Eric Schmidt booed: Is the mood turning against AI?

00:00:00: You know, I have a lot of friends at Metta and can tell you that the mood is very down.

00:00:06: People don't want to be surveyed.

00:00:08: just because we can manage or monitor our employees doesn't mean we should.

00:00:14: And when you have an unmotivated low morale employee pool they're going underperform.

00:00:20: They are not going...you cannot manage cognitive labor in the same way as manually.

00:00:29: Welcome, welcome to Big Bang Tech and welcome to my co-host Alvin Van Greylen.

00:00:35: Hello Alvin!

00:00:36: Hey

00:00:37: Jens good to see you again.

00:00:38: Good to see.

00:00:39: it's a real big week.

00:00:41: Elan loses against OpenAI.

00:00:43: SpaceX may become the IPO of the year.

00:00:47: No no no off the decade.

00:00:49: Microsoft shows AI agents in the cyber defense And Meta raises new labor question.

00:00:58: Eric Schmidt from my favorite company, Google.

00:01:02: Well he was a former CEO gets booed by students over AI.

00:01:06: that's pretty new.

00:01:08: so this is not just the tech news This is a new power map.

00:01:12: So let's go Elvin huh?

00:01:14: Yes That's happening absolutely.

00:01:16: Let's

00:01:18: do it!

00:01:18: What are your thoughts about Elon versus OpenAI?

00:01:23: He was something like a founding father of openai And then he has just lost a major round in the lawsuit against OpenAI.

00:01:33: So what is going on there?

00:01:35: Yeah, I think this kind of an issue that's been brewing for long time between Sam and Elon.

00:01:45: They've had other ways trying to resolve it.

00:01:47: they haven't And essentially the court ruling was that he waited too long.

00:01:52: So it didn't even evaluate the evidence.

00:01:55: after three weeks of going through this trial, then there is essentially be... The period of what term is expired which I think from a technicality perspective It's fine but from a Satisfaction perspective?

00:02:13: It wasn't very satisfying for any of the audiences right and also It created an issue in terms of what we talked about last time, which is it actually did not give a precedence on whether or not company that was.

00:02:28: A non-profit can become before profit and would actually I think with the bigger.

00:02:33: the bigger issues should have been discussed and shouldn't have been ruled out And then there wasn't.

00:02:39: so i'm disappointed than back yesterday at night when Ilan called for interview and the first question to him was, what do you think about that decision?

00:02:52: And he's like yeah it is.

00:02:55: It felt as if he wasn't too excited or disappointed by it but I think it kind of expected

00:03:01: so... Yeah!

00:03:03: Normally nobody would say that Elon came very late because here's a man who has always been coming up really early with SpaceX.

00:03:16: This is really huge.

00:03:19: The infrastructure stock of the AI age.

00:03:22: and what do you think about that?

00:03:24: What's going on there?

00:03:25: Yeah, I know this will be a very big move for Elon to go out quickly before OpenAI or Anthropic gets out because he'll eat up both attention as well as funds right now creating something that he has to grow into, right?

00:03:50: Because the value of the company and the value at AI side is actually... The economics doesn't justify the valuation they have.

00:03:59: But his painting a very good long-term story.

00:04:03: And you know He's talking about having a gigawatt or a terawatt Of compute being sent in space and giant space data centers.

00:04:11: All these things might create demand for for more Starship launches.

00:04:18: But I think the type of scale he's talking about, which essentially would require hundreds a day of Starship launchers is not something that he's able to do in the time frames at his discussing.

00:04:33: but evaluations are based on him delivering.

00:04:39: So I think he may do a good job in getting the IPO to go out at high price, but i don't know if you can sustain those prices.

00:04:48: And for the price is important it's not any more a rocket company.

00:04:53: It's about Starlink satellites defense connectivity data and geopolitical power.

00:05:02: You see it in Ukraine everywhere without Elan nothing going on there.

00:05:08: Yeah, yeah.

00:05:08: So he's right now more than half or probably like sixty percent of the global launch capacity and He wants to become like probably eighty or ninety percent at a global on capacity Right?

00:05:19: And I think the satellites in communication i have A lot Of faith is going To do that.

00:05:23: well The whole story around AI data centers I Think it's not as real As he makes It out-to-be Both from my technical perspective as Well as for me cost Perspective just doesn't really make sense at least any time in the next five

00:05:39: years.

00:05:40: And I think for this stock exchange, it's very important to be a company for the future of operating geopolitics technology.

00:05:58: and so you would sign it?

00:06:02: You will sign

00:06:04: which?

00:06:05: The stock.

00:06:06: I don't think...I would not buy it.

00:06:08: Um, i-i think..it will pop because he's a really good marketer but i dont know if its gonna stay high Because once you go public then you have to release all of your financials And those financials won't justify the size of this IPO.

00:06:30: We're talking about two trillion dollar valuation on a company that really still has relatively minimal real revenues.

00:06:40: And, you know... On the SpaceX side there is revenues but that revenue it's not necessarily as scalable because its all very dependent upon lot of governmental contracts whereas the AI side I don't think.

00:06:56: Okay, so hands off of this stock from the expert Evan.

00:07:04: Let's switch to Microsoft.

00:07:07: The owner or part-owner of OpenAI and there is something new.

00:07:12: Microsoft was showing where AI agents may first create real enterprise value cybersecurity.

00:07:19: So that's pretty good news for us.

00:07:23: Yeah, I

00:07:25: mean they're proving something that's actually quite interesting because in the past people used to think you have giant models.

00:07:33: Essentially god-models right?

00:07:35: What anthropic and open air creating can be able to have the highest performance on things And what their showing is with this new Mdash project or last-generation models, and it actually outperforms the cybersecurity capabilities of most expensive giant models.

00:07:57: And this I think can change a lot of perspectives in terms.

00:08:05: how should models be trained?

00:08:06: How should they be deployed?

00:08:08: The value of models versus harness.

00:08:11: Harness is about all these models are put together.

00:08:16: That actually, to me is where the breakthroughs will happen versus trying to create a giant supermodel that knows everything.

00:08:25: I think what makes more sense is creating models which connect lots of little specialized models.

00:08:32: It's actually more energy efficient and it's more training-efficient as well as cost efficient.

00:08:38: But this isn't really sexy because if we write about cybersecurity nobody cares about it?

00:08:44: Why does it look like

00:08:46: so?

00:08:46: Well, I mean cybersecurity is something that isn't in the background.

00:08:49: You don't care about it until a break yeah But when it breaks everybody cares because

00:08:54: of course It's like an insurance and life insurance you have to have but you its not sexy at all?

00:08:59: Do Not Like it

00:09:00: Yeah, but but i think cyber security Is only The first use case Of this.

00:09:05: they could be used on incentive In terms of this concept of integrating lots of little models into A general harness too to enhance intelligence is something, as a practice can be applied to any industry.

00:09:19: So this why I actually feel like the current chase for these giant data centers and the giant ten or twenty trillion parameter models Is actually misguided And unnecessary.

00:09:32: so you were astonished about this news?

00:09:34: Or is it for your pretty normal that It will show up in there where we

00:09:38: think?

00:09:39: i was surprised at how quickly and How well it performs?

00:09:43: I'm not astonished, but this actually is a sign to the world that maybe we may not be on the right track of just expecting pure scaling to solve everything.

00:09:53: Which it's good news because it also means that We don't necessarily need to continue to build you know trillions of dollars at data centers.

00:10:04: Yeah,

00:10:05: you're always very skeptical about data centers.

00:10:07: Yes if they are too big and we do not need it And to much energy well

00:10:12: We were these data centers of built To try to solve a problem that way using current techniques for and there may be breakthroughs that come That can significantly improve the efficiency And the methodology of how this training happens, you know The fact that our brain is a general-purpose computer and it runs on less than twenty watts.

00:10:37: It takes millions of watts to do similar things in a giant rack.

00:10:44: There's definitely ways...things we don't understand yet In terms of doing more efficiently Using only brute force its work.

00:10:52: but it cannot continue to use that technique because now we're at the edge of physical limitations.

00:11:01: We have another company, a real big one, Meta.

00:11:04: Meta has for news.

00:11:06: I've read some news since three weeks and then other news came up about reports that employee behavior could be used to train AI agents And the concern is screen activity, mouse movements workflows.

00:11:24: The daily patterns of human work that they track it and then with this information They will have a new digital labor force.

00:11:34: or do you

00:11:38: think about?

00:11:45: I mean using it as a management tool.

00:11:49: But right now, I have a lot of friends at Metta and can tell you that the mood is very down.

00:11:55: people don't want to be surveyed.

00:11:57: People do not feel like they are being trained.

00:12:00: So

00:12:02: for those who are not fake news this works.

00:12:06: when i'm writing or use my computer realize what I'm doing the whole day, and then at least they think okay i can do it by myself or the systems can do that by themselves.

00:12:17: Well yeah...I think its a twofold thing.

00:12:20: one is about labor replacement.

00:12:22: The other things are actually more about labor management efficiency management right?

00:12:28: It used to be.

00:12:29: when you're on an assembly line people could tell ok Are your moving in the same speed?

00:12:33: because it has to move?

00:12:35: so you have to be working And it forces you to be productive.

00:12:39: But when you're at a white collar work in front of the computer, how much time are using moving your mouse?

00:12:45: How many times do I have to use my brain thinking?

00:12:47: Right.

00:12:47: It's hard to differentiate it too.

00:12:49: and if you're being measured on...how many key strokes that you're making?

00:12:52: and measuring how many mouse strokes we were making You now be gauged or measured

00:13:00: just like

00:13:01: an assembly line?

00:13:03: People don't like that because this is different way than benchmarking.

00:13:08: Do you like it?

00:13:10: No, I would not.

00:13:11: Because most people who are working and being paid as white collar workers using the brain is not necessarily equated to how many keystrokes they make.

00:13:23: so one good idea could be worth a trillion dollars but that doesn't require anything.

00:13:30: Yeah, that's all.

00:13:31: That's I think we are facing new questions.

00:13:34: We have never thought about it right?

00:13:36: yeah i mean just because we can Manage and monitor our employees doesn't mean we should.

00:13:42: And then when you have a unmotivated Low morale employee pool they're going to underperform.

00:13:49: They're not going too.

00:13:50: You know you cannot manage cognitive labor the same way you managed manual labor.

00:13:58: So The former CEO of Google, Eric Schmidt.

00:14:02: I think you know him and he was booed during a university commencement speech when he spoke about artificial intelligence.

00:14:11: so it wasn't in the tech conference but with people on the way who love AI like you, for example.

00:14:20: So what was going on there?

00:14:23: Yeah so he spoke at the commencement speech for Arizona State University is actually one of the biggest universities in America and yeah

00:14:31: where's it?

00:14:32: uh in Arizona?

00:14:33: all right you

00:14:35: know one of those kind of desert southern mountain states I guess.

00:14:40: but You know, I think the issue is really more about his approach.

00:14:45: Right?

00:14:46: And he's talking and very vocal about AI replacing laborer.

00:14:53: He wasn't giving people hope or solution.

00:14:57: He was saying you need to go with it.

00:15:02: I think in a Silicon Valley audience, his messages would probably be okay because everybody already accepts that they're going to all lose their jobs.

00:15:10: But when you have young people who are at the beginning of their career and if your taking away there livelihoods not everyone is studying technology so they don't even feel like misunderstood the audience and, uh...the type of messaging that he should be.

00:15:34: He's giving essentially a message that would normally give to Silicon Valley people And I don't think it is right approach.

00:15:43: But its interesting an important too.

00:15:45: in board rooms AI means productivity That needs more money.

00:15:51: on campuses AI can mean fear

00:15:55: And I think one is about fear.

00:16:00: He represents what's creating this problem, right?

00:16:03: Because he was part of Google and Google as one the big companies building these things.

00:16:07: so people see him seeing things that are self-serving.

00:16:10: Right!

00:16:10: He already a billionaire...he's the one who's creating technology to take away their jobs or futures.

00:16:18: Why should be the one that lectures them on how they live life?

00:16:22: He's destroying their life from there perspective and I can appreciate that, understand you know?

00:16:29: That sentiment.

00:16:31: So he should have came in with a different approach And i think he was over-emphasizing AI when he should've been more emphasizing How.

00:16:40: how can the human side of Of the education play a bigger role?

00:16:46: How do you meld into the world where AI now produces everything and controls everything?

00:16:57: But

00:16:58: maybe this industry needs a new PR agency that is explaining it in a better way.

00:17:04: Yeah, to be honest I think right now the AI industry is doing itself very big disservice.

00:17:11: You have Dario saying within twelve or eighteen months all cognitive labor will be replaced by AI And then without a solution It doesn't help.

00:17:20: It scares people and it makes people angry.

00:17:22: And America's, eighty percent of the people are C-AI as negative to society.

00:17:28: That is a very bad thing.

00:17:30: Evan last question What your message for this weekend?

00:17:38: So I am actually working on new paper.

00:17:41: This message is something that important for people understand.

00:17:45: Right now All of this race condition is happening because there's a perception from a few labs in the world that whoever gets to AGI will become a Dekatroner.

00:17:57: Right, they're going to make there essentially going to take over what we just talked about.

00:18:01: the cognitive labor white collar Laborer industry in the world is somewhere around thirty two forty trillion dollars of value and They think if I make a GI.

00:18:09: I will be able to transfer that thirty to forty trillion into my pocket.

00:18:14: And it is a mirage.

00:18:15: It's a myth and it won't happen.

00:18:19: incentivizing and perpetuating this myth because it's making people make bad decisions, is making people manipulate governments to try be that first person or the company.

00:18:32: It also makes these companies manipulate government to create tension between countries when necessary.

00:18:42: we need to see AI as a public good.

00:18:46: something can benefit from, something that the world works on together and then when it comes is shared with everyone.

00:18:54: That's not to use as a way of making money for individuals or companies.

00:19:01: we have to change their perspective because until you do these issues are going to be about young people hating this technology.

00:19:08: We're gonna have companies fighting over it of sharing between countries and duplicating efforts, building unnecessary data centers.

00:19:20: And spending money that should be built in terms of creating social infrastructure we're now creating.

00:19:25: AI infrastructure is going to be wasted when real breakthroughs are created right?

00:19:32: So this I know i'm kind of preaching the choir but AI should not be used as a way to profit.

00:19:46: It should use it for the benefit of humanity, and that's my message too... ...to people in both Germany & US, China & the rest of the world.

00:19:56: In Germany you will be in Berlin very soon.

00:19:59: then you'll have an appointment with the German government, with the Department of Digitalization And this would also be your message.

00:20:09: Yeah, absolutely.

00:20:11: Because once we realize that then we realized... We don't need to race!

00:20:14: Once we realize it and say let's work together take what we have deployed in a smooth responsible way so the workers are not losing their job very quickly.

00:20:25: They're not creating a dislocation or disruption for economy and society because they will be bad for everyone.

00:20:37: We will see you soon here in Europe.

00:20:39: I look forward

00:20:40: to seeing you in Berlin, yes absolutely!

00:20:43: But the show is over and we'll see our audience again in two weeks.

00:20:49: so if there's any questions don't hesitate to send us an e-mail at debua.gdb.de And especially for nice emails or compliments from me and critics too Alvin

00:21:03: Very good.

00:21:04: And next time are we going to do this in Berlin live?

00:21:06: Is that...

00:21:07: Yeah, that could be!

00:21:08: Next time let's do it live.

00:21:10: and next was the Brandenburger Tour.

00:21:15: Thank you very much.

00:21:16: everybody

00:21:17: Have a

00:21:18: great week Bye.