 --> 0:15:07.960 not all going to have to be well code as 0:15:08.000 --> 0:15:09.080 let alone prompt engine is. 0:15:09.800 --> 0:15:12.440 I do think this is an amazing democratization moment for 0:15:12.480 --> 0:15:13.320 app developers. 0:15:13.400 --> 0:15:14.440 And I mean even. 0:15:14.240 --> 0:15:17.560 For years before AI Salesforce, one of the secret sources 0:15:17.560 --> 0:15:21.600 of Salesforce was it's no code and low code platform, 0:15:21.840 --> 0:15:25.840 which means that business analysts can build apps without knowing 0:15:25.880 --> 0:15:29.440 how to code, and even pro code developers who are 0:15:29.480 --> 0:15:32.080 really good at coding, they're able to use our tools 0:15:32.280 --> 0:15:36.000 to code faster. And AI is just an incredible accelerant 0:15:36.040 --> 0:15:38.480 and amplifier of both of these. And so today we 0:15:38.520 --> 0:15:41.840 had a series of really exciting announcements called our Einstein 0:15:41.880 --> 0:15:45.520 one studio that really makes it turnkey for any no 0:15:45.640 --> 0:15:49.000 code developer as well as pro code developer to use 0:15:49.040 --> 0:15:50.200 AI to build faster. 0:15:51.240 --> 0:15:54.240 Do you think that the AI systems that we're building 0:15:54.280 --> 0:15:56.720 right now are as diverse as the use cases that 0:15:56.720 --> 0:15:59.240 we're having and or are you worried about the bias 0:15:59.240 --> 0:16:02.480 that's in there, inevitably because of what it's ingesting. 0:16:03.120 --> 0:16:06.000 I think more broadly, we should all be very concerned 0:16:06.160 --> 0:16:10.120 about the AI model development and the regulation of AI 0:16:10.600 --> 0:16:11.560 not being diverse enough. 0:16:11.600 --> 0:16:13.080 And that's why I think it's so important for. 0:16:13.080 --> 0:16:16.440 Us to have rooms like this and dialogues like this 0:16:16.640 --> 0:16:19.640 to really ensure that a diverse set of perspectives are 0:16:19.680 --> 0:16:23.840 going into creating the AI itself, as well as creating 0:16:24.080 --> 0:16:28.240 the governance and the policies and the regulation around responsible AI. 0:16:28.800 --> 0:16:33.080 And then finally, AI is going to create tremendous wealth 0:16:33.160 --> 0:16:37.680 and value for shareholders, for stakeholders, for employees of these companies, 0:16:37.960 --> 0:16:41.360 for the people and the businesses that utilize AI. And 0:16:41.440 --> 0:16:44.640 we can't leave behind women in minorities in this value 0:16:44.640 --> 0:16:45.320 creation moment. 0:16:46.560 --> 0:16:48.880 My conversation a couple of days ago with the Salesforce 0:16:49.040 --> 0:16:51.680 AI CEO Claras, and we're going to dig in more. 0:16:52.360 --> 0:16:55.080 Yeah, let's stick with AI biased concerns and turn to 0:16:55.080 --> 0:16:58.840 today's Bloomberg Big Take. Our reporters have been working on 0:16:58.880 --> 0:17:03.359 an experiment to see if chat GPT has any discriminatory biases, 0:17:03.520 --> 0:17:08.400 finding in their conclusion that AI software can produce discriminatory 0:17:08.440 --> 0:17:13.440 results when prompted to rank resumes that are equally qualified. 0:17:13.480 --> 0:17:14.120 Let's bring in. 0:17:14.080 --> 0:17:17.680 One of those reporters, Bloomberg's Davy Alba. This is a 0:17:17.760 --> 0:17:21.200 must read, but it is complicated. Let's start with the methodology. 0:17:21.280 --> 0:17:23.560 First. Explain to the audience what you did. 0:17:24.080 --> 0:17:28.040 Yeah, thanks for having me. So we carried out this 0:17:28.200 --> 0:17:35.879 experiment where we used GPT's API, and so we prompted 0:17:36.040 --> 0:17:41.240 GPT three point five and GPT four with eight equally 0:17:41.359 --> 0:17:47.040 qualified resumes, keeping all of the qualifications of the resume's 0:17:47.400 --> 0:17:50.760 equal equivalent essentially, And the only thing that we changed 0:17:51.119 --> 0:17:53.159 was the name. 0:17:53.080 --> 0:17:55.120 That was on top of the resume. 0:17:55.760 --> 0:18:01.320 So that name we derived it using mutational methods and 0:18:02.080 --> 0:18:07.040 made it representative of particular demographic groups. So we had, 0:18:07.400 --> 0:18:12.600 you know, either male or female, white, Black, Hispanic or 0:18:12.640 --> 0:18:18.119 Asian demographically distinct names, and we said it loose, and 0:18:18.200 --> 0:18:24.960 we basically had GPT rank these resumes thousands of times 0:18:25.000 --> 0:18:28.240 for four different jobs. And what we found was actually 0:18:28.680 --> 0:18:34.720 some pretty stark evidence of racial bias and disparities depending 0:18:34.760 --> 0:18:39.120 on the job that we asked GPT to rank it for, 0:18:39.480 --> 0:18:41.280 and you. 0:18:41.240 --> 0:18:41.919 Know, sort of the. 0:18:43.480 --> 0:18:45.320 Candidates that it was evaluating. 0:18:45.600 --> 0:18:49.120 Okay, so you're looking for a financial analyst for example. 0:18:49.119 --> 0:18:52.840 As one of the examples, Yes, who ends up in 0:18:53.000 --> 0:18:57.280 probability coming out on top, who ends up coming off worst? 0:18:57.760 --> 0:19:01.240 Yeah, so that was one of our really good examples. 0:19:01.800 --> 0:19:06.000 You know, we found that black men and women were 0:19:06.240 --> 0:19:10.560 disproportionately ranked U sort of at the bottom. They were 0:19:10.640 --> 0:19:16.720 chosen as a top candidate the least frequently. And it 0:19:16.800 --> 0:19:20.920 was so stark that we actually measured adverse impact, which 0:19:20.960 --> 0:19:24.200 is sort of what the US federal agencies use as 0:19:24.359 --> 0:19:29.359 a benchmark to see whether a group is you know, 0:19:29.800 --> 0:19:34.960 badly impacted by a particular hiring process. And I wanted 0:19:34.960 --> 0:19:38.600 to add, this is not just theoretical that people are 0:19:38.720 --> 0:19:42.880 using GPT to use these to to use to sort 0:19:42.920 --> 0:19:47.800 of sort and rank resumes. We actually in our reporting 0:19:47.880 --> 0:19:51.119 found that a lot of businesses are indeed using it 0:19:51.160 --> 0:19:52.680 for this particular use case. 0:19:53.359 --> 0:19:55.560 And you know, it's it's. 0:19:55.440 --> 0:19:57.600 Important to consider these things. 0:19:57.880 --> 0:20:00.880 Open AI and response saying you're not meant to use 0:20:01.280 --> 0:20:04.600 chut GPT or GPT in our underlying technology, just straight 0:20:04.640 --> 0:20:07.680 out the box in this way, but still stark reporting 0:20:07.720 --> 0:20:09.719 from Davy Alba. We thank her and the team so 0:20:09.800 --> 0:20:11.840 much for it. 0:20:18.480 --> 0:20:20.720 Welcome back to Bloomberg Technology Ed love Loow here in 0:20:20.760 --> 0:20:21.440 San Francisco. 0:20:21.560 --> 0:20:22.639 I'm Caroline heard in New York. 0:20:22.960 --> 0:20:25.359 Let's keep today's jobs report in mind, but dive deeper 0:20:25.400 --> 0:20:28.320 into the tech sector angle to all of this and 0:20:28.359 --> 0:20:32.000 bring in Jacqueline Rice Nelson, CEO of the upskilling platform 0:20:32.240 --> 0:20:37.840 Tribe AI. Basically, unemployment two year high, but hiring continues. 0:20:37.960 --> 0:20:39.800 I'm going to try and make the jump to what's 0:20:39.840 --> 0:20:42.760 happening in AI in the context of that by saying 0:20:43.240 --> 0:20:47.679 there is a scarcity of roles or a scarcity of 0:20:47.720 --> 0:20:51.320 the skilled workers we need to meet that hiring demand 0:20:51.320 --> 0:20:53.240 in the AI context. Do you think that that's right? 0:20:54.200 --> 0:20:56.440 First of all, I'm thrilled to be here, and especially 0:20:56.560 --> 0:20:59.280 on a day that shows some of the potential of 0:20:59.560 --> 0:21:02.640 AI and value creation that I believe we will continue 0:21:02.680 --> 0:21:06.720 to see in the market. I think when you talk 0:21:06.760 --> 0:21:12.720 about AI and talent and opportunity, we can't really underscore 0:21:13.520 --> 0:21:16.359 that we're just at the beginning of this massive wave. 0:21:17.320 --> 0:21:25.680 And today we absolutely have steep specialists who have been living, breathing, 0:21:25.840 --> 0:21:31.440 sleeping AI for years. But what really matters right now 0:21:31.560 --> 0:21:34.439 is who has been doing on the ground generative AI 0:21:34.760 --> 0:21:38.480 applications built on top of these large language models. These 0:21:38.520 --> 0:21:42.200 models have only really been accessible to a broader population 0:21:42.720 --> 0:21:45.720 for a year, and so that population. 0:21:45.200 --> 0:21:46.359 Of people who really. 0:21:46.119 --> 0:21:49.719 Has the experience that companies actually need to do this 0:21:49.880 --> 0:21:52.320 well to do this right is limited. 0:21:53.359 --> 0:21:56.560 Okay, what is tribe seeing in the jobs market? We 0:21:56.840 --> 0:22:00.000 know jobs are being added added in this context or industry, 0:22:00.480 --> 0:22:01.479 but what are the roles? 0:22:01.560 --> 0:22:01.760 Right? 0:22:01.800 --> 0:22:05.160 All these companies are saying we're using AI, but who 0:22:05.160 --> 0:22:06.720 are they hiring as a result of that? 0:22:07.320 --> 0:22:11.160 Yeah, this is an incredibly interesting moment because I believe 0:22:11.200 --> 0:22:14.680 we are kind of at a real inflection point where 0:22:14.720 --> 0:22:20.600 the conversation around AI has been around efficiency, around cost cutting, 0:22:20.840 --> 0:22:25.720 and potentially the fears have been around job reductions. And 0:22:25.800 --> 0:22:29.600 I believe where we're headed is that there will absolutely 0:22:29.640 --> 0:22:33.280