We're sunsetting PodQuest on 2025-07-28. Thank you for your support!
Export Podcast Subscriptions
cover of episode (Preview) xAI Buys X, The Bitter Lesson for Everyone, Everywhere, Google’s Only Hope in AI

(Preview) xAI Buys X, The Bitter Lesson for Everyone, Everywhere, Google’s Only Hope in AI

2025/3/31
logo of podcast Sharp Tech with Ben Thompson

Sharp Tech with Ben Thompson

AI Deep Dive AI Chapters Transcript
People
B
Ben
无相关信息。
主持人
专注于电动车和能源领域的播客主持人和内容创作者。
Topics
主持人: 我对马斯克将xAI与X合并的举动感到困惑,因为xAI似乎一直以来都在使用X的数据。我不确定这是否会对xAI带来实质性的改变,或者仅仅是形式上的区别。 我很好奇,随着AI公司对实时数据的依赖性越来越高,X的数据价值是否也越来越高? 最后,我想探讨一下xAI的成功案例,这是否意味着只要资金充足,就能在几年内开发出具有竞争力的AI模型? Ben: 马斯克的交易是否合法合规,我们无法得知,因为这并非公开市场交易。 我认为xAI收购X是有意义的,因为X的数据非常有价值,收购后xAI就不需要再为数据付费了。此外,此举也体现了X的长期潜在价值,即其实时数据。 推特在被马斯克收购之前,盈利能力很差,甚至在其运营的十年里,盈利能力都很差。马斯克收购推特的最大缺陷在于支付的价格过高,但这并不影响推特的数据价值。推特的价值在于其影响力远超其盈利能力,它应该专注于数据变现。 未来AI模型的差异化在于对实时数据的访问,X的实时数据对于AI公司来说非常有价值。AI公司收购推特是有意义的,因为推特拥有非常有价值的实时数据,这些数据可以被用于多种用途,例如为华尔街交易员提供数据。 推特上有大量有价值的数据,而此次收购是利用这些数据的一种方式。最初的推特投资者现在将持有XAI的股份,这比持有X的股份更好。 X上的广告质量很差,这反映了X平台的现状。为了避免看到垃圾广告,我购买了X的付费服务。关于推特将要倒闭的传言是错误的,但推特的稳定性和可靠性不如以前。 我现在更经常使用OpenAI的模型,因为其应用程序更好用,而Grok的网页应用用户体验不好。我在台湾使用互联网时经常遇到验证码问题,这可能是因为台湾曾经是许多网络诈骗的中心。虽然Grok不如OpenAI的模型好用,但我仍然在使用它。 谷歌应该更注重产品发布,而不是只专注于技术研发。

Deep Dive

Chapters
This chapter analyzes Elon Musk's acquisition of X by XAI, evaluating its impact and implications. The discussion explores whether the merger is a significant change or simply a reorganization, considering the previous data-sharing arrangement between the two entities. The value and monetization potential of real-time data are also examined.
  • XAI acquired X in an all-stock transaction valuing XAI at $80 billion and X at $33 billion.
  • XAI had previously been using X's data.
  • The acquisition eliminates XAI's data payments to X.
  • The real-time data of X is considered valuable for AI development.

Shownotes Transcript

Translations:
中文

Hello, and welcome to a free preview of Sharp Tech. Elon Musk tweeted, XAI has acquired X in an all-stock transaction. The combination values XAI at $80 billion and X at $33 billion, $45 billion less $12 billion debt.

Since its founding two years ago, XAI has rapidly become one of the leading AI labs in the world, building models and data centers at unprecedented speed and scale. X is the digital town square where more than 600 million active users go to find the real-time source of ground truth and

and in the last two years has been transformed into one of the most efficient companies in the world, positioning it to deliver scalable future growth.

So Ben, this caused a bit of a hubbub on Friday afternoon. I'm now trying to get a sense for whether that news actually matters because it was my understanding that XAI had been working with X data all along. So does this news actually change anything for XAI or is the new arrangement effectively a distinction without any real difference?

Yeah. I mean, some people were trying to like raise like, oh, this is, you know, so wrong. And, you know, what are these numbers? It's like they're all private transactions. I was messaging a friend like in case you missed it. Excellent private a couple of years ago. So does Elon Musk have fiduciary duty to the investors in this deal? Yes, he does. Does he have their permission to do that? I would assume so.

Do we know or have any insight? No, it's not a public market company. So just putting that out there, number one. Number two, this is why I didn't write about or mention the reports about like X being profitable or not or what the valuation was because a huge amount of their revenue was being paid from XAI for access to their data. So which again is the right hand paying the left, right?

which is fine. The X data is very valuable, but that does speak to, there is value in acquiring X in that they're not now paying for X data, you know, sort of again, leaving aside the numbers. Is this a deal that makes sense? Yes, it is. And I would go one step further, which is to say the structure of this deal probably does speak to the,

long-term potential value of X, which has always been the

The real time data. And for ages. It's hard to keep track of what. The various things that I prescribe for Twitter. Over the years. Trying to make this into. 12 years of recommendations that were not heeded. And just bailing on it multiple times. And not writing about it. At first because it was like look this is a failed business. There's no point writing about the business. Then once Elon takes over. It's like so radioactive. And also now it's private. So you don't even have numbers to look at. I mean people forget.

Twitter, I think, had one profitable quarter in terms like ever, like in its entire run. Maybe it was two or three. I might have been like three. But I remember because we got that email on the last episode asking for our group chat segment. And I had to go back and look at some of our discussions in like November 2022. And there was a lot of Twitter discussion back then. And I remember being astounded by.

looking back at how unprofitable they had been in the 10 years prior to Elon. So whether it was one quarter, two quarters, three quarters, it wasn't good. It's the clear takeaway. It was not good. So yeah, it is an important thing to keep in mind. The biggest flaw in the entire Musk administration

acquisition whole thing was how much he paid because it was, you know, but the reality you could argue, he's already gotten his money's worth. As we mentioned, you know, a few months ago, because the influence of Twitter, it

has always far exceeded its monetization potential. And this bit about real-time access being valuable to me has always been interesting. And why, remember, I think they acquired Mopub back in the day. It's interesting because they then spun Mopub off to Apple Oven, and Apple Oven has done amazing things with it from a business perspective. But my view was, look, this is a data...

Like this whole company should be a data play. They should be selling access at much higher rates to the data that they have. And there's way more monetization potential in this regard. And this is maybe like a backwards way to that, where what is going to differentiate models from,

In the long run. Well, one thing that is valuable and to me, one of the biggest grok use cases is its access to at least the sense of access to much more real time data. And can it surface that consistently and usefully from X point?

We'll see how that plays out in the long run. But from a very high level, again, setting aside all the particulars, obviously setting aside all the politics, does this make sense that a –

Would be AI company acquires Twitter again, leaving aside it's the same person on both sides of the equation. Does this make sense? Yes, I think it does make sense. I think it makes a great deal of sense. And it makes sense in this order that it's the the AI company acquiring Twitter as a really valuable input to being a potential winning product in the market.

Well, in terms of real-time data for X, how much more valuable is it now that these AI companies are reliant on real-time data or get much better with real-time data than it was in, say, 2016, 2017? It seems like they've waited around long enough so that there's now more of a market for...

for what they can provide. Yeah, this is just another way to do what I proposed ages ago. I thought they should have been selling like API feeds and have a much more developed API and access to data for traders, for Wall Street. Maybe this is going to end up being an even larger opportunity, so it will work out in the long run. But again, we're talking at a very high level. If you just zoom out

Is there valuable data on Twitter slash X? Yes. Does that make it, is this one way to manifest it? Certainly. So we'll, we'll see, we'll see how it plays out, but at least from a big picture perspective, again, setting aside the numbers, does this make sense? Yes, I think it does.

Indeed. Well, and the initial Twitter investors look like they're going to be taken care of here. Another funny thing. Well, they're taken care of to the extent they're now all in on XAI. Which I would feel better about than being all in on X, frankly. As much as X is apparently doing better than ever these days, I don't know how much I buy that.

Yeah, we'll see. I mean, I don't know. The quality of ads I get are certainly not very high. Don't speak to this being a very thriving platform. So, and I'd like to say I have a fairly, you know, I shouldn't be getting the level of trash ads that I do. But, you know, we'll see.

Well, there was about a year there where I was constantly served Cheech and Chong marijuana gummy ads on X. Oh, I just get these atrocious crypto ads. Oh, yeah. There is a lot of atrocious crypto stuff. The last week, there's been all these Apple token things that are just – it's ridiculous. It's disgusting. I do – I did on my burner account apply –

I do have like the highest level one now just to get rid of ads. I've realized like, look, it's pretty expensive, but I use this a lot. So it's worth it for me not to see it. And then every time I go to one of my other, I'm really being punished for having the multiple, the multiple X accounts. It's rough. It's pretty rough.

The algorithm. Your algorithms are so confused by your Twitter usage. I will say, going back to fall of 2022, one story that I had completely forgotten about before I had to search through our archives last week was there was that weekend where...

where X had fired everybody and then it looked like it went dark for a little while and there were all these Twitter eulogies about how the service was just not going to survive. No, it never went dark. This was a completely spun-up, fanciful, it's going to go dark news cycle that, you know, like, again... It's pretty funny to relive a couple years later. Yeah, I mean, it should give everyone pause about, you know, this is just...

Yeah, it never happened. And it's funny because it's hard – like I think Twitter has actually been quite – it goes through periods of quite a bit of instability. I think it's been quite unstable over the last week or so. And is the infrastructure as stable and as reliable as it was back then? I still don't think it is. But –

But it's also the case they did cut costs by like 80%. So I mean, it's mostly still functioning. So I think everyone kind of looks kind of dumb about the whole thing for sure. Well, on the XAI side, are you still using Grok more than any LLM? And if so, why?

I have to say I've kind of gone back to open AI a fair bit. And it's honestly because like, so now I feel validated in one of my initial takes. The apps are better. They're just better, particularly on Mac OS. Like the Mac OS, like,

ChatGPT app is great. It's always there. It's reliable. The Grok, it's still only a web app. There is no actual app. And that web app seems to – number one, it's just harder to use a web app in general even if you have like a dedicated site browser. It logs me out constantly. I always have to re-verify and I don't know I'm logged out until I type a question and it doesn't answer and I have to refresh the page and I have to click a cop shop. That's a deal breaker right there. Yeah. It might be worse because I am abroad and –

this is a, Taiwan used to be the home of a lot of, uh, you know, scammers and stuff as far as like, like, especially back in the nineties and two thousands and farm type stuff. Yeah. Bot farm sort of stuff. And so I've, it's, it's kind of painful actually using the internet here. I, I deal with more captures than you can believe with everything all the time. And it's still the case that a lot of things like, uh,

you know, like I go to Home Depot or I go to my local furniture store and I'm always blocked. I have to go into VPN to get through it. Like there's just large swaths, particularly of like older stuff because no one ever updated with better security practices. Like, look, just block this whole list of IPs. So that might be part of the reason, but it doesn't matter to me. I'm just trying to get work done. And 4.0 has been updated. I think it is probably better. But honestly, for a lot of stuff,

It gets the job done, and it's just easier to use. So I do still use Grok, but less than I was where I was all in for a few weeks anyway. Fair enough. Well, we're going to continue to monitor your LLM favorites as we go here. One big picture question I have. I am not, by the way. Everyone says Gemini 2.5 is incredible. They did just put it in the Gemini app. I've just...

In case it comes up on this podcast, I have not dived into it yet. I do plan to do that. Okay. I did appreciate – I scolded Google on Wednesday where ChatGPT is going viral even though Google kind of had some aspect of this capability already. And meanwhile, they had 2.5 that came out the day before. That was only in the API and only in AI Studio. It wasn't in their apps. I'm like, guys, you've got to ship a product. Like you –

Your infrastructure, it was so classic Google in many respects. They got it into the product already. Apparently, it was put in over the weekend. So I do look forward to checking that out. But I just want to give that caveat. I see you, Gemini. I see you, 2.5. No feedback yet. There we go. Well, perhaps we'll revisit it after vacation. I had one other question on XAI.

What is the lesson from XAI? Because Elon's tweet mentioned spinning this up in two years, which is frankly pretty unbelievable. Is it that anyone who has enough money can just buy GPUs and hire the best engineering talent and eventually release competitive models within a couple of years? Or was this blueprint something that Elon was uniquely situated to execute?

All right, and that is the end of the free preview. If you'd like to hear more from Ben and I, there are links to subscribe in the show notes, or you can also go to sharptech.fm. Either option will get you access to a personalized feed that has all the shows we do every week, plus lots more great content from Stratechery and the Stratechery Plus bundle. Check it out, and if you've got feedback, please email us at email at sharptech.fm.