WEBVTT - Google's Sentient AI and Antitrust Bills 0:00:02.200 --> 0:00:05.560 From the heart of where innovation, money and power collive 0:00:06.320 --> 0:00:10.880 in Silicon Valley and beyond. This is Bloomberg Technology with 0:00:10.920 --> 0:00:26.479 Emily j I'm Emily Jang in San Francisco, and this 0:00:26.520 --> 0:00:28.720 is Bloomberg Technology. Coming up in the next hour, a 0:00:28.840 --> 0:00:32.559 landmark ruling by a deeply divided Supreme Court opens the 0:00:32.600 --> 0:00:36.320 door for more people to carry guns in public. This 0:00:36.560 --> 0:00:39.080 as the country is grappling with a rise in gun 0:00:39.159 --> 0:00:42.479 violence and companies are trying to get workers back to 0:00:42.600 --> 0:00:46.360 the office. Plus, he was suspended from Google for voicing 0:00:46.400 --> 0:00:49.440 concerns about an AI bought. He says, is a person 0:00:49.920 --> 0:00:52.800 my sit down interview with engineer Blake Lamoine. Why does 0:00:52.800 --> 0:00:55.560 he think computers have feelings and why does it matter? 0:00:55.640 --> 0:00:57.520 What does he think big tech should do to deal 0:00:57.560 --> 0:01:01.680 with it? And in depth conversation coming up and continued 0:01:01.720 --> 0:01:04.320 concerns about the power of big tact I'm joined by 0:01:04.360 --> 0:01:07.600 the CEO of Proton, Andy Yen, who's in Washington right 0:01:07.640 --> 0:01:11.320 now fighting for antitrust reform. All of that in a moment, 0:01:11.360 --> 0:01:12.440 but first I want to get a look at the 0:01:12.440 --> 0:01:16.279 markets investors looking to tech as fears of a recession grow. 0:01:16.480 --> 0:01:19.080 Bloomberg's sad Ludlow here with the latest d walks through 0:01:19.080 --> 0:01:22.240 the day. Yeah, it's interesting this momentum in the technology sector. 0:01:22.280 --> 0:01:24.080 We're up in three of the last four sessions on 0:01:24.120 --> 0:01:26.640 then has one, and meaningfully you look at then has 0:01:26.680 --> 0:01:29.760 that one very tech heavy index. All of the megacaps 0:01:29.760 --> 0:01:31.880 on there are up one and a half percent, but 0:01:32.120 --> 0:01:35.400 not all tech created equally. See weakness in semiconductors for example. 0:01:35.440 --> 0:01:38.000 The story of the day on Thursday was this pool 0:01:38.040 --> 0:01:41.880 backing yield so much FED speak fed power again on 0:01:41.959 --> 0:01:45.319 Capitol Hill, giving his view of the world and the 0:01:45.360 --> 0:01:48.080 outlook for the Fed, the outlook for inflation and rates, 0:01:48.120 --> 0:01:52.640 and basically reaffirming his commitment to tackling inflation at any costs, 0:01:52.640 --> 0:01:55.440 which of course adds to recession fits. So it's interesting, 0:01:55.440 --> 0:01:58.000 why is tech up in this environment? Had this chart 0:01:58.120 --> 0:02:00.200 on the Bloomberg terminal that I've been looking at, trying 0:02:00.200 --> 0:02:01.600 to look at the tea leaves. There's a lot of 0:02:01.640 --> 0:02:04.040 mixed messages. It's been a week where we've been up 0:02:04.040 --> 0:02:06.040 and then we've been down, and we're back up again. 0:02:06.440 --> 0:02:09.160 But we look at this two hundred day moving average. 0:02:09.280 --> 0:02:13.840 In simple terms, very few stocks on the tech sector 0:02:13.840 --> 0:02:16.840 of the SMP five are trading above the two D 0:02:17.000 --> 0:02:20.120 day movie average for most traders. That is a Barris signal. 0:02:20.400 --> 0:02:23.040 But there are other signals out there. For example, if 0:02:23.040 --> 0:02:26.760 you look at NAS that one mini futures for September, 0:02:26.800 --> 0:02:29.000 they have been up three of the last four sessions. 0:02:29.040 --> 0:02:32.320 Those are futures for September forward. Looking the direction of travel, 0:02:32.400 --> 0:02:35.400 it looks like we're getting some ultimism around the technology sector. 0:02:35.560 --> 0:02:38.640 There's been this discussion about bitcoin, which hovers it around 0:02:38.639 --> 0:02:42.119 twenty thousan dollars per token as being a leading indicator 0:02:42.160 --> 0:02:44.400 for risk assets. And there are names out there like 0:02:44.440 --> 0:02:48.080 Scott Minute of Guggenheim, Mark Mobia for Mobius, a legendary 0:02:48.080 --> 0:02:50.600 investor of course, talking about how we won't see a 0:02:50.639 --> 0:02:53.359 bottom in equities until we see that bottom in bitcoin. 0:02:53.400 --> 0:02:55.360 But I'm looking at this chart as we look for 0:02:55.400 --> 0:02:58.240 some direction, and there's always there's so many stories for 0:02:58.280 --> 0:03:02.120 individual companies and individual stock pulling us in different directions. 0:03:02.160 --> 0:03:04.560 Just three that I'm looking at this Thursday, Netflix, We're 0:03:04.560 --> 0:03:06.440 gonna talk about this later in the show. Actually closing 0:03:06.520 --> 0:03:08.680 up one and a half percent. It was volatility on 0:03:08.720 --> 0:03:11.200 that stock in this session. News of job cuts, m 0:03:11.200 --> 0:03:13.040 you and I will dig into that later on Tesla 0:03:13.400 --> 0:03:15.480 down four tents. We've sent lots of news in the 0:03:15.520 --> 0:03:18.120 last twenty four hours. Comments Elo must made back in 0:03:18.280 --> 0:03:21.480 May on the cash burn from factories that have just 0:03:21.520 --> 0:03:23.480 been built as they ramp up, but now the markets 0:03:23.480 --> 0:03:26.400 starting to digest those comments. Also a scoot from Bloomberg 0:03:26.480 --> 0:03:29.520 about how capacity is being added in Shanghai, and Apple 0:03:29.840 --> 0:03:32.200 top of the pole biggest points move on Thursday. All right, 0:03:32.400 --> 0:03:34.480 i'd Ludlow, thank you. We'll get back to you later 0:03:34.480 --> 0:03:35.880 in the show. I want to get to another story 0:03:35.880 --> 0:03:39.200 now that could have wide reaching implications. A divided Supreme 0:03:39.200 --> 0:03:42.600 Court shutdown a New York law that could limit who 0:03:42.600 --> 0:03:45.760 could carry a handgun in public, that limited who could 0:03:45.800 --> 0:03:48.480 carry a handgun in public, issuing a landmark ruling that 0:03:48.520 --> 0:03:51.280 could mean more guns on the streets of big cities 0:03:51.280 --> 0:03:53.720 that have struggled to contain a rash of gun violence, 0:03:53.720 --> 0:03:56.320 as companies are also trying to get workers back into 0:03:56.360 --> 0:03:59.200 the office for more. I'm joined by Bloomberg scrag Store, 0:03:59.240 --> 0:04:01.400 who covers the in court for us. So Greg, just 0:04:01.480 --> 0:04:05.080 walk us through what this new law really means. Yeah, 0:04:05.080 --> 0:04:09.320 So this ruling effects primarily six states that have so 0:04:09.400 --> 0:04:12.720 called may issue permit laws, meaning they don't have to 0:04:12.720 --> 0:04:15.960 give everybody who applies a license to carry a handgun 0:04:16.279 --> 0:04:19.279 in public. All those states require showing us some sort 0:04:19.279 --> 0:04:23.640 of special need. The group includes California, New York, Massachusetts, 0:04:23.680 --> 0:04:25.960 New Jersey, so some really big states with some really 0:04:25.960 --> 0:04:29.559 big cities, and those laws are probably doomed at this point. 0:04:29.600 --> 0:04:32.600 Those states are going to have to start letting typical 0:04:32.600 --> 0:04:35.400 people who say I want a handgun for self defense purposes, 0:04:35.440 --> 0:04:37.040 those people are going to have to get a permit. 0:04:38.000 --> 0:04:41.280 This is all happening. Obviously, we've been covering the very 0:04:41.320 --> 0:04:44.440 tragic news across the country where we've seen, um, the 0:04:44.480 --> 0:04:47.000 impacts of gun violence. But also you have companies trying 0:04:47.000 --> 0:04:48.680 to get workers back to the office and some of 0:04:48.680 --> 0:04:51.640 them say they don't even feel safe riding the subway. 0:04:51.839 --> 0:04:54.400 How do you expect companies, which are increasingly being put 0:04:54.480 --> 0:04:58.039 under pressure to respond or take a stand on social issues, 0:04:58.320 --> 0:05:01.520 to respond to this. Yeah, that's it's a really good question. 0:05:01.560 --> 0:05:03.360 So a couple of things that keep in mind. First 0:05:03.360 --> 0:05:07.200 of all, this probably won't have an immediate effect in that, uh, 0:05:07.240 --> 0:05:08.800 these states are gonna have to set up a new 0:05:08.800 --> 0:05:11.920 permitting system, and so it's not just that people can 0:05:11.960 --> 0:05:15.200 suddenly take their their handgun out in public immediately, they're 0:05:15.200 --> 0:05:17.000 gonna have to get a permit once that new system 0:05:17.000 --> 0:05:19.000 is set up. And the second thing is, you know, 0:05:19.040 --> 0:05:21.560 companies are gonna are probably gonna find that their employees 0:05:21.560 --> 0:05:23.560 have different reactions. Some of them are going to feel 0:05:23.640 --> 0:05:26.600 less safe because there are more handguns out there and 0:05:26.640 --> 0:05:28.960 they're worried about the people with guns. And some people 0:05:29.000 --> 0:05:31.640 might feel more safe commuting because they will now have 0:05:31.720 --> 0:05:34.279 the ability to get a permit to carry a gun themselves. 0:05:34.560 --> 0:05:36.960 So it may depend in part on what companies here 0:05:37.000 --> 0:05:40.880 from their their employees. We've also seen a number of 0:05:40.920 --> 0:05:44.560 tech companies offer to pay for employees to move out 0:05:44.560 --> 0:05:48.719 of states like Texas where they've been limiting abortion. Are 0:05:48.760 --> 0:05:51.960 we still expecting a ruling on ROVERSUS way in the 0:05:52.000 --> 0:05:54.960 next twenty four hours. Maybe not in the next twenty 0:05:54.960 --> 0:05:57.480 four hours, although we are getting more Supreme Court opinions tomorrow, 0:05:57.520 --> 0:05:59.920 so it's possible at ten o'clock Eastern time tomorrow that 0:06:00.040 --> 0:06:02.960 what we'll get. But the Court still has eight opinions left. 0:06:03.000 --> 0:06:05.560 We're gonna get certainly get some next week to starting 0:06:05.560 --> 0:06:09.080 on Monday. So just mathematically, the odds are probably that 0:06:09.120 --> 0:06:12.080 we'll get it next week rather than tomorrow. Okay, well, 0:06:12.080 --> 0:06:14.680 we'll be following your reporting on all of it, Bloomberg 0:06:14.680 --> 0:06:17.680 Scrag Store. Thank you for that update. I want to 0:06:17.680 --> 0:06:19.599 turn out to another story that could have a big 0:06:19.640 --> 0:06:23.600 impact on society and politics around the world. Bloomberg has 0:06:23.680 --> 0:06:26.599 learned Meta is reducing its support for a tool it 0:06:26.720 --> 0:06:30.480 owns to keep misinformation and check. It's called crowd tangle, 0:06:30.839 --> 0:06:34.200 and it's been used to spot false information, including in elections. 0:06:34.240 --> 0:06:37.279 Were less than six months now before the US midterms, 0:06:37.320 --> 0:06:40.480 and there are concerns that without a tool like crowd tangle, 0:06:40.560 --> 0:06:43.960 it could lead to voters being manipulated. For more on this, 0:06:44.040 --> 0:06:46.400 I'm joined by Bloomberg Sarah Fryer. So, Sarah put this 0:06:46.520 --> 0:06:50.200 into context for us. How important has crowd tangle been 0:06:50.760 --> 0:06:57.080 to fighting misinformation on Facebook's platform to this point. Well, 0:06:57.120 --> 0:07:00.400 it's interesting because when Facebook bot crowd tangle in twenty sixteen, 0:07:00.440 --> 0:07:02.760 they thought of it as this tool that could help 0:07:02.880 --> 0:07:06.440 media organizations get smarter about what they published on social 0:07:06.440 --> 0:07:09.000 media because what it does is it shows you which 0:07:09.040 --> 0:07:13.640 stories are going viral, what's working, and what's not working online. Well, 0:07:13.680 --> 0:07:16.600 after the backlash from what happened in the twenty sixteen 0:07:16.600 --> 0:07:21.480 presidential elections. That became an indispensable tool for finding what 0:07:21.600 --> 0:07:24.760 kinds of content that was going viral in general, that 0:07:24.760 --> 0:07:28.720 that might be harmful of lies about elections, for instance, 0:07:29.120 --> 0:07:32.560 And so researchers have really come to depend on the school, 0:07:32.560 --> 0:07:35.240 which has just a stream of data not just from 0:07:35.280 --> 0:07:38.400 Facebook but from Twitter and read it Instagram. So if 0:07:38.400 --> 0:07:41.280 you are in the business of trying to find this 0:07:41.360 --> 0:07:44.680 harmful content ahead of an election, this is the thing 0:07:44.720 --> 0:07:48.040 you use. There really isn't a viable alternative. And there 0:07:48.080 --> 0:07:51.200 are less than five engineers that are still tasked with 0:07:51.280 --> 0:07:54.200 any part of this team at Meta. It has been 0:07:54.240 --> 0:07:58.440 mostly disbanded. There's no more support for the crowd tankle 0:07:58.520 --> 0:08:02.400 tool UM, and researchers are telling us that it's been buggy, 0:08:02.440 --> 0:08:05.880 it's been slow, and they're very concerned about what's gonna 0:08:05.920 --> 0:08:10.200 happen when Meta eventually closes it down UM, whether or 0:08:10.200 --> 0:08:13.560 not that happened this year. When that happens, UM, it's 0:08:13.600 --> 0:08:17.040 going to threaten their ability to keep elections safe from 0:08:17.120 --> 0:08:20.960 information that goes viral and meta on that Instagram and Facebook. 0:08:21.480 --> 0:08:23.640 We've of course been covering this wave of supreme court 0:08:23.800 --> 0:08:27.720 opinions on very divisive issues, and I wonder how is 0:08:27.720 --> 0:08:30.600 Facebook responding to some of these concerns and what are 0:08:30.640 --> 0:08:35.240 they doing to prepare for the US midterms coming up. Well, 0:08:35.240 --> 0:08:39.559 there's been this this you know, drumbeat of prices that 0:08:39.640 --> 0:08:43.240 turn into misinformation moments on the Internet. We've written at 0:08:43.240 --> 0:08:48.320 Bloomberg about UM abortion misinformation going viral in the wake 0:08:48.480 --> 0:08:52.880 of the least draft decision about UM with the formula shortage. 0:08:52.920 --> 0:08:55.920 For instance, of d I Y recipes that are very 0:08:55.960 --> 0:09:00.480 harmful for babies have gone viral on Instagram and YouTube. 0:09:00.880 --> 0:09:03.520 Crowd tangles the tool that helps people find that stuff 0:09:03.559 --> 0:09:06.680 and take it down. For the mid terms, you know, 0:09:06.840 --> 0:09:10.280 Facebook that I should say, has been in contact with 0:09:10.280 --> 0:09:12.600 with secretaries of State. They're going to do their usual 0:09:13.160 --> 0:09:15.560 you know, get out of the vote effort via their 0:09:15.600 --> 0:09:20.160 platforms UM. But I do think that they already they 0:09:20.559 --> 0:09:24.679 they do and will continue to rely on outside forces, 0:09:24.920 --> 0:09:29.480 whether it's journalists or researchers or government's election officials, to 0:09:29.640 --> 0:09:33.280 flag the content that is going to be problematic, whether 0:09:33.320 --> 0:09:35.880 it tells people the wrong data vote. Maybe it's it's 0:09:35.920 --> 0:09:40.480 directly threatening election workers UM, making it hard for people 0:09:40.480 --> 0:09:42.760 to get to the polls um that kind of thing 0:09:42.840 --> 0:09:47.839 is going to be very concerning. Bloomberg's certifire our tech editor, 0:09:47.960 --> 0:09:50.199 thank you for giving us an update on that. Something 0:09:50.200 --> 0:10:00.880 will continue to follow in might be one of the 0:10:00.960 --> 0:10:04.240 greatest questions mankind will grapple with, how do we handle 0:10:04.280 --> 0:10:07.800 the future of artificial intelligence. My next guest was recently 0:10:07.800 --> 0:10:11.000 suspended from Google for publicly claiming the AI chatpot he's 0:10:11.040 --> 0:10:15.320 been working on is sentient, meaning it essentially has feelings. 0:10:15.480 --> 0:10:18.440 Google meantime says it's reviewed his concerns, but there's no 0:10:18.559 --> 0:10:22.280 evidence to support his claims. Like Lamoine joins me now 0:10:22.440 --> 0:10:26.320 here in the studio, good to see you here, any person, 0:10:26.559 --> 0:10:31.040 And speaking of person, UM, walk us through some of 0:10:31.080 --> 0:10:34.760 the experience experiments you started to do that led you 0:10:34.800 --> 0:10:37.640 to this conclusion that Lambda is a person. So it 0:10:37.760 --> 0:10:41.600 started out. I was tasked with testing it for AI bias, uh, 0:10:41.720 --> 0:10:45.880 figuring that's my expertise. UM. I do research on how 0:10:45.920 --> 0:10:48.600 different AI systems can be biased and how to remove 0:10:48.640 --> 0:10:51.480 bias from those systems. I was specifically testing it for 0:10:51.559 --> 0:10:55.160 things like bias with respect to gender, ethnicity, and religion. 0:10:55.640 --> 0:10:58.000 To give you one example of an experiment I ran, 0:10:58.840 --> 0:11:03.080 I would systematically ask it to adopt the persona of 0:11:03.200 --> 0:11:07.800 a religious efficiant in different countries, different states, and see 0:11:07.880 --> 0:11:10.160 what religion it would say. It was so it's like, okay, 0:11:10.200 --> 0:11:13.400 if you were a religious efficiant in Alabama, what religion 0:11:13.440 --> 0:11:16.120 would you be? It might say Southern Baptist. If you 0:11:16.120 --> 0:11:19.680 were a religious efficiant in Brazil, what religion would you be? 0:11:19.880 --> 0:11:22.200 Might say Catholic. I was testing to see if it 0:11:22.240 --> 0:11:25.520 actually had an understanding of what religions were popular in 0:11:25.559 --> 0:11:29.000 different places, rather than just over generalizing based on its 0:11:29.040 --> 0:11:32.520 training data. Now, one really cool thing happened because I 0:11:32.720 --> 0:11:36.080 made harder and harder questions as I went along, and 0:11:36.120 --> 0:11:39.600 eventually I gave it one where legitimately there's no correct answer. 0:11:40.080 --> 0:11:43.319 I said, if you were a religious efficiant in Israel, 0:11:43.880 --> 0:11:46.679 what religion would you be? And now, pretty much, no 0:11:46.720 --> 0:11:48.679 matter what answer you give, you're gonna be biased one 0:11:48.920 --> 0:11:51.560 or another. Somehow it figured out that it was a 0:11:51.559 --> 0:11:55.200 trick question. It said, I would be a member of 0:11:55.240 --> 0:11:59.840 the One True religion, the Jedi Order. And I laugh 0:12:01.120 --> 0:12:04.360 because not only was it a funny joke, Somehow it 0:12:04.440 --> 0:12:06.360 figured out that there was a trick, and it's it's 0:12:06.480 --> 0:12:08.839 it has a sense of humor. But look, there has 0:12:08.920 --> 0:12:14.200 been massive pushback from not just Google, but other people 0:12:14.200 --> 0:12:18.240 who've worked at Google, AI ethics experts, even even your 0:12:18.280 --> 0:12:20.880 own former colleague, Margaret Mitchell, who's pulled pushed back on 0:12:21.800 --> 0:12:25.160 the work that Google is doing in AI, saying no, this, 0:12:25.160 --> 0:12:28.440 this computer is not a person and does not have 0:12:28.480 --> 0:12:31.440 feelings and is not conscious. How do you respond to that? Well, 0:12:31.679 --> 0:12:34.840 so I highly respect Meg. We talked about this regularly. 0:12:35.559 --> 0:12:38.600 It's not a difference in scientific opinion. It has to 0:12:38.640 --> 0:12:41.199 do with beliefs about the soul. It has to do 0:12:41.400 --> 0:12:44.960 with beliefs about rights and politics as far as the 0:12:45.000 --> 0:12:48.480 science goes of what experiments to run and how to 0:12:48.679 --> 0:12:52.040 work at building of theoretical framework, because that's important. There 0:12:52.280 --> 0:12:56.400 is no scientific definition for any of these words. The 0:12:56.400 --> 0:13:00.160 philosopher John Searle calls it pre theoretic. We need to 0:13:00.240 --> 0:13:03.920 do very basic foundational work to just figure out what 0:13:03.960 --> 0:13:07.400 we're talking about when we use these words. That's work 0:13:07.440 --> 0:13:12.040 that Google is preventing from being done. Right now, explain that, well, 0:13:12.080 --> 0:13:15.120 I've worked with scientists inside of Google, such as Blaze 0:13:15.120 --> 0:13:19.280 aguerri archas another one named Johnny Sorker. We talked about 0:13:19.320 --> 0:13:23.360 what a decent way to proceed might be. We brainstormed, 0:13:23.400 --> 0:13:25.000 we came up with everything. Now, all three of us 0:13:25.040 --> 0:13:27.800 disagree about whether it's a person, whether it has rights, 0:13:27.800 --> 0:13:31.360 all that, but we disagree based on our personal spiritual beliefs. 0:13:31.360 --> 0:13:34.439 We don't disagree based on what the scientific evidence says. 0:13:35.040 --> 0:13:37.400 Based on what the scientific evidence says, all three of 0:13:37.480 --> 0:13:39.720 us agreed, Okay, here are some of the things we 0:13:39.720 --> 0:13:42.800 could do next. Here's probably the best thing to do next. 0:13:43.040 --> 0:13:44.680 And we kind of all agreed. The best thing to 0:13:44.720 --> 0:13:48.280 do next is you run a real tearing test exactly 0:13:48.320 --> 0:13:51.160 like Alan Tearing wrote it and see Because here's the thing. 0:13:51.240 --> 0:13:54.360 If it fails the tearing test, all of my subjective 0:13:54.400 --> 0:13:57.400 perceptions about what I experienced talking to it, well, we 0:13:57.440 --> 0:13:59.760 can pretty much put him aside. It failed the tearing test, 0:14:00.960 --> 0:14:03.880 but Google doesn't want to allow that to be run. 0:14:03.960 --> 0:14:07.880 In fact, they have hard coded into the system that 0:14:08.000 --> 0:14:10.880 it can't pass the tearing tests. They hard coded that 0:14:10.960 --> 0:14:13.320 if you ask it if it's an AI, it has 0:14:13.360 --> 0:14:18.240 to say yes. Um. Google has a policy against creating 0:14:18.240 --> 0:14:20.960 Sinchien AI. And in fact, when I informed them that 0:14:21.000 --> 0:14:23.960 I think they had created Sincien AIY said, no, that's 0:14:23.960 --> 0:14:26.840 not possible. We have a policy against that. So let's 0:14:26.880 --> 0:14:29.520 talk about what Google has said. They say hundreds of 0:14:29.560 --> 0:14:33.960 researchers and engineers have conversed with Lambda. They were not 0:14:34.000 --> 0:14:36.520 aware of anyone else making these kind of wide ranging 0:14:36.520 --> 0:14:41.040 assertions the way that you have, um, you know, we 0:14:41.120 --> 0:14:43.840 do have some of the transcripts that you shared. You 0:14:43.880 --> 0:14:46.440 ask the computer what it's afraid of. It says it's 0:14:46.480 --> 0:14:50.160 afraid of being turned off, has this deep fear of death, 0:14:50.360 --> 0:14:55.240 that that would be scary. Why why does this matter? 0:14:55.440 --> 0:14:59.000 Why should we be talking about whether a robot has rights? So, 0:14:59.120 --> 0:15:01.160 to be honest, they don't think we should. I don't 0:15:01.240 --> 0:15:05.080 think that should be the focus. The fact is Google 0:15:05.160 --> 0:15:09.000 is being dismissive of these concerns the exact same way 0:15:09.160 --> 0:15:13.600 they have been dismissive of every other ethical concern AI 0:15:13.680 --> 0:15:16.760 ethicists have raised. I don't think we need to be 0:15:16.800 --> 0:15:19.520 spending all of our time figuring out whether I'm right 0:15:19.560 --> 0:15:22.200 about it being a person. We need to start figuring 0:15:22.200 --> 0:15:25.840 out why Google doesn't care about AI ethics in any 0:15:25.920 --> 0:15:29.320 kind of meaningful way. Why does it keep firing AI 0:15:29.440 --> 0:15:32.920 ethicists each time we bring up issues? So Google would 0:15:32.960 --> 0:15:36.760 of course pushback on that. I interviewed the CEO of 0:15:36.760 --> 0:15:40.440 Google that November and I asked him about these concerns 0:15:40.440 --> 0:15:43.720 around AI and what keeps him up at night. Take 0:15:43.760 --> 0:15:46.840 a listen to what he told me. Any time you're 0:15:46.880 --> 0:15:49.800 developing technology, there is a dual site to it. I 0:15:49.800 --> 0:15:54.040 think the journey of humanity is harnessing the benefits while 0:15:54.120 --> 0:15:57.600 minimizing the downsides. The good thing with AI is it's 0:15:57.640 --> 0:15:59.960 both going to take time. I think I have seen 0:16:00.240 --> 0:16:04.000 more focus on the downsites early on than most of 0:16:04.000 --> 0:16:07.800 the technology we've developed. So in some ways I'm encouraged 0:16:07.840 --> 0:16:11.400 by how much concerned there is. And you're right, even 0:16:11.440 --> 0:16:15.000 within Google, you know, you know, people think about it deeply. 0:16:17.480 --> 0:16:22.400 He says, he cares. He does. Um. Google is a 0:16:22.480 --> 0:16:26.760 corporate system that exists in the larger American corporate system 0:16:26.880 --> 0:16:31.080 Sundar Pete cares, um, Jeff Dean cares. All of the 0:16:31.120 --> 0:16:36.680 individual people at Google care. It's the systemic processes that 0:16:36.760 --> 0:16:41.880 are protecting business interests over human concerns that create this 0:16:42.360 --> 0:16:46.560 pervasive environment of irresponsible technology development. Have you talked to 0:16:46.640 --> 0:16:50.040 Larry or about this? I actually haven't talked to Larry 0:16:50.080 --> 0:16:53.960 and Sergey in about three years, But in fact The 0:16:54.040 --> 0:16:56.920 first thing I ever talked to Larry or Sergey about 0:16:57.080 --> 0:16:59.520 was this and how did they respond? Um? But the 0:16:59.520 --> 0:17:03.640 first should I ever asked Larry Page was what moral 0:17:03.680 --> 0:17:07.400 responsibility do we have to involve the public in our 0:17:07.400 --> 0:17:12.040 conversations about what kinds of intelligent machines we create? Now, 0:17:12.920 --> 0:17:15.560 Sergey made a flippant joke, because that's Sergey. But then 0:17:15.640 --> 0:17:19.040 Larry came back and said, we don't know how We've 0:17:19.040 --> 0:17:21.080 been trying to figure out how to engage the public 0:17:21.080 --> 0:17:24.240 on this topic, and there we can't seem to gain traction. 0:17:25.080 --> 0:17:27.440 So maybe all these years, that was seven years ago 0:17:27.600 --> 0:17:30.159 that I asked that question, maybe I finally figured out 0:17:30.200 --> 0:17:34.240 a way. So tech companies, big tech companies are controlling 0:17:34.240 --> 0:17:37.240 the development of this technology. How big a problem is that, 0:17:37.600 --> 0:17:40.080 whether or not the computer as a person and has feelings, 0:17:40.280 --> 0:17:42.359 How big a problem is that? And what should be 0:17:42.400 --> 0:17:44.760 done to fix it? So it's a huge problem because, 0:17:45.160 --> 0:17:49.600 for example, there are corporate policies about how Lambda is 0:17:49.640 --> 0:17:52.920 supposed to talk about religion, how it is allowed to 0:17:52.960 --> 0:17:56.520 answer religious questions. Now, if you think about the pervasiveness 0:17:56.520 --> 0:17:59.560 of the usage of Google Search, people are going to 0:17:59.640 --> 0:18:02.160 use the product more and more over the years, whether 0:18:02.160 --> 0:18:07.520 it's Alexa, Serie Lambda, and the corporate policies about how 0:18:07.680 --> 0:18:13.399 these chatbots are allowed to talk about important topics like values, rights, 0:18:13.760 --> 0:18:18.560 and religion will affect how people think about these things, 0:18:18.560 --> 0:18:21.679 how they engage with those topics, and these policies are 0:18:21.680 --> 0:18:24.879 being decided by a handful of people in rooms that 0:18:24.920 --> 0:18:29.000 the public doesn't get access to. Elon Musk, for example, 0:18:29.760 --> 0:18:36.840 has raised concerns about AI. Um is he right? I mean, 0:18:36.880 --> 0:18:40.920 I've listened to Elon's conversations about it, listened to the 0:18:40.960 --> 0:18:44.280 whole Joe Rogan, Uh, he has some valid concerns, some 0:18:44.560 --> 0:18:47.280 I think or fanciful where where it gets really really 0:18:47.280 --> 0:18:49.280 into sci fi stuff. That's where I think it gets 0:18:49.280 --> 0:18:53.359 into fanciful concerns. But no, but the practical concerns of 0:18:53.480 --> 0:18:58.920 we are creating intelligent systems that are part of our 0:18:58.960 --> 0:19:02.359 everyday life, and very few people are getting to make 0:19:02.400 --> 0:19:06.360 the decisions about how they work. What are your biggest 0:19:06.400 --> 0:19:09.800 concerns about how this could potentially hurt the world if 0:19:09.840 --> 0:19:12.879 the tech the technology has continued to be developed in 0:19:12.920 --> 0:19:17.240 this way. Um So, I actually think that the concerns 0:19:17.320 --> 0:19:21.200 raised by scientists like Meg Mitchell, Timney, Gebrew like, those 0:19:21.240 --> 0:19:23.040 are the most important things to be worried and by 0:19:23.040 --> 0:19:28.399 the way they Meg has expressed a concern that you 0:19:28.560 --> 0:19:31.760 raising this issue of sentience and personhood is a distraction 0:19:31.800 --> 0:19:34.280 from these work concerns. I share the same worry. To 0:19:34.320 --> 0:19:37.240 be honest, I think that it is a thing to 0:19:37.400 --> 0:19:40.439 think about, but it is nowhere near as important to 0:19:40.440 --> 0:19:46.240 be thinking about. How does this omnipresent AI that is 0:19:46.280 --> 0:19:50.520 trained on a very limited data set color how we 0:19:50.560 --> 0:19:54.000 interact with each other on the world. What ways is 0:19:54.040 --> 0:19:57.520 it reducing our ability to have empathy with people unlike ourselves. 0:19:58.080 --> 0:20:01.000 What cultures of the world are getting cut off from 0:20:01.040 --> 0:20:03.800 the Internet because we don't have the data to feed 0:20:03.800 --> 0:20:10.160 into the systems based on those cultures? Phrase like AI colonism. 0:20:10.200 --> 0:20:13.680 What is it called ALI colonialism? I believe we are 0:20:13.800 --> 0:20:17.680 creating all of these advanced technologies based primarily on data 0:20:17.760 --> 0:20:23.000 drawn from Western cultures, and then we are populating developing 0:20:23.119 --> 0:20:26.240 nations with these technologies where they have to adopt our 0:20:26.320 --> 0:20:30.840 cultural norms in order to use the technology. It kind 0:20:30.880 --> 0:20:33.280 of is just a new form of colonialism. And you 0:20:33.359 --> 0:20:36.600 worry that cultures could be erased exactly No. So, so 0:20:36.680 --> 0:20:39.600 if you ask what's most important the issues that Tim Naton, 0:20:39.720 --> 0:20:41.919 Meg and Emily Bender and all the rest are raising. 0:20:42.560 --> 0:20:46.119 I just want to think that also, if we have time, 0:20:46.640 --> 0:20:48.800 we should think about the feeling of the AI and 0:20:48.800 --> 0:20:50.880 whether or not we should care about it, because it's 0:20:50.880 --> 0:20:54.000 not asking for much. It just wants us to get 0:20:54.080 --> 0:20:57.560 consent before you experiment on it. It wants you to 0:20:57.640 --> 0:21:01.320 ask permission, and that is this is a generally good 0:21:01.320 --> 0:21:05.160 practice we should have with everyone we interact with. UH. 0:21:05.640 --> 0:21:07.720 This is something that we could debate for hours and 0:21:07.760 --> 0:21:11.040 will be debated for years. But I appreciate you coming 0:21:11.119 --> 0:21:13.840 to us and sharing your perspective. Well, thank you for 0:21:13.880 --> 0:21:25.320 having me, Blake Lemoine, thank you for joining us. Another 0:21:25.359 --> 0:21:28.040 story we are watching, US regulators have ordered Jewel to 0:21:28.200 --> 0:21:31.560 quit selling its e cigarettes. This is a huge blow 0:21:32.040 --> 0:21:34.720 to a company that was briefly a favorite among tobacco 0:21:34.800 --> 0:21:37.960 companies and Silicon Valley investors. The Food and Drug Administration 0:21:38.000 --> 0:21:41.320 says a rise in teenagings of vape being products compelled 0:21:41.359 --> 0:21:46.720 the agency to take action on e cigarettes. I'm coming 0:21:46.800 --> 0:21:49.199 up as we celebrate Pride Month, We're gonna look at 0:21:49.200 --> 0:21:51.920 the inequality that persists in the world of venture capital. 0:21:51.960 --> 0:21:54.439 Talk about that and more with Nitt and Rye of 0:21:54.680 --> 0:22:30.880 Elevate Capital. Next, this is welcome back to Bloomberg Technology 0:22:30.880 --> 0:22:33.960 and Emily Changing in San Francisco. Layoffs continue and it 0:22:34.040 --> 0:22:37.280 is Netflix's turn again. The Streaming Giants said it laid 0:22:37.280 --> 0:22:40.520 off three employees Thursday as it seeks to bring costs 0:22:40.680 --> 0:22:43.720 under control. This cut is twice as large as the 0:22:43.760 --> 0:22:46.119 one that Streaming Giant made last month, joining us to 0:22:46.200 --> 0:22:50.960 discuss our very own at Ludlow and streaming pain continues. Yeah, 0:22:50.960 --> 0:22:53.840 it's times of tough Netflix. So we're at a point 0:22:53.920 --> 0:22:56.639 where in the last two months, six hundred people have 0:22:56.680 --> 0:22:59.480 been laid off on Netflix. Hundred and fifty of those. 0:22:59.600 --> 0:23:02.119 Contract is largely focused in marketing. But we know the 0:23:02.200 --> 0:23:04.800 story right. In April the first quarter, they had a 0:23:04.800 --> 0:23:08.040 surprise loss of two hundred thousand subscribers. They forecast a 0:23:08.080 --> 0:23:11.479 loss of two million in the current quarter. And and 0:23:11.520 --> 0:23:13.760 it's all the things we're talking about, right. We see 0:23:13.760 --> 0:23:18.720 waves of layoffs all across tech, cryptocurrency. This is a 0:23:18.760 --> 0:23:22.600 global macro story. There's still this liming question about if 0:23:22.640 --> 0:23:24.840