WEBVTT - Bloomberg Wall Street Week - June 2nd, 2023 0:00:00.560 --> 0:00:02.880 This is Bloomberg Wall Street Week. 0:00:02.920 --> 0:00:05.600 We turn our attention to the markets this week. USCPI 0:00:05.720 --> 0:00:09.760 en neembors reinforcing concerns about inflation. The financial stories that 0:00:09.880 --> 0:00:12.600 shape are worth a really different reaction to market. Some 0:00:12.600 --> 0:00:16.120 more indications of just how hot the US economy really. 0:00:15.920 --> 0:00:18.440 Is rude the eyes of the most influential voices. 0:00:18.720 --> 0:00:21.639 Katherine Keating, CEO of B and Y Moan, Ryan Winnahan, 0:00:21.760 --> 0:00:24.560 a Bank of America, Sam zel Sharmon and founder of 0:00:24.560 --> 0:00:25.960 Equat Group Investments. 0:00:26.000 --> 0:00:28.840 Bloomberg Wall Street Week with David Weston. 0:00:28.600 --> 0:00:29.920 From Bloomberg Radio. 0:00:30.120 --> 0:00:34.160 Drones hit Moscow, lower numbers hit China, and Washington works 0:00:34.200 --> 0:00:37.280 hard not to hit the debt ceiling. This is Bloomberg 0:00:37.280 --> 0:00:40.879 Wall Street Week. I'm David Weston. This week's secial CONTRIBA 0:00:40.960 --> 0:00:44.280 Larry Summers of Harvard on the stubbornly strong labor market 0:00:44.400 --> 0:00:45.760 and what it means for inflation. 0:00:46.360 --> 0:00:50.520 There's something about the underlying strength of the economy that 0:00:50.800 --> 0:00:51.640 they're missing. 0:00:52.120 --> 0:00:55.960 And former IBM CEO Sam Palmersano on just how big 0:00:56.120 --> 0:00:59.120 generative AI could get and what could stop it. 0:00:59.560 --> 0:01:03.680 You need some kind of a cognition, a capability which it. 0:01:03.640 --> 0:01:04.319 Does not have. 0:01:17.319 --> 0:01:21.360 Global Wall Street started the week watching drone hits in Moscow, 0:01:21.800 --> 0:01:25.400 perhaps as a precursor to a counter offensive in Ukraine. 0:01:25.560 --> 0:01:28.560 We're able to reach you right in the heart of Russia, 0:01:28.600 --> 0:01:31.080 and these attacks are really about sort of like sending 0:01:31.400 --> 0:01:31.960 that message. 0:01:31.959 --> 0:01:34.240 So it's quite embarrassing for Russia that this is happening. 0:01:34.560 --> 0:01:37.120 We spent another week on the debt ceiliing in Washington, 0:01:37.240 --> 0:01:39.720 but this time it wasn't about getting a deal, but 0:01:39.760 --> 0:01:42.600 about getting it all behind us, at least for a 0:01:42.600 --> 0:01:45.400 couple of years. We're going to deal with the debt ceiling. 0:01:45.600 --> 0:01:47.680 I think things are going as planned. 0:01:47.720 --> 0:01:51.080 Tonight we all made history because this is the biggest 0:01:51.240 --> 0:01:55.320 cut and savings this Congress has ever voted. 0:01:54.960 --> 0:01:58.400 For, and as expected, not everyone was happy. 0:01:59.080 --> 0:02:01.440 Now one would again should vote for this deal. 0:02:02.120 --> 0:02:05.960 Not once China got some bad economic news, raising more 0:02:06.040 --> 0:02:08.360 questions about the speed of its recovery. 0:02:08.840 --> 0:02:12.720 China's manufacturing activity contracted for a second straight month in May. 0:02:12.960 --> 0:02:15.760 It's more evidence that the post COVID recovery has slowed 0:02:15.919 --> 0:02:18.080 in the world's second largest economy. 0:02:20.720 --> 0:02:23.040 But here in the United States, the jobs, well, it 0:02:23.240 --> 0:02:25.760 just kept on going, adding three hundred and thirty nine 0:02:25.760 --> 0:02:28.200 thousand new jobs in the month of May that was 0:02:28.280 --> 0:02:30.239 way more than the one hundred and ninety five thousand 0:02:30.240 --> 0:02:34.000 that was estimated, while unemployment actually rose up from three 0:02:34.040 --> 0:02:36.960 point four percent to three point seven percent, which set 0:02:37.000 --> 0:02:39.400 the markets off to the races, with the S and 0:02:39.400 --> 0:02:41.639 P five hundred up one point eight three percent for 0:02:41.720 --> 0:02:45.120 the week, almost all of it on Friday alone. The 0:02:45.200 --> 0:02:47.520 Nasdaq was up a bit more, adding two point oh 0:02:47.639 --> 0:02:50.960 four percent for the week, while the bond markets almost 0:02:50.960 --> 0:02:53.560 did a round trip, starting the week just under three 0:02:53.600 --> 0:02:56.000 point eight percent on the ten year yield, dipping well 0:02:56.000 --> 0:02:58.960 below three point six percent at one point, and then 0:02:58.960 --> 0:03:01.680 coming back to just under three point seven percent on 0:03:01.760 --> 0:03:04.840 Friday as people return to thinking about possible rate hikes 0:03:05.160 --> 0:03:06.680 to take us through the week in the market, is 0:03:06.680 --> 0:03:10.359 welcome now, Barbara reinhardtch she's Voye Capital, head of asset Allocation, 0:03:10.600 --> 0:03:14.760 and Aaron Brown PIMCO Multi Assets Strategy portfolio Manager, So 0:03:14.840 --> 0:03:17.040 welcome back to both of you. Great to have you here. Barbara. 0:03:17.080 --> 0:03:18.160 Let me start with you, but what we're seeing in 0:03:18.160 --> 0:03:19.880 the markets. We saw some numbers on Friday and then 0:03:19.880 --> 0:03:22.480 you get the commer roll looks pretty strong. Ata when 0:03:22.520 --> 0:03:24.960 it comes to jobs. How are the markets reacting to 0:03:24.960 --> 0:03:25.760 all this information. 0:03:25.919 --> 0:03:28.920 It was a strange employment report, right, so it didn't 0:03:28.960 --> 0:03:31.480 coincide with necessarily some of the other data that we're 0:03:31.520 --> 0:03:33.880 looking at, looking at such as and if I be 0:03:34.240 --> 0:03:37.200 hiring intentions the home based data, they would have told 0:03:37.200 --> 0:03:38.760 you that the labor market would have been a lot 0:03:38.800 --> 0:03:41.400 weaker than the number that we had printed today. But 0:03:41.520 --> 0:03:43.880 the good news is and that what the equity market 0:03:44.000 --> 0:03:48.280 held onto is the gentle decline in average hourly earnings. 0:03:48.560 --> 0:03:51.240 And that is good news on the inflation side, because 0:03:51.240 --> 0:03:54.280 you've had very sticky inflation. It's been coming down, but 0:03:54.360 --> 0:03:57.440 coming down probably slower than what people expected. But to 0:03:57.480 --> 0:03:59.760 see the easing in the wage data is a very 0:03:59.760 --> 0:04:00.960 good positive development. 0:04:01.440 --> 0:04:03.320 Barbara, we've heard an awful lot about this lag and 0:04:03.360 --> 0:04:06.040 the effects from monetary policy. How long a lag should 0:04:06.040 --> 0:04:08.080 we expect, because boy, it doesn't see you're showing up 0:04:08.200 --> 0:04:09.360 very profoundly so far. 0:04:09.640 --> 0:04:11.960 So it tends to be about you know, when the 0:04:11.960 --> 0:04:14.920 Federal Reserve started hiking industrates, it was about twenty about 0:04:15.080 --> 0:04:17.960 fourteen months ago. There tends to be a lag of 0:04:18.000 --> 0:04:22.080 about twenty four months between when the Fed starts hiking 0:04:22.160 --> 0:04:24.480 and when you start to see a dramatic slowdown in inflation. 0:04:24.960 --> 0:04:27.120 That means we should have at least the rest of 0:04:27.120 --> 0:04:29.600 this year into the first quarter of twenty four before 0:04:29.640 --> 0:04:31.799 you start to see a significant slow down in inflation. 0:04:32.320 --> 0:04:34.520 Look when you look at things like the sticky price 0:04:34.600 --> 0:04:38.719 CPI or the underlying inflation gauge, they were at, you know, 0:04:38.760 --> 0:04:42.400 between six and seven percent just six months ago. They're 0:04:42.520 --> 0:04:44.560 down now to about three and a half to four 0:04:44.560 --> 0:04:47.359 and a half percent, well above the Fed's target, but 0:04:47.400 --> 0:04:50.000 they're going in the right direction. In order to get 0:04:50.040 --> 0:04:52.560 down to the two percent on inflation, which is the 0:04:52.600 --> 0:04:55.560 fed's target, you likely need to see a little bit 0:04:55.560 --> 0:04:56.440 more time go by. 0:04:56.640 --> 0:04:57.599 Think about this, David. 0:04:57.960 --> 0:05:00.480 The real issue is that the FED hikes the seventy 0:05:00.480 --> 0:05:02.640 five basis points that they did over the course of 0:05:02.880 --> 0:05:05.680 three or four meetings last year. They happened in the 0:05:05.680 --> 0:05:08.520 third and fourth quarters of last year. You need at 0:05:08.600 --> 0:05:10.920 least twelve months for that to start to be felt 0:05:10.920 --> 0:05:14.000 by the economy and start to see the slowdown you're 0:05:14.040 --> 0:05:16.280 seeing in some of the data change indexes. You're starting 0:05:16.320 --> 0:05:19.599 to see some slowing. Certainly not indicative of the labor 0:05:19.640 --> 0:05:22.599 market this morning on that report, but I would say 0:05:23.480 --> 0:05:27.000 that data can change and can change relatively quickly. 0:05:27.760 --> 0:05:31.080 It's not just US investors who drive US market. We 0:05:31.120 --> 0:05:34.479 also pay attention to foreign investors in US securities. And 0:05:34.520 --> 0:05:37.000 Torsten Slock of Apollo is here with one of his 0:05:37.200 --> 0:05:42.560 charts to show where foreign investors are putting their money. 0:05:42.640 --> 0:05:46.080 So before the pandemic, that was appetite among foreigners for 0:05:46.200 --> 0:05:51.360 buying in particular US government bonds and also US equities. 0:05:51.839 --> 0:05:54.760 Foreigners have also historically bought a lot of US credit, 0:05:55.080 --> 0:05:59.360 but broadly speaking, the trends were that foreigners had appetide 0:05:59.560 --> 0:06:02.520 for buying both US fixed income and buying US equities. 0:06:02.720 --> 0:06:05.720 So over the months ahead, we still have that the 0:06:05.720 --> 0:06:09.200 federal serve is doing quantitive tightening to the tune of 0:06:09.279 --> 0:06:11.920 roughly sixty billion dollars every single month. 0:06:12.200 --> 0:06:13.080 At the same time, we. 0:06:13.080 --> 0:06:16.160 Also have a budget deficit which is roughly around five 0:06:16.160 --> 0:06:18.920 percent of GDP, and that amounts to roughly in round numbers, 0:06:18.960 --> 0:06:22.440 about one hundred billion dollars in issuings of treasuries every 0:06:22.440 --> 0:06:27.359 single month. Combining both QT and the budget deficit. With 0:06:27.520 --> 0:06:30.760 this new issue, and on the other side of this 0:06:30.920 --> 0:06:34.360 that Celing deal now potentially being quite substantial, it does 0:06:34.680 --> 0:06:38.360 raise questions about who is it that will be buying treasuries, 0:06:38.400 --> 0:06:41.200 in particular in an environment where the hedging costs for 0:06:41.200 --> 0:06:43.560 foreigners continue to be quite elevated. 0:06:44.279 --> 0:06:46.800 Aaron Brown up PIMCO and Barbara Ryan Hart a boy 0:06:46.880 --> 0:06:48.599 are still with us. Er, let me start with you here. 0:06:48.760 --> 0:06:50.720 We've been following this after the death Sente got put 0:06:50.760 --> 0:06:53.360 behind this unfortunately did this way. There's the question we're 0:06:53.400 --> 0:06:56.240 replenishing the larder essentially in the US treasure to get 0:06:56.720 --> 0:06:58.960 their stock back up. It's going to be a lot 0:06:58.960 --> 0:07:01.120 of tea bills. Do we have the buyers for those 0:07:01.160 --> 0:07:01.680 T bills? 0:07:02.640 --> 0:07:05.160 Yeah, it's Torsten mentioned. I mean the challenge is from 0:07:05.160 --> 0:07:08.240 a foreigner's perspective. If you're a Japanese investor and you're 0:07:08.279 --> 0:07:11.960 looking to buy ten year government bonds, after the FX 0:07:12.000 --> 0:07:16.000 hedging cost measured by this sort of three month you know, 0:07:16.200 --> 0:07:21.160 FX forward hedge, it will cost you negative two percent. 0:07:21.800 --> 0:07:23.800 Your yield that you're getting is about negative two point 0:07:23.840 --> 0:07:25.880 two percent on a ten year bond. You can buy 0:07:25.960 --> 0:07:30.160 jgbs for about forty basis points. So from a foreigner's perspective, 0:07:30.240 --> 0:07:32.960 it's better to buy the local bonds than to take 0:07:33.000 --> 0:07:35.440 into account the FX hedging costs and buy a US 0:07:35.480 --> 0:07:39.000 treasury bond, even though US treasuries at least from a 0:07:39.720 --> 0:07:42.520 yield perspective in the US is yielding. More so, it's 0:07:42.600 --> 0:07:46.240 really unattractive right now just because of FX hedging costs 0:07:46.280 --> 0:07:50.240 to buy US government bonds from a foreigner's perspective. You know, 0:07:50.280 --> 0:07:53.800 there's another prop a wrinkle that I'm going to throw into, 0:07:54.320 --> 0:07:57.320 you know, the challenge that we have with perspective supply. 0:07:57.600 --> 0:08:01.200 Not only do we need to refill the cofer of 0:08:00.440 --> 0:08:04.600 the t GA and we have to fund the physical deficit, 0:08:04.720 --> 0:08:08.040 but also keep in mind from an international perspective, you 0:08:08.080 --> 0:08:11.440 have you're going to start to have the TLG or 0:08:11.560 --> 0:08:16.360 ros also that are going to you know, start adding 0:08:16.400 --> 0:08:20.120 to additional supply uh in in Europe as well. So 0:08:20.200 --> 0:08:23.360 you have a lot of government supply that's going to 0:08:23.440 --> 0:08:25.520 be you know, hitting the market in the second half 0:08:25.520 --> 0:08:28.000 of this year, which is going to you know, create 0:08:28.040 --> 0:08:31.840 a challenge certainly, you know, in terms of or raise 0:08:31.880 --> 0:08:34.000 the question in terms of who's going to be the 0:08:34.040 --> 0:08:37.560 incremental guy or for that the problem with the debt 0:08:37.600 --> 0:08:40.599 ceiling being passed, and I mean it's good from a 0:08:40.640 --> 0:08:43.920 political perspective and good from a stability perspective, but it 0:08:44.120 --> 0:08:49.520 does mean that you now have have to refill those coffers. 0:08:49.559 --> 0:08:52.360 They're gonna be a significant amount of issuance over the 0:08:52.440 --> 0:08:54.800 next couple of months, and the Treasury is saying that 0:08:54.840 --> 0:08:58.120 they want to rebuild to about six hundred billion dollars 0:08:58.480 --> 0:09:01.640 that TGA, which is means a significant amount of issuance 0:09:01.679 --> 0:09:03.400 over the next three to four months. 0:09:03.440 --> 0:09:06.000 So Barbro, i'mos gonna. I hadn't thought about the Teltro issue. 0:09:06.040 --> 0:09:07.439 But whatever it is, what's you going to do to 0:09:07.480 --> 0:09:09.360 liquid in the marketplace? And what could that mean as 0:09:09.360 --> 0:09:11.160 a fract many that money's got to come from some place. 0:09:11.360 --> 0:09:13.319 It does, but a lot of times you're talking about 0:09:13.360 --> 0:09:16.280 cash buyers moving over to a different cash instrument. In 0:09:16.320 --> 0:09:18.480 the past twenty years, there have been four times that 0:09:18.559 --> 0:09:21.640 the Treasury General Account has had to raise significant amount 0:09:21.679 --> 0:09:24.080 of assets like it is this time. Each time it 0:09:24.160 --> 0:09:26.280 was able to do it without an issue. We don't 0:09:26.320 --> 0:09:29.120 rule out that something could potentially be and miss this 0:09:29.200 --> 0:09:32.360 time and cause some indigestion, But the FED is very 0:09:32.400 --> 0:09:35.240 well versed in the plumbing of the treasury market, and 0:09:35.559 --> 0:09:37.520 we don't think that it's necessarily going to be a 0:09:37.640 --> 0:09:40.920 very big issue for the front end market. Plus also 0:09:40.960 --> 0:09:44.000 think about this, You're now getting interest rates that are 0:09:44.520 --> 0:09:49.000 almost in excess of headline inflation to be parked in 0:09:49.400 --> 0:09:52.760 tea bills. So it's just a very attractive investment for 0:09:52.880 --> 0:09:56.040 many investors. And many of the bank depositors are looking 0:09:56.080 --> 0:09:58.200 at the table market thinking I can go over to 0:09:58.200 --> 0:10:01.439 money markets and get much better of interest. So I 0:10:01.480 --> 0:10:03.120 think there's going to be a lot of natural buyers. 0:10:03.160 --> 0:10:05.760 Thank you so very much to Barbara Reinhardt, Avoye Capital 0:10:05.800 --> 0:10:08.440 and Aaron Brown of Pimpco. Now, this week we marked 0:10:08.480 --> 0:10:11.440 the one hundredth birthday of Henry Kissinger, the man who 0:10:11.520 --> 0:10:13.959 laid the foundation for the most important economic relationship in 0:10:14.040 --> 0:10:16.000 the world. That was, of course, the one between the 0:10:16.040 --> 0:10:18.360 United States and China. So we went back and found 0:10:18.360 --> 0:10:20.880 an interview Doctor Kissinger did with Lewis Rockaiser back in 0:10:20.920 --> 0:10:24.200 nineteen ninety when Chinese GDP was a mere three hundred 0:10:24.240 --> 0:10:27.920 and sixty billion dollars compared to the eighteen trillion dollars 0:10:28.040 --> 0:10:31.520 that it is today. 0:10:31.960 --> 0:10:35.440 Do you think we still have important economic opportunities in China? 0:10:35.600 --> 0:10:39.400 I think we do, and I'm in favor of maintaining 0:10:39.520 --> 0:10:43.440 a working relationship with China. It is too important the 0:10:43.520 --> 0:10:49.920 country it had made good procreates in economic refem and 0:10:49.960 --> 0:10:53.360 I think we should not likely sacrifice the relationship with 0:10:53.679 --> 0:10:57.880 the country with the longest continuing history in Asia, with 0:10:58.080 --> 0:10:59.840 the cultural influence that it has. 0:11:01.840 --> 0:11:03.959 This is Wall Street Week on Bloomberg. 0:11:05.679 --> 0:11:09.880 This is Bloomberg Wall Street Week with David Weston from 0:11:10.000 --> 0:11:10.960 Bloomberg Radio. 0:11:18.240 --> 0:11:21.960 Artificial intelligence it's been with us for longer than we think. 0:11:22.200 --> 0:11:25.160 Bank of America's Brian moynihan says they've been using it 0:11:25.200 --> 0:11:26.840 for at least five years. 0:11:27.160 --> 0:11:29.160 We have this thing called Erica, which is a virtual 0:11:29.640 --> 0:11:33.880 artificial intelligence, virtual assistant, natural language processing predictive technology that 0:11:33.880 --> 0:11:35.280 we've been running for four or five. 0:11:35.160 --> 0:11:38.720 Years, and people like JP Morgan's Jamie Diamond insist that 0:11:38.840 --> 0:11:43.080 it is transformational. AI is real. This is not no, 0:11:44.120 --> 0:11:48.199 that's not crypto ed. This is a technology which is staggering, 0:11:48.440 --> 0:11:52.240 with AI pioneer Samuel Altman telling Congress we think it 0:11:52.280 --> 0:11:54.400 can be a printing press moment. We have to work 0:11:54.400 --> 0:11:56.840 together to make it so. But now there are those 0:11:56.880 --> 0:11:59.880 who are warning about some of the downsides with tech. 0:12:00.040 --> 0:12:04.160 Veterans like Carly Fiorina, agreeing with Open AI. Samuel Altman 0:12:04.200 --> 0:12:06.000 that there needs to be regulation. 0:12:06.559 --> 0:12:10.559 Big tech has to say we alone are not in charge, 0:12:10.840 --> 0:12:12.439 and we alone cannot control. 0:12:13.080 --> 0:12:15.960 And a group of AI industry leaders this week wrote 0:12:16.000 --> 0:12:19.040 an open letter warning that what they are building could 0:12:19.160 --> 0:12:24.600 lead to a risk of extinction and to take us 0:12:24.600 --> 0:12:27.400 through some of the possible impediments to general of AI, 0:12:27.480 --> 0:12:30.280 as well as the risks involved. Welcome to Sam Palmasano. 0:12:30.360 --> 0:12:33.920 He is now the chairman of the Center for Global Enterprise, 0:12:33.960 --> 0:12:36.040 but of course for many years he was the chair 0:12:36.080 --> 0:12:39.120 and CEO of IBM. So Sam, welcome back to Wall Street. 0:12:39.120 --> 0:12:41.920 We've really great to have you here. Let's start, first 0:12:41.920 --> 0:12:45.800 of all with exactly how big this generative AI could be. 0:12:45.920 --> 0:12:50.360 We've seen Nvidia this week really explode in its market capitalisation. 0:12:50.920 --> 0:12:52.560 Is it over hyped? Do you think? Is it as 0:12:52.559 --> 0:12:53.400 big as people say? 0:12:54.160 --> 0:12:54.480 David? 0:12:54.480 --> 0:12:56.079 But first of all, thank you for having me. Now, 0:12:56.080 --> 0:12:57.040 I think it's that big. 0:12:57.080 --> 0:13:01.160 This is transformational technology, but it's always but I mean 0:13:01.200 --> 0:13:04.200 I always draw the analogy to the Internet because if 0:13:04.240 --> 0:13:08.040 you think back where it began exchanging technical documents around 0:13:08.080 --> 0:13:10.200 the world in the seventies to what has become today 0:13:10.240 --> 0:13:13.720 fifty years later, a lot happened in innovation and development 0:13:13.800 --> 0:13:17.520 and regulation and guidance, etc. All that has to happen 0:13:17.559 --> 0:13:19.920 with generative AI or large language models. 0:13:20.120 --> 0:13:23.080 But it is that transformation transformational because. 0:13:23.080 --> 0:13:25.400 Like the browser for the Internet made it kind of 0:13:25.480 --> 0:13:28.000 user friendly for people to actually get in there and 0:13:28.080 --> 0:13:30.960 look up information or search, this will do the same 0:13:31.080 --> 0:13:33.599 thing as far as AI is concerned, for people to 0:13:33.679 --> 0:13:37.040 use natural language to interface to it and get responses 0:13:37.080 --> 0:13:40.280 that are very much like human being responses. 0:13:40.960 --> 0:13:44.240 So if it is transformational, as you say, what are 0:13:44.240 --> 0:13:46.440 we going to need to have it be that transformational? 0:13:46.480 --> 0:13:49.040 What are some of the key inputs. Let's start, for example, 0:13:49.080 --> 0:13:52.480 with microchips. That's what's driving in Vidiot or large degree 0:13:52.520 --> 0:13:54.680 and maybe some other chip manufacturers as well. What is 0:13:54.720 --> 0:13:55.400 the need there? 0:13:56.760 --> 0:14:00.320 Well, basically this is going to be it requires massive infrastruction. Sure, 0:14:00.440 --> 0:14:02.960 and today you think about as hyperscalers as the term, 0:14:03.000 --> 0:14:07.720 or cloud computing service providers like Amazon, Google, Microsoft, etc. 0:14:08.800 --> 0:14:13.480 Having said that, this is huge demands on capacity and 0:14:13.600 --> 0:14:17.520 performance of that capacity. Therefore, the chips and the advantage 0:14:17.520 --> 0:14:20.480 that Nvidia has is these things graphical processing units that 0:14:20.520 --> 0:14:26.000 are called GPUs. Those technologies themselves are much better performers 0:14:26.000 --> 0:14:29.000 than the classic Intel chip, which referred to as X 0:14:29.120 --> 0:14:32.720 eighty six technology. So fundamentally you need that level of 0:14:32.760 --> 0:14:37.000 performance to run this like these natural language models, because 0:14:37.000 --> 0:14:39.240 they are so large, the data setsor is so large 0:14:39.280 --> 0:14:41.760 and requires so many interconnections that you. 0:14:41.760 --> 0:14:44.200 Need the performance to be acceptable to the end user. 0:14:44.960 --> 0:14:47.000 One of the things we hear about is energy. Also, 0:14:47.280 --> 0:14:49.440 we didn't seem that long ago we're talking about crypto 0:14:49.520 --> 0:14:51.760 mining and how what a big drain that was for 0:14:51.880 --> 0:14:54.480 energy back China really cut back on some of the 0:14:54.560 --> 0:14:57.800 mining there in China. What is the demand potentially in 0:14:57.840 --> 0:15:01.800 this transformational gener of AI LM, what is the demand 0:15:01.800 --> 0:15:02.360 for energy? 0:15:02.600 --> 0:15:04.800 Oh, it's the same as all these large cloud service 0:15:04.880 --> 0:15:05.480 data centers. 0:15:05.520 --> 0:15:08.080 I mean it's massive. I mean you could model those 0:15:08.120 --> 0:15:11.080 as if they're nuclear reactors. That so much energy they consume. 0:15:11.240 --> 0:15:13.320 Now, there's a lot that can be done, I mean, 0:15:13.400 --> 0:15:17.000 quite honestly, around the utilization of the infrastructure to cut 0:15:17.000 --> 0:15:20.400 the energy costs. There's ways you can use green energy 0:15:20.640 --> 0:15:23.920 to supplement the classical forms of electricity in those sorts 0:15:23.960 --> 0:15:26.640 of things. But there are lots of techniques associated with 0:15:26.640 --> 0:15:27.400 how to reduce that. 0:15:27.520 --> 0:15:28.640 But energy is going. 0:15:28.600 --> 0:15:31.760 To be significant there's no doubt about it already is 0:15:31.920 --> 0:15:34.120 I mean, it has been for years in these large 0:15:34.160 --> 0:15:37.000 whether they were supercomputers in the old days or now 0:15:37.040 --> 0:15:40.040 the cloud infrastructure providers, it's the same issue. 0:15:40.200 --> 0:15:42.320 They require significant amounts of energy. 0:15:42.680 --> 0:15:44.720 So, Sam, we've heard an awful lot about the wonderful 0:15:44.720 --> 0:15:48.160 things that generative AI could bring us, not least of 0:15:48.200 --> 0:15:51.840 which is increased productivity. They're remarkable of projections here of 0:15:51.880 --> 0:15:55.560 trillions of dollars of increased production because of a general AI. 0:15:55.760 --> 0:15:58.000 Then we had a letter that came out this week 0:15:58.320 --> 0:16:01.040 from a group of AIX first people who know about 0:16:01.040 --> 0:16:03.440 this warning about the risks and even saying there's a 0:16:03.520 --> 0:16:06.880 risk of extinction if we're not careful. Is that overblown? 0:16:07.200 --> 0:16:09.400 Well, okay, let me tell you what the risk are. 0:16:09.560 --> 0:16:12.040 I don't know if it's extinction, but the risk are. 0:16:12.120 --> 0:16:15.960 Quite Honestly, this technology makes you feel like you're actually 0:16:16.040 --> 0:16:19.040 getting something that's correct, you know, right, But it's not. 0:16:19.760 --> 0:16:20.760 It's not valid. 0:16:21.000 --> 0:16:24.400 And what it's doing basically at the engineering design is 0:16:24.400 --> 0:16:27.920 connecting large scale words together, but there's no checking to 0:16:28.000 --> 0:16:30.360 the source of the words. There's no fact checking as 0:16:30.400 --> 0:16:33.600 the conclusions these words reach. In some cases it's a 0:16:33.720 --> 0:16:36.440 terrific essay that it's prepared. Maybe it's a great book 0:16:36.440 --> 0:16:39.200 that it's written, or a wonderful poem. But at the 0:16:39.320 --> 0:16:41.800 end of the day, if you need accuracy, and you 0:16:41.920 --> 0:16:44.400 have to have accuracy, like in healthcare, if you were 0:16:44.400 --> 0:16:48.080 depending upon your diagnostics, you would not use chat GBT. 0:16:48.400 --> 0:16:50.520 I mean you want to have a doctor or some 0:16:50.800 --> 0:16:53.240 human of some kind that looked at the conclusion that 0:16:53.320 --> 0:16:55.840 it reached, even though it probably read all the medical 0:16:55.880 --> 0:16:58.800 journals like we did in Watson Health. You need some 0:16:58.920 --> 0:16:59.800 kind of a. 0:16:59.800 --> 0:17:02.760 Cognition capability, which it does not have. 0:17:03.520 --> 0:17:06.040 I think people get confused by what it does have 0:17:06.119 --> 0:17:08.240 because it's so easy to use. 0:17:08.600 --> 0:17:11.040 I mean, it's very, very attractive, there's no doubt about it. 0:17:11.440 --> 0:17:13.720 But it needs to be enhanced with some of these 0:17:13.760 --> 0:17:17.360 capabilities that will address the concerns that are in that article. 0:17:17.080 --> 0:17:20.000 Today about worrying about it. It could be like nuclear warfare 0:17:20.000 --> 0:17:21.720 in the future. I mean, if you think. 0:17:21.600 --> 0:17:25.600 About applications that could have a short term impact that 0:17:26.040 --> 0:17:28.320 you would like to have facts but don't have any facts. 0:17:28.320 --> 0:17:32.760 Take political propaganda as an example of that, our political campaigns. 0:17:32.840 --> 0:17:35.119 I mean, you'd like to have some facts and all 0:17:35.160 --> 0:17:38.199 that information but we all know they don't exist, and 0:17:38.280 --> 0:17:40.920 I'm not sure that proponents of all that information really 0:17:40.960 --> 0:17:42.720 want the facts out there anyway. 0:17:43.240 --> 0:17:47.040 Sam, is that risk? Is that weakness about actually truth checking? 0:17:47.440 --> 0:17:50.680 Is that inherent in large language models? Is it inherent 0:17:50.720 --> 0:17:53.880 in general AI or is it something with new iterations? 0:17:53.960 --> 0:17:56.040 Actually it could take over yourself so it could actually 0:17:56.119 --> 0:17:56.639 check itself. 0:18:00.000 --> 0:18:00.720 It needs to be done. 0:18:00.760 --> 0:18:03.040 Is the self checking, like you said, that does not 0:18:03.200 --> 0:18:07.199 exist today. I make the argument that the inventors of 0:18:07.240 --> 0:18:10.960 this technology that are raising all these alarms knew about 0:18:11.000 --> 0:18:13.680 these alarms, and they could have written some of these checkers, 0:18:13.680 --> 0:18:17.000 these tracing of the information or transparency of the data 0:18:17.240 --> 0:18:21.080 into the actual technology itself. Now, however, you know, I 0:18:21.160 --> 0:18:23.760 understand that they raised the concerns. They are very very 0:18:23.840 --> 0:18:28.000 valid concerns. However, if you are the inventor or the 0:18:28.000 --> 0:18:30.760 guys coming up with this technology, you knew this and 0:18:30.800 --> 0:18:32.040 you could have prepared for this. 0:18:32.480 --> 0:18:33.600 But now it has to be done. 0:18:33.840 --> 0:18:37.080 I mean, it's great technology, it's transformational, but now it 0:18:37.160 --> 0:18:38.720 has to be done, and it will be done, and 0:18:38.720 --> 0:18:41.199 there's going to be lots of small innovative companies that 0:18:41.240 --> 0:18:42.720 are going to take advantage of that gap. 0:18:42.720 --> 0:18:43.960 And add that capability. 0:18:44.440 --> 0:18:45.560 What's the government to do? 0:18:47.880 --> 0:18:50.960 It depends, you know, right, this is a great dilemma. 0:18:51.240 --> 0:18:53.040 And I take you back to the early days of 0:18:53.080 --> 0:18:55.679 the Internet. I mean, in the early days of the Internet, 0:18:55.720 --> 0:18:58.440 the government decided to take a light hand of guidance 0:18:58.440 --> 0:19:01.200 to the Internet, and the reason being that they would 0:19:01.240 --> 0:19:05.359 like to see the innovation could grow and accelerate before 0:19:05.400 --> 0:19:06.320 they actually put the. 0:19:06.880 --> 0:19:08.720 Say, the regulations in place. 0:19:08.720 --> 0:19:10.600 And they're still trying to do that today as you 0:19:10.600 --> 0:19:14.080 look at data privacy and those kinds of issues, which are. 0:19:14.000 --> 0:19:15.160 Very very valid issues. 0:19:15.400 --> 0:19:17.800 And this is the same thing I think in the beginning, 0:19:18.200 --> 0:19:20.800 how do you regulate something that you don't know what 0:19:20.880 --> 0:19:21.360 it is? 0:19:21.600 --> 0:19:22.920 And we really don't know. 0:19:23.000 --> 0:19:26.640 We have concerns about what it could be, but that's 0:19:26.680 --> 0:19:30.439 not what it is today. So my suggestion to the 0:19:30.520 --> 0:19:33.720 people that are raising all these concerns is they work 0:19:33.840 --> 0:19:38.119 with that obviously, the technologists, the private sector, of the academic, 0:19:38.200 --> 0:19:43.440 university's government regulations and come up with guardrails. Guardrails, not regulations, 0:19:43.680 --> 0:19:46.239 because once you put this in law or code, it 0:19:46.320 --> 0:19:48.800 never gets reversed. And I don't think we know enough 0:19:48.840 --> 0:19:52.439 today to put that in a responsible way, put that 0:19:52.520 --> 0:19:54.520 in place. I know some of the founders of the 0:19:54.560 --> 0:19:58.840 technology are calling for that. I understand that. I understand 0:19:58.840 --> 0:20:03.000 why they're raising those concerns. However, it's the actual details 0:20:03.000 --> 0:20:05.760