WEBVTT - SpaceX Fires Employees and Google's AI Issues 0:00:02.200 --> 0:00:05.480 From the heart of where innovation, money and power COLLI 0:00:06.320 --> 0:00:10.840 in Silicon Valley and beyond. This is Bloomberg Technology with 0:00:10.920 --> 0:00:27.120 Emily jay I. Remember ly check in San Francisco and 0:00:27.160 --> 0:00:30.400 this is Bloomberg Technology coming up the next hour. Multiple 0:00:30.480 --> 0:00:35.800 SpaceX employees fired after an open letter criticizing Elon Musk. 0:00:36.000 --> 0:00:38.800 What this says about his leadership and what it means 0:00:39.120 --> 0:00:44.720 for Twitter Plus? Does AI have feelings? Google just suspended 0:00:44.720 --> 0:00:47.240 an engineer who claims the answer to that question is yes. 0:00:47.680 --> 0:00:50.560 We will have an in depth conversation with Google's former 0:00:50.640 --> 0:00:55.880 AI ethicist about the limits and power of this technology 0:00:56.040 --> 0:00:59.080 and cryptotraders go from feeling the fear of missing out 0:00:59.080 --> 0:01:02.280 to straight up fear. We'll talk to micro Strategy CEO 0:01:02.320 --> 0:01:05.520 Michael Saylor about his big bet on bitcoin and if 0:01:05.560 --> 0:01:11.240 he has any regrets. SpaceX now has fired several employees 0:01:11.280 --> 0:01:14.639 involved in an open letter criticizing the behavior of CEO 0:01:14.720 --> 0:01:17.639 Elon Musk. This according to an internal memo that began 0:01:17.760 --> 0:01:21.920 circulating among staff this week. The letter, seen by Bloomberg 0:01:21.920 --> 0:01:25.840 called Musk's behavior and tweets quote a frequent source of 0:01:25.880 --> 0:01:30.360 distraction and embarrassment, and called on SpaceX leadership to condemn 0:01:30.360 --> 0:01:34.680 and distance itself from his quote personal brand. Bloomberg's bad 0:01:34.760 --> 0:01:38.640 Ludlow here now to discuss. So Bloombook obtained a copy 0:01:38.640 --> 0:01:41.440 of the Soper letter. What exactly did it say? Yes, 0:01:41.520 --> 0:01:44.360 so this is a small select group of employees who 0:01:44.440 --> 0:01:47.760 called on SpaceX Is management, including Gwin short World, the 0:01:47.840 --> 0:01:53.600 CEO and president, to basically publicly distance itself from Musk 0:01:53.800 --> 0:01:56.120 what he says, what he believes, what he was doing, 0:01:56.400 --> 0:02:00.040 because they felt that it was quote, an embarrassment and 0:02:00.480 --> 0:02:02.400 that it was basically impacting the work that space It 0:02:02.480 --> 0:02:03.800 was trying to do. This was the view of a 0:02:03.800 --> 0:02:06.240 small number of employees. It was shared as sort of 0:02:06.280 --> 0:02:10.679 something you could sign vera QR code through internal messaging channels, 0:02:10.760 --> 0:02:12.560 and it picked up some momentum. You know, they did 0:02:12.560 --> 0:02:19.160 get signatures until management acted. What exactly did management then? 0:02:20.600 --> 0:02:24.080 So in a second internal memo which was sent by 0:02:24.120 --> 0:02:27.000 Gwyn shot Well, the company's CEO, which Blomberg has seen, 0:02:28.120 --> 0:02:31.120 they basically said this wasn't helpful. You know, they conducted 0:02:31.160 --> 0:02:35.280 an investigation into the origins of this open letter from staff, 0:02:35.840 --> 0:02:38.840 and as you said, and I think we had the 0:02:38.919 --> 0:02:40.920 quote from Gwyn shot Well about sort of the broader 0:02:40.960 --> 0:02:44.320 impact of it. They ended up firing that what they 0:02:44.320 --> 0:02:46.840 said a number of employees, we don't have a firm 0:02:47.120 --> 0:02:49.360 number and how many it is. But this is what's interesting. 0:02:49.639 --> 0:02:53.720 She calls this overreaching activism. You know, so clearly there's 0:02:53.720 --> 0:02:56.360 a group of employees within the business who have strong opinion. 0:02:56.400 --> 0:03:00.720 And this what she says elsewhere in the memo that's 0:03:00.760 --> 0:03:02.720 frankly too long for us to share on the screen, 0:03:02.800 --> 0:03:06.400 is that this select group of employees were putting pressure 0:03:06.400 --> 0:03:09.920 on their peers to sign something that they didn't believe in. 0:03:09.960 --> 0:03:12.760 These are paraphrasing Gwin Shotwell's words, did a bunch of 0:03:12.800 --> 0:03:15.760 employees sign it? And are we've seen arranging numbers up 0:03:15.800 --> 0:03:19.760 there from several hundreds several thousand. SpaceX has twelve thousand 0:03:19.760 --> 0:03:24.639 employees around the world essentially, so it gained traction um. 0:03:25.880 --> 0:03:28.480 But ultimately, you know, what Gwyn Shotwell goes on to 0:03:28.520 --> 0:03:31.200 say in that memo is that this is a distraction 0:03:31.240 --> 0:03:33.240 from their end goal and their own goals we know 0:03:33.520 --> 0:03:36.080 is getting to mars Well. This isn't the first story 0:03:36.360 --> 0:03:40.120 in the last few weeks about Elon Musk's behavior at SpaceX. 0:03:40.160 --> 0:03:43.440 There was another story about accusing him of sexual harassment. Right, 0:03:44.040 --> 0:03:47.080 and Gwen shot Well also defended him in that instance 0:03:47.080 --> 0:03:50.680 as well. Correct. So last month, Business Insider reported that 0:03:51.040 --> 0:03:54.680 the company space X settled with a former contract employee 0:03:54.960 --> 0:03:58.360 in two thousand eighteen for two fifty thousand U S 0:03:58.400 --> 0:04:03.720 dollars and that employee in question was a contract air 0:04:03.760 --> 0:04:07.360 steward aboard a space x jet that Must would used 0:04:07.360 --> 0:04:10.720 to travel. And straight away we should point out must 0:04:10.760 --> 0:04:15.320 denies not only denies the claims made, but he actually 0:04:15.360 --> 0:04:17.839 goes on to say in a series of tweets earlier 0:04:17.880 --> 0:04:19.719 in the month or last month that he basically saw 0:04:19.760 --> 0:04:21.760 this as a political attack on him, that it was 0:04:21.800 --> 0:04:27.680 sort of a calculated initiative to to impact his reputation. Um, 0:04:27.720 --> 0:04:30.640 you know, and he questioned the source who is the 0:04:30.640 --> 0:04:33.839 original source of the business inside of story. So regardless, 0:04:33.880 --> 0:04:35.400 we don't get to the bottom of it. I should 0:04:35.440 --> 0:04:38.600 point out SpaceX does have a comms team, a PR team. 0:04:38.920 --> 0:04:41.840 I messaged them regularly. They are real people, but they 0:04:41.880 --> 0:04:44.919 have not responded to multiple recomment on this story, on 0:04:44.960 --> 0:04:47.640 the open letter story, so you know, what we have 0:04:47.720 --> 0:04:50.040 to go on is these internal messages that Bloomberg has 0:04:50.000 --> 0:04:53.680 scene interesting. There has been some other news for SpaceX 0:04:53.720 --> 0:04:56.240 today they did have another successful life. I mean, this 0:04:56.279 --> 0:04:57.880 is what it all comes back to down to. So 0:04:57.960 --> 0:05:00.719 today was kind of another bog standard style launch, another 0:05:00.800 --> 0:05:04.440 three fifty three Starlink satellites deployed to orbit. But it 0:05:04.480 --> 0:05:07.800 was also a milestone for reusability. It was the hundred 0:05:07.839 --> 0:05:12.520 of launch using a proven, a flight proven rocket or booster. 0:05:12.520 --> 0:05:15.039 In other words, this is the hundred of time that 0:05:15.160 --> 0:05:18.240 SpaceX has sent up a rocket that has previously flown 0:05:18.720 --> 0:05:22.279 and landed it successfully, and it's just changed the game, right. 0:05:22.320 --> 0:05:24.360 This is the whole point of the space X story 0:05:24.760 --> 0:05:28.640 that this reusability angle allows them to go with such 0:05:28.680 --> 0:05:32.279 regularity that it becomes routine, but also that it makes 0:05:32.600 --> 0:05:36.280 access to space much more affordable for little satellite providers 0:05:36.360 --> 0:05:39.400 and themselves building out the Starlink network. All right, well, 0:05:39.600 --> 0:05:43.800 TBC to be continued as I'm sure this narrative will 0:05:43.839 --> 0:05:57.440 be Thank you thanks. Earlier this week, a Google engineer 0:05:57.480 --> 0:06:00.320 working on the company's AI development team was suspen ended 0:06:00.320 --> 0:06:04.680 after claiming a chatbot actually has feelings. Blake Lemoyne was 0:06:04.720 --> 0:06:07.919 placed on paid leave last week after he posted on 0:06:08.040 --> 0:06:12.000 Medium that he had encountered a quote Sentien AI, igniting 0:06:12.040 --> 0:06:15.440 a fiery debate about the possibilities and limits of this 0:06:15.560 --> 0:06:20.120 cutting edge technology. Dr Margaret Mitchell, Hugging Face, Chief Ethics 0:06:20.160 --> 0:06:24.720 Scientists and researcher and former Google AI employee who worked 0:06:24.760 --> 0:06:29.520 on the AI development team, joins us now to discuss Margaret, 0:06:29.560 --> 0:06:31.720 thank you so much for joining us, or I should say, 0:06:32.080 --> 0:06:38.640 Dr Mitchell, given your expertise, do you think Blake Lemoyne 0:06:38.920 --> 0:06:46.000 is right? Does this AI? Does this spot have feelings? Uh? Okay, 0:06:46.000 --> 0:06:50.120 well no, I don't think it does. Um. I certainly 0:06:50.160 --> 0:06:55.360 don't think it has feelings definitely not consciousness or sentience, 0:06:55.800 --> 0:07:00.240 which which is what the claims have been. So what 0:07:00.360 --> 0:07:03.240 does this, though tell us about the potential or power 0:07:04.279 --> 0:07:08.080 four AI am bought to full human beings into thinking 0:07:08.279 --> 0:07:13.960 that they're real. Yeah, there's a few things going on. UM. 0:07:14.040 --> 0:07:18.400 On the one hand, we have uh psychological effects of 0:07:18.600 --> 0:07:23.679 UM interacting with things that are human like UM. So uh, 0:07:23.720 --> 0:07:27.400 we we tend to anthropomorphize UM. We tend to put 0:07:27.480 --> 0:07:32.240 intentionality UM into things that we're that we're inter interacting 0:07:32.280 --> 0:07:34.800 with that seem human like. I think people are sort 0:07:34.840 --> 0:07:37.160 of used to doing this with our pets and things, 0:07:37.240 --> 0:07:40.520 you know, creating like whole dialogues and conversations UM, but 0:07:40.600 --> 0:07:44.240 also with like you know, uh, stuffed animals and tamagotchis 0:07:44.280 --> 0:07:48.080 and things like that. And there's also been psychological studies 0:07:48.360 --> 0:07:53.000 showing that we have a propensity to um impute intentionality 0:07:53.760 --> 0:07:58.920 into um non conscious beings UM when they show some 0:07:59.000 --> 0:08:03.920 sort of properties like speaking uh, like vulnerability UM or 0:08:04.080 --> 0:08:08.240 or movement that's aligned with human like movement UM. On 0:08:08.280 --> 0:08:12.280 the other hand, we also have a lot of companies 0:08:12.280 --> 0:08:17.800 working in AI using the language of human cognition UM 0:08:18.000 --> 0:08:21.120 so saying things like chain of thought, saying things like 0:08:21.240 --> 0:08:26.000 reasoning um, you know, essentially comparing the models that they're 0:08:26.040 --> 0:08:29.200 working with to the brain, which makes some sense, but 0:08:29.280 --> 0:08:32.520 you really have to temper that with with the details 0:08:32.559 --> 0:08:35.400 of this essentially being um a bunch of a bunch 0:08:35.440 --> 0:08:38.880 of calculations UM. So we have a few things going on, 0:08:39.000 --> 0:08:43.480 the psychological illusions UH and the language that companies are 0:08:43.600 --> 0:08:48.559 using around the technology they're building. So, given the complexity 0:08:48.600 --> 0:08:53.160 of this, what are your biggest concerns about? For example, 0:08:53.200 --> 0:08:57.040 these transcripts that that Blake Lamoyne published where the computer 0:08:57.160 --> 0:08:59.880 is saying I'm scared of dying, I'm scared of being 0:09:00.000 --> 0:09:05.760 turned off. Yeah, I mean I I think I echo 0:09:06.240 --> 0:09:10.680 UM a lot of researchers in this space, UM, where 0:09:11.400 --> 0:09:13.959 I think we all sort of feel like sentience is 0:09:14.320 --> 0:09:17.959 not the point here. UM. I think it's I think 0:09:17.960 --> 0:09:20.480 it's important to note that we are not going to 0:09:20.520 --> 0:09:24.440 get an agreement on sentience or consciousness anytime soon. People 0:09:24.480 --> 0:09:27.040 are going to see sentience, people are going to see 0:09:27.040 --> 0:09:30.360 consciousness UM, and that will probably go on, you know, 0:09:30.480 --> 0:09:34.800 indefinitely where we just have a disagreement. UM. But when 0:09:34.880 --> 0:09:39.680 you do have people starting to see sentience in consciousness, UM, 0:09:39.720 --> 0:09:42.840 it starts to bring up things like, UM, you know, 0:09:42.880 --> 0:09:45.480 like robot rights, all this work that's been done on 0:09:45.960 --> 0:09:50.600 what the personhood of these models might be. UM. Well, 0:09:50.640 --> 0:09:53.800 at the same time, you have technology that you know, 0:09:54.000 --> 0:09:58.880 essentially discriminates, you know, against black, black and brown people, UM, 0:09:58.920 --> 0:10:04.000 poorly represents women and reflects misogynistic viewpoints. Uh. So there's 0:10:04.000 --> 0:10:06.679 something to be said for an obsession with the personhood 0:10:06.960 --> 0:10:11.160 of AI and systems UM and thinking about the rights 0:10:11.160 --> 0:10:14.160 that they might have while not actually doing good work 0:10:14.280 --> 0:10:18.400 on the rights of actual people. UM. On top of Oh, 0:10:18.559 --> 0:10:20.480 I have so much to say, but yeah, you have 0:10:20.520 --> 0:10:23.640 another question. I'm sure well, you know, and of course 0:10:23.840 --> 0:10:25.559 you know the history behind this is that you were 0:10:25.600 --> 0:10:29.079 fired for your work and sounding the alarm about sexism 0:10:29.120 --> 0:10:33.000 and racism in a in AI at Google. So it 0:10:33.000 --> 0:10:36.080 sounds to me like you're saying, this isn't the problem. 0:10:36.160 --> 0:10:38.680 We shouldn't be asking if robots have feelings and rights. 0:10:38.720 --> 0:10:42.960 We should be asking if AI is gender blind and 0:10:43.000 --> 0:10:47.040 color blind and making sure, um that we're focusing on 0:10:47.040 --> 0:10:51.120 all of these other things that are far more important. Yeah, 0:10:51.160 --> 0:10:54.160 I mean, so it's not gender blind, it's actually uh 0:10:54.679 --> 0:10:58.560 targeting gender and negative ways. UM. And so for example, 0:10:58.600 --> 0:11:00.880 we know that a lot of these systems are trained 0:11:01.040 --> 0:11:06.400 on UM text data from from websites that have misogynistic 0:11:06.440 --> 0:11:12.160 tendencies UM and UH websites that are predominantly white and 0:11:12.280 --> 0:11:16.760 male UM and and actually US based. UM. So there's 0:11:16.760 --> 0:11:19.840 all these kinds of things that are being UM propagated 0:11:19.880 --> 0:11:22.800 by these systems that are really problematic. UM. And they 0:11:22.920 --> 0:11:27.640 become even more problematic when people start to be affected 0:11:27.640 --> 0:11:30.320 by the systems as they interact with them. UM. So, 0:11:30.360 --> 0:11:33.199 in the case of consciousness, UM, you have the concern 0:11:33.280 --> 0:11:37.200 that people might be persuaded to do horrible things. UM. 0:11:37.280 --> 0:11:41.120 You also have you know, concerns around bullying and hate bots, 0:11:41.679 --> 0:11:43.440 uh and these kinds of things that can you know, 0:11:43.559 --> 0:11:47.480 really hurt people. Um. And then you know you also 0:11:47.640 --> 0:11:52.000 have um, these subtle effects of you know, in search 0:11:52.120 --> 0:11:55.120 ranking results, what will tend to appear at the top 0:11:55.160 --> 0:11:57.520 of that ranking And if it's a function of these 0:11:57.520 --> 0:12:00.240 sorts of language models um as Google, for exap people 0:12:00.320 --> 0:12:03.480 has said, UM, they are, then you're going to have 0:12:03.559 --> 0:12:07.120 these bias effects influencing search results in such a way 0:12:07.160 --> 0:12:09.960 that you tend to see the viewpoints of white men, 0:12:10.280 --> 0:12:13.200 you know, at the top of the search drinking results 0:12:13.280 --> 0:12:16.040 as opposed to you know, black women, And that is 0:12:16.080 --> 0:12:19.160 sort of a echo a chamber effects where it's like 0:12:19.720 --> 0:12:23.400 the privileged gets more privileged, right, privilege gets privileged, while 0:12:23.440 --> 0:12:27.400 the marginalized become more marginalized. Now Google has come out 0:12:27.400 --> 0:12:29.400 saying that in this particular case, when it comes to 0:12:29.440 --> 0:12:32.400 Blake Lemoyne, that you know, hundreds of researchers have interacted 0:12:32.440 --> 0:12:36.079 with the same technology haven't expressed these concerns. I also 0:12:36.120 --> 0:12:39.760 sat down with Alphabet and Google CEO Sundar pacha I 0:12:40.559 --> 0:12:43.960 last year and asked him about concerns around AI from 0:12:43.960 --> 0:12:47.559 within Google itself. From people like yourself. I asked him 0:12:47.600 --> 0:12:49.840 what keeps him up at night? Take a listen to 0:12:49.920 --> 0:12:54.480 what he had to say. Anytime you're developing technology, there 0:12:54.559 --> 0:12:56.800 is a dual site to it. I think the journey 0:12:56.800 --> 0:13:01.880 of humanity is harnessing the benefits while minimizing the downsides. 0:13:02.559 --> 0:13:04.640 The good thing with AI is it's both going to 0:13:04.679 --> 0:13:08.480 take time. I think I've seen more focus on the 0:13:08.600 --> 0:13:11.800 downsides early on than most of the technology we've developed. 0:13:11.800 --> 0:13:15.360 So in some ways I'm encouraged by how much concerned 0:13:15.400 --> 0:13:19.320 there is. And you're right, even within Google, you know, 0:13:19.840 --> 0:13:24.400 you know people think about it deeply. Margaret, Do you 0:13:24.480 --> 0:13:27.840 think he and Google are leading on these issues in 0:13:27.880 --> 0:13:33.200 the right way? No, clearly not. I mean everyone I think, 0:13:33.320 --> 0:13:35.600 at least in tech is familiar with this notion of 0:13:35.640 --> 0:13:40.800 tech solutionism UM. And there's no lack of pr and 0:13:40.920 --> 0:13:44.440 calms around the benefits of AI and really trying to 0:13:44.480 --> 0:13:47.600 push it as beneficial for humanity and all these sorts 0:13:47.640 --> 0:13:51.120 of things. It's it's a very small minority who speaks 0:13:51.200 --> 0:13:54.840 up about the downside. So I would say that UM 0:13:55.120 --> 0:14:01.319 soon Dar's characteration characterization is false UM and frustrating false UM. 0:14:01.640 --> 0:14:03.680 And one of the reasons. I think that there's a 0:14:03.720 --> 0:14:06.920 desire to stay away from the downsides. UM. In addition 0:14:06.960 --> 0:14:10.480 to you know, concern around profit is that it also 0:14:10.520 --> 0:14:13.200 starts to open up liability. Right, so if you have 0:14:13.280 --> 0:14:17.520 systems that you can show work worse on black women, 0:14:17.880 --> 0:14:20.480 then now it starts to bring up questions of discrimination 0:14:20.560 --> 0:14:25.480 within the systems. UM. So it behooves companies to try 0:14:25.520 --> 0:14:27.920 and say, oh, the downsides are you know, are being 0:14:28.040 --> 0:14:32.200 over examined, and try and kind of shut that conversation down. 0:14:32.640 --> 0:14:35.080 But I think what's actually happening is that the small 0:14:35.160 --> 0:14:38.800 set of people who have been speaking about ethical concerns 0:14:39.320 --> 0:14:42.920 are starting to be listened to because people are seeing 0:14:43.040 --> 0:14:46.720 the negative effects. UM. And I think that's really what's 0:14:46.760 --> 0:14:49.760 happening is a desire on the corporate side to shut 0:14:49.840 --> 0:14:52.280 the conversation down for a lot of sort of incentives 0:14:52.320 --> 0:14:55.920 they have, and then people actually seeing the downside it's 0:14:55.960 --> 0:14:59.320 and not having an effect on what gets reported. Do 0:14:59.400 --> 0:15:04.120 you think a little wine should have been suspended? Uh? No, 0:15:04.440 --> 0:15:07.880 I don't. UM. I so I should say that um. 0:15:08.280 --> 0:15:11.320 Blake and I are really good friends. We worked together 0:15:11.400 --> 0:15:14.480 at Google. We wrote a paper together actually on how 0:15:14.520 --> 0:15:21.640 to um mitigate problematic biases in UM in machine learning systems. Um. 0:15:21.680 --> 0:15:26.040 He's a very very bright guy. UM. So I'm a 0:15:26.040 --> 0:15:28.800 little bit worried that there's sort of this reductive narrative 0:15:28.840 --> 0:15:32.640 that there's something like fundamentally wrong with him or something. UM. 0:15:32.680 --> 0:15:35.440 He he has a lot of dimensions. UM. And I 0:15:35.480 --> 0:15:38.400 think Google could have done a much better job at 0:15:38.480 --> 0:15:43.480 engaging with him rather than this you know, very alienating 0:15:43.920 --> 0:15:46.720 sort of experience that they gave him instead. UM. I 0:15:46.720 --> 0:15:48.920 think it shows a weakness on Google's part to be 0:15:49.040 --> 0:15:53.720 able to um uh, to be able to be open 0:15:53.800 --> 0:15:58.120 to different kinds of experiences and perspectives. So one, are 0:15:58.240 --> 0:16:02.800 your biggest fears if Google continues to develop the technology 0:16:03.400 --> 0:16:06.840 at the pace that it is developing, this technology continues 0:16:07.000 --> 0:16:10.280 to you know, potentially not listen to this as you say, 0:16:10.400 --> 0:16:13.720 minority of voices that are speaking up paint the picture 0:16:13.840 --> 0:16:16.680 of of what you fear the world could look like 0:16:17.520 --> 0:16:22.480 if Google continues on this path. Oh no, that is 0:16:22.520 --> 0:16:25.040 a very big question. UM. And I'm not a good painter, 0:16:25.920 --> 0:16:28.680 I should mention I'm a computer scientist, so I might 0:16:28.720 --> 0:16:31.400 not you know, be as elephant at this UM. But 0:16:31.720 --> 0:16:35.280 you know, we're already seeing a lot of what we 0:16:35.320 --> 0:16:39.240 can expect to happen in the future, but even worse. UM. 0:16:39.320 --> 0:16:42.359 So just recently, someone released a ton of hate thoughts 0:16:42.480 --> 0:16:46.320 UM and then made the model available to the public UM. 0:16:46.400 --> 0:16:49.400 And so we are going to likely see an increase 0:16:49.600 --> 0:16:56.320 of hateful UM intelligence seeming systems across our interactions online 0:16:56.320 --> 0:16:59.120 and on social media. UM. And this includes things like 0:16:59.160 --> 0:17:03.960 bullying as is really problematic, persuasion into sort of more 0:17:04.320 --> 0:17:09.760 extremist UM areas. UM. I think we're going to see 0:17:10.200 --> 0:17:15.720 UH further sort of marginalization and worsening of power differentials. 0:17:15.760 --> 0:17:19.680 So as you know, a company like Google a mass 0:17:19.800 --> 0:17:24.320 is more and more um ability to affect people's sense 0:17:24.480 --> 0:17:27.640 of of what's true in the world through search ranking results, 0:17:28.280 --> 0:17:32.160 through the sort of UH products they're making. UM. It 0:17:32.200 --> 0:17:35.840 means that the voices of people who have less access 0:17:35.880 --> 0:17:38.840 to the Internet, for example, are going to disappear more 0:17:38.840 --> 0:17:41.280 and more, while Google a mass is more and more 0:17:41.320 --> 0:17:45.800 power UM. And so I'm very very concerned about how 0:17:45.880 --> 0:17:50.120 much the sort of technology moving forward empowers Google UM 0:17:50.200 --> 0:17:54.159 and sort of lack of respects that I've seen for 0:17:54.640 --> 0:17:58.960 very serious ethical concerns. You know, misinformation obviously is one 0:17:59.040 --> 0:18:01.640 alongside some of the models that have come out recently, 0:18:01.680 --> 0:18:04.080 We're not gonna know what's real. There's going to be 0:18:04.320 --> 0:18:08.720 text text based misinformation, so so news that's wrong, uh, 0:18:08.800 --> 0:18:12.600 image based missed information so images that look real that 0:18:12.600 --> 0:18:16.119 that are not real. UM. And video based as well, 0:18:16.359 --> 0:18:20.520 and also audio based. So essentially all of the main 0:18:20.640 --> 0:18:26.040 ways that we consume information online will now no longer 0:18:26.200 --> 0:18:31.680 be very easy uh to trace back to reality. UM. 0:18:31.720 --> 0:18:39.480 And that means mass misunderstanding. UM. So yeah, scary, Dr Mitchell. Uh. 0:18:40.280 --> 0:18:42.320 This we could have this conversation for hours, and I 0:18:42.640 --> 0:18:44.560 know we're gonna be having it for years. I'd love 0:18:44.720 --> 0:18:46.480 to have you back to talk more about your work 0:18:46.480 --> 0:18:49.000 at Hugging Face. I know that there you are taking 0:18:49.000 --> 0:18:51.280 a different approach to a lot of these issues. UM, 0:18:51.280 --> 0:18:54.000 But because of commercials, we're gonna have to leave it here. Um. 0:18:54.080 --> 0:18:57.720 Dr Margaret Mitch and Mitchell Hugging Face chief ethics scientists 0:18:57.800 --> 0:19:01.520 and researcher. UM, thank you uh for joining us today 0:19:01.520 --> 0:19:05.119 and help us work through some of these very complex issues. 0:19:05.119 --> 0:19:07.920 Will have much more ahead stay with us. This is Bloomberg. 0:19:17.440 --> 0:19:19.720 TikTok says it has reached an agreement with Oracle to 0:19:19.800 --> 0:19:23.560 store data from US users on Oracle servers. The deal 0:19:24.000 --> 0:19:27.440 has been in the work since following concerns of security 0:19:27.560 --> 0:19:30.160 risks linked to the Chinese owned app. This news comes 0:19:30.200 --> 0:19:34.840 the same day BuzzFeed shared that leaked audio from dozens 0:19:34.840 --> 0:19:37.960 of internal TikTok meetings revealed US user data has been 0:19:38.000 --> 0:19:50.520 repeatedly accessed from China. Welcome back to bloomber Technology and 0:19:50.560 --> 0:19:52.480 Emily Chang in San Francisco. I want to dig into 0:19:52.480 --> 0:19:55.280 this and how the fed rate hike is impacting the 0:19:55.320 --> 0:19:59.879 world of VC with Mike Volpi and partner at Index Ventures. 0:20:00.160 --> 0:20:02.520 Might great to have you back with us. Look, it's 0:20:02.560 --> 0:20:04.560 been an incredibly vi all of the week, a lot 0:20:04.600 --> 0:20:08.040 of uncertainty about the future, a lot of people saying 0:20:08.720 --> 0:20:13.560 the R word is inevitable. What do you think. Yeah, 0:20:13.600 --> 0:20:15.680 I'm gonna start by saying that I'm a Lakers fan, 0:20:15.760 --> 0:20:19.600 so I'm relatively different all of the events. Well, we're 0:20:19.640 --> 0:20:25.440 not sorry, but thank you for clarifying, Okay, So, um, yeah, 0:20:25.520 --> 0:20:29.840 it's it's definitely a very tricky time right now. Um, 0:20:30.040 --> 0:20:33.159 and I think that when we look at the portfolio 0:20:33.160 --> 0:20:36.400 of companies that we look after, there are certainly ones 0:20:36.440 --> 0:20:38.399 that are more in the consumer side of business that 0:20:38.440 --> 0:20:43.040 are seeing some softness happened. They're seeing that consumers in 0:20:43.160 --> 0:20:47.080 general have read enough in the news, seeing enough tweets 0:20:47.119 --> 0:20:51.200 about inflation and interest rates and all that, and they 0:20:51.240 --> 0:20:54.080 are moderating their behavior. And you can see that even 0:20:54.080 --> 0:20:56.000 in the statements that the larger companies are making. So 0:20:56.119 --> 0:21:00.080 whether it's Amazon or Target or Walmart saying they don't 0:21:00.080 --> 0:21:03.479 the right inventory or they may not be expanding as 0:21:03.520 --> 0:21:06.680 quickly as they thought, there's clearly something happening out there. 0:21:07.000 --> 0:21:08.800 I think what we don't know is whether it's a 0:21:08.840 --> 0:21:11.280 little Our recession, sort of like a quick one that's 0:21:11.359 --> 0:21:13.359 you know, quick one a quarter or two where it 0:21:13.359 --> 0:21:16.600 fixes things, or a big Our recession. But there's clearly 0:21:16.640 --> 0:21:19.800 something signaling that's going on out there. And I would 0:21:19.800 --> 0:21:22.959 say that on a relative basis, it's happening sooner than 0:21:23.000 --> 0:21:25.560 people expected, because there was sort of this notion that, oh, 0:21:25.640 --> 0:21:29.240 maybe next year. I don't think if something, if the 0:21:29.240 --> 0:21:32.520 economy is to slow down with some significance, it's probably 0:21:32.560 --> 0:21:34.920 not next year, but it's like next quarter or the 0:21:35.000 --> 0:21:38.600 quarter after that. So just how bad then do you 0:21:38.640 --> 0:21:41.400 think is the wreckage is going to be? How many 0:21:41.400 --> 0:21:44.320 more companies will have layoffs? How many people will get 0:21:44.400 --> 0:21:47.199 laid off? How many companies will make it out of this? 0:21:47.280 --> 0:21:51.879 How long does this last? Yeah? In all candor, I 0:21:51.920 --> 0:21:53.800 don't think it's going to be that bad. There will 0:21:53.880 --> 0:21:57.920 be some high profile situations where well known companies are 0:21:58.000 --> 0:22:02.000 letting people go, and that will be painful, undoubtedly, But 0:22:02.400 --> 0:22:06.119 I think that generally speaking, over the last couple of years, 0:22:06.160 --> 0:22:11.440 in and especially private companies have been able to raise 0:22:12.080 --> 0:22:15.399 uh amounts of money that we've never seen in the past, 0:22:15.880 --> 0:22:18.600 and that puts their balance sheets in a pretty good condition. 0:22:19.040 --> 0:22:21.840 Now they may have over hired, so they might trim 0:22:21.880 --> 0:22:24.520 a little bit here or there, but by and large, 0:22:24.560 --> 0:22:28.040 I think the majority of companies have the strongest balance 0:22:28.080 --> 0:22:30.399 sheets that they've had in a while, and in many 0:22:30.440 --> 0:22:33.080 cases have enough in the tank to get through a 0:22:33.119 --> 0:22:36.000 difficult period and to come out at the end of it. 0:22:36.240 --> 0:22:39.520 So I do think that we will surely see uh 0:22:39.720 --> 0:22:42.840 some challenges, and perhaps the ones that we most challenge 0:22:42.880 --> 0:22:45.520 would be the public companies because they obviously have to 0:22:45.520 --> 0:22:48.320 respond to a stock price and the fact that right 0:22:48.359 --> 0:22:52.240 now investors want shorter term earnings or less losses in 0:22:52.280 --> 0:22:55.040 the short term, so I think we'll see more from those, 0:22:55.080 --> 0:22:58.400 But by and large. I don't see the kind of 0:22:58.760 --> 0:23:02.119 quote wreckage that Mamy we saw in two thousand and 0:23:02.160 --> 0:23:05.320 If anything, I would expect this to happen more quickly, 0:23:05.359 --> 0:23:08.800 both the downturn and the upside to happen more quickly 0:23:08.840 --> 0:23:11.600 than in the past. Let's talk about some of these 0:23:11.600 --> 0:23:15.280 more high profile situations, and I want to focus on Cisco. 0:23:15.359 --> 0:23:18.760 Cisco CEO Chuck Robbins was on the show earlier this week. 0:23:18.800 --> 0:23:21.520 I asked for his perspective. Take a listen to what 0:23:21.560 --> 0:23:25.280 he had to say. We are always planning for different scenarios, 0:23:25.280 --> 0:23:27.080 but we've been around long enough and been through enough 0:23:27.119 --> 0:23:29.879 downturns that we have playbooks and we know how to 0:23:30.000 --> 0:23:35.200 we know how to deal with those appropriately. Now, Cisco 0:23:35.359 --> 0:23:37.320 is one of the companies that hit its peak in 0:23:37.400 --> 0:23:41.399 the dot com boom, and the stock Mike has never recovered. 0:23:41.440 --> 0:23:44.240 And what's interesting is you worked there for over a 0:23:44.320 --> 0:23:48.200 decade through the dot com boom and bust, and I'm 0:23:48.280 --> 0:23:52.800 just so curious how you reflect on on the fact 0:23:52.840 --> 0:23:55.639 that Cisco, you know, never has, at least from a 0:23:55.720 --> 0:24:00.960 stock perspective, gone back to what it was. Yeah, I 0:24:01.000 --> 0:24:04.240 think that there's a couple of things though that contribute 0:24:04.240 --> 0:24:05.639 to that, and it's true, by the way, it was 0:24:05.680 --> 0:24:08.960 incredibly unpleasant to go from eighty some dollars a share 0:24:09.040 --> 0:24:11.040 to like nine dollars a share and the span of 0:24:11.080 --> 0:24:13.720 six months when I used to work there. Um, But 0:24:13.880 --> 0:24:15.760 I think that there's a couple of factors that are 0:24:15.760 --> 0:24:19.080 happening in Cisco's case. If you look at where Cisco 0:24:19.160 --> 0:24:21.560 is today, they're a single digit grower, you know, five 0:24:21.560 --> 0:24:24.960 percent grow or three percent grower. And fundamentally what happened 0:24:24.960 --> 0:24:28.000 there is, yes, there was a downturn, but in parallel 0:24:28.040 --> 0:24:31.800 with that, the technologies that they were pervading became much 0:24:31.920 --> 0:24:36.320 more broadly available commoditized, competitive, etcetera, etcetera, and the company 0:24:36.359 --> 0:24:41.040 never really reachieved the kind of growth rate that occurred 0:24:41.240 --> 0:24:44.680 during the pre dot COM's scenario. If you take another 0:24:44.720 --> 0:24:48.840 example like Amazon all Contrair, they did much better afterwards 0:24:48.960 --> 0:24:53.160 because they were able to strategically expand the product lines, 0:24:53.200 --> 0:24:56.520 the capabilities, the offerings, the sectors that the company was in. 0:24:56.960 --> 0:24:58.960