Musk’s OpenAI Bid, Sam Altman’s Roadmap, Apple + Alibaba

0:00Elon Musk bids$97 billion for OpenAI, but is his bid real? Sam Altman reveals openAI's roadmap as the future of generative AI looks a bit clearer and Apple finally has an AI partner in China. That's coming up on a Big Technology Podcast Friday edition right after this.

0:48Welcome to Big Technology Podcast Friday edition, where we break down the news in our traditional cool-headed and nuanced format. We have a massive week of news to break down for you. We have this Elon Musk bid, or is it a real bid? We'll get into that for OpenAI. Also, we have so much more clarity on the roadmap of OpenAI and its fellow generative AI companies following this week. So we're going to talk about what's coming next from them. We also have an Apple intelligence partner in China. It's Alibaba, of course. After a couple of false starts, Apple seems on the way to be releasing Apple intelligence in its second biggest market.

1:24And there's going to be plenty more to discuss. Joining us as always on Fridays is Ranjan Roy of Margins. Ranjan, great to see you again. Good to see you. I think I'm going to put in a bid for OpenAI as well. Let's make it competitive, Elon. Why don't we do it? Let's gather our money together and maybe we can put the$200 a month we've been spending for OpenAI's advanced features towards it. So, all right, this is our official bid,$400 for OpenAI. $400 going once, going twice. I bet. It's a nonprofit, right? So that should at least register. I think so. So the news this week is kind of out of control, as it always is with AI.

2:01and I was starting out for a fairly calm normal week and then Monday news breaks Elon Musk is bidding 97 billion dollars for OpenAI so immediately everybody covers this as if this is real because of course Elon Musk has done this before he took over Twitter with a 44 billion dollar offer so okay so what's another you know another few billion for OpenAI I'll just run through this quickly this is from Reuters a consortium led by Elon Musk said on Monday it offered$97.4 billion to buy the nonprofit that controls OpenAI. Altman responds, he says, no, thank you, but we'll buy Twitter for$9.74 billion if you want.

2:40That is the opening salvo. Then OpenAI's lawyers, I'm just going to fast forward through the story. They say, listen, Elon Musk is suing to prevent us from turning into a for-profit. So how is he now trying to buy us for$97 billion and do the same thing that he's telling us we can't do in court. And Musk responds saying, and this is from the New York Times, he will drop the open AI bid if the company preserves its nonprofit mission. Basically saying, all right, you got me. I'll drop the bid if you stay a nonprofit and we all go home happy. So Ranjan, I'm curious how you view the moves of the week and whether you think Elon Musk, even if he's not successful in the takeover bit is going to get something out of making this offer.

3:24It's not real. It's not real. I think that's a pretty safe bet in this, though, again, maybe something that starts as a troll like this that starts as an investment troll slash legal troll could turn into something real. It's happened before. It has happened before. That's fair. But I think the more I was reading about this, this really did present itself as a pretty smart troll. Because again, you put in the bid, you essentially force OpenAI and Sam Altman's hand. And even there's a quote from Ari Emanuel around how the double dealing of Sam Altman and between the nonprofit and the for-profit arm.

4:08I mean, it's something we've talked about for now years. It's something that never has quite made sense. Somehow they keep being able to just skate by this as they raise billions and billions more. So Elon pushing this issue and actually bringing it more to the forefront by making this somewhat ludicrous bid is a pretty interesting development in this. Here, let's actually read directly from Ari Emanuel's comments at this Freakonomics live taping that happened yesterday. And just to full disclosure, Endeavor, Ari Emanuel's company emailed this to me today, copying Musk's lawyer. So they clearly want it out there.

4:49He says, Emanuel says, we're not proposing to buy the charity OpenAI. We are proposing to buy the charity's majority share of Sam and Microsoft's for-profit OpenAI enterprise. So the charity would get 97.4 billion for that. So to me, this is interesting because I wonder, If they're offering$97 billion and Altman's group wants to pay$30 billion to the nonprofit to be able to move to this for-profit structure, whether this fully prevents this spinoff from ever taking place. Because if they have to match that$97 billion, I mean, what are we talking about in terms of funding? Well, that's why it's a savvy move, because why should Sam Altman be able to price the nonprofit arm of OpenAI when clearly there's a vested interest in putting that as low as possible so you can accrue more value to the for-profit side?

5:39And again, it's always been incredibly unclear what code sits in the nonprofit if it was all initially generated within the nonprofit. What assets, what customer lists, what revenue? It's never been clear what sits where. So to just be able to say, OK, this is worth$30 billion. That sounds good. Let's spin it out and let's leave the other. What is it? What did Masa give it? $260, I think? $260 billion? Leave the other$230 billion for the for-profit in my own personal wealth. I think it doesn't make any sense. So I'm actually happy that someone is calling it out. Someone's pushing their hand at this moment on it.

6:22Where this goes, does this just fade away a little later? I mean, I think we might be forgetting about this in a week or two. But who I think will not forget about it is every investor who has to, in the next round or the next pitches, because it just proposes a seed of doubt to the overall structure, reminds everyone things are a lot more complicated than a simple, here's a revenue projection, here's a cap table. And I think that alone actually does cost Sam a lot of problems. Look, I don't think we're going to be forgetting about this. I think we are all going to come to the realization, and I wrote this in Big Technology today in the Substack newsletter, that this is not a real bid.

7:08But it is a very impactful move. And one thing we should note is all of this stuff, all of it, is playing out in the docket in the Northern District of California where Elon Musk, in the same case where Elon Musk is suing to stop the for-profit conversion. I don't think it's a foregone conclusion that OpenAI is going to be able to pull this off. And then you start to ask some existential questions about the company, don't you? Because we know they have to raise billions more dollars to continue with their mission. We know they probably have to raise billions more dollars to keep the lights on.

7:48And it is very hard to do that with this nonprofit hanging over you. So in the extreme case, and I'm not going there yet, but this certainly has made the probability higher that it could be lights out for OpenAI if they're unable to continue to fundraise. Do you think Masa cares about any of this, though? Do you think Masa, in terms of when he's deciding to write the next gigantic check or actually push through? Because I don't know if it was finalized, the gigantic fundraise with SoftBank. I mean, it was reported, but this still is more of a legal detail versus, I think, an existential question.

8:31For any slightly more rational investor, excluding SoftBank here, I think it does provide some more doubt, some more complexity, complication. But I think for the SoftBanks of the world or that kind of money, I don't think it's going to change significantly anything. I mean, I'd have to imagine that eventually, like this, you run out of room. Maybe not this year, maybe not next year, but you can't persist in this structure. We saw it happen when Sam was fired. We're seeing it happen now as business partners are trying to be compensated. And if Musk forces them to maintain that structure, it's going to be a problem.

9:14I mean, it clearly rattled Sam. Did you see his comments about Elon saying he operates everything from a position of insecurity? And I mean, he came out swinging on this one. So it clearly touched a nerve. I did. And I think it's important for us to talk about that because, you know, again, as I was writing this story, we are looking at an open AI that just in the past few months has regained its footing since the Altman firing in 2022. And that might be an extreme thing to say because we're in February 2025. But you have to remember, the company was rattled all through Christmas 2022. Comes back 2023, and it's the parade of executives leaving.

9:54Mira Mirati out. Ilya Sutskever out. A researcher every other day. And now OpenAI, finally, okay, it has some momentum. And then all of a sudden, boom, deep seek happens. Okay, now we have a technological battle. We went through shipmas. 12 days of shipmas we got. Yeah, we had shipments, right? So that seemed like a moment of stability, even if we've made fun of the Santa voice chat GPT. Like, okay, they are shipping. They started shipping the things they were talking about. You and I last week talked about how chat GPT finally hit 300 million users after stagnating for about a year. So they're finally hitting their footing and then Musk does this.

10:31This is like a, if you're a CEO, this is a pretty complicating thing. We're like really putting a lot of efforts to balance all these stakeholders, balance Microsoft, figure out how to compensate your employees what does the ceo do how do you make this very novel legal transition from non-profit to for-profit what does that look like and then just musk comes in and he's like leroy jenkins and you know and and what happens right it's like sam is completely thrown off he gets to the paris ai summit we've talked about before how he's the savviest of all communicators when it comes to these ai ceos totally thrown off his footing completely thrown off his footing.

11:09He's talking about how Elon Musk is an unhappy person. And he opens himself up to a back and forth with Musk and his associates that he's not going to win and ultimately makes him look small. And then he got slapped by Ari Emanuel. It was pretty good. So he's like, Elon's unhappy. And Ari said, Sam said that Elon was an unhappy person. Yeah, thank you for your analysis. Elon's unhappy because you're a phony trying to get away with cheating the charity in its original mission. Elon is not. So look, you know, we all from the peanut gallery, we're like, ha ha ha, look at them. They're, you know, going back and forth and talking about how bad the other one is.

11:43But like, ultimately, it's important for OpenAI's focus to be on the technology. And Sam is at his best when he's talking about the technology, tweeting out chat GPT previews and tweeting out Dolly previews and Sora previews and getting people excited and being at the Apple intelligence release event, finding ways to partner with them. The more he has to deal with this, the worse it is for OpenAI. And to me, that's the secondary impact of Musk doing this is he's just completely thrown them off the rhythm. They've worked so hard to restore. Yeah, no, if you think about it, there's essentially three major challenges at any given time Sam's dealing with.

12:19There's the product and technology side, which they've been kicking some ass on recently. I mean, deep research operator, again, very mesmerizing product, not great, But still, like having a huge impact in opening people's imagination. Then there's the financial side, the actual like keeping the lights on, training the next models, which it seems now with the SoftBank relationship, they really have found their footing. And then the third is exactly this, the underlying legal and structural complexity that the deal with Microsoft, the for-profit versus non-profit. And it's something that has always been there, but that has just been quiet for a long time.

13:05So yeah, the more that becomes the primary issue, then all of your progress on the financial and the product slash technology becomes a lot more difficult. Yeah, and for the record, I did email Musk's lawyer, Mark Toberoff, and I've been lighting up the guy's inbox the whole week, trying to get him on the phone. And I was like, hey, Mark, let's both be honest here. Elon Musk doesn't really want to buy OpenAI. And four minutes later, he writes back, that's completely wrong. So they're sticking with their story. That's completely wrong. Concise, to the point. Thanks, Mark. By the way, for those who are joining us and are new listeners to the show, maybe you found the show last week as we record a video on Spotify.

13:54Just to let you know, Ranjan and I show up every Friday to break down the week's news. So we do this Fridays. It's the Friday show. And then on Wednesdays, I do a flagship interview with somebody in the tech world or somebody with commentary on the tech world. So you might have heard the Evan Ratliff interview this week where he turned his voice into an AI clone and sent it out to therapy and speaking with scammers. And over the next couple of weeks, we're going to speak with people with some political views. Chris Hayes is coming on, the MSNBC host, and Sagar and Jetty of Breaking Points to talk about tech and also how it's impacting politics.

14:26So if you're new here, we appreciate you and we want to say thank you for being here. All right, let's talk about the technology, shall we? So this wasn't just a dispute week. This was also a technology week where Altman decides this is time for me to share the roadmap for GPT 4.5 and GPT 5. So a couple of really interesting updates here. And this is all about the next couple models that we're going to see from OpenAI. So on Twitter, Altman says, we'll next ship GPT 4.5, the model we called Orion internally, as our last non-chain of thought model. After that, the top goal for us is to unify O-series models and GPT-series models by creating systems that can use all our tools, know when to think for a long time or not, and generally be useful for a wide range of tasks.

15:19In both ChatGPT and our API, we'll release GPT-5 as a system that integrates a lot of our technology, including O3, which is its latest reasoning model. We will no longer ship O3 as a standalone model. So a couple of really interesting news nuggets here in a very concise statement. So I would say, first of all, we know that GPT-5 is coming. Somebody asked Altman what the timing is going to be for both these models. He said GPT-4.5, which is going to be their next LLM coming in weeks and GPT-5 coming in months. Of course, how many months? We don't know, but at least we have timing there. The bigger news from my perspective is that OpenAI is now going to sort of determine your intent and basically free you from the model picker if you're in chat GPT.

16:07So you can decide whether, you know, right now you're deciding, I want reasoning or I want something normal. And the truth is we as the user shouldn't have to determine that. Like that should be something that the technology does on its own. And what Altman is saying is, yep, that is going to happen. So this, I think, this is a tough one. Having been in the weeds building with generative AI, I actually think this is a bad strategic move. The idea of trying to unify models and exactly what you said, determine intent and understand and pick the right model is a very difficult thing. And it only adds more complexity and more room for not building a great product.

16:47It makes things easier and seemingly, you know, like maybe it will, especially at the consumer level, add a little bit of ease. But I think more and more, the more people actually want genuinely accurate results or are incorporating this into genuinely impactful things where they need accuracy, having different models that are specialized and starting at that point and actually building and going back to our eternal, it's more about the product than the model. putting the right models with the right products and interfaces actually is a better way to approach this versus the idea that magical GPT-5 will be so comprehensive that every query you have about every issue or anything you want to build, it'll just know what to do.

17:35I don't think that's a good way to approach this. Okay. So that is maybe if you're coming at it as a company or a nonprofit, we'll see that the main business is the API. But we've talked about it. For ChatGPT, the main business is going to be... Sorry, for ChatGPT, look, I already called it ChatGPT. For OpenAI, the main business is going to be ChatGPT. And so if the main business is going to be ChatGPT, this is actually good. It seems like it's a more consumer-friendly way to use generative AI. No, I'm going to push back on that. You could also just not name them terrible names like 03, 03 mini and call them, have a button that says for more complex tasks, click this.

18:20For simpler informational tasks, click this. Like it's not, that's a UI challenge. That's not the future of humanity GPT-5 AGI challenge. And I think to me, it's more from a product marketing perspective, we know they need to do something big with GPT-5. It's been talked about so long that they're obviously trying to turn it into this much bigger thing. But I don't think even from the consumer side, that's necessarily the right approach. And then again, what will happen on this is the end results if it's calling the wrong model for that specific task if it's understanding user intent is the most difficult thing in the world so trying to solve that layer of the overall process i think is not it will make for a worse product and just give it a simple name or have a nice button rather than forcing me to click on a drop down menu and try to decipher for 04 versus 03 versus GPT-4, like that kind of stuff.

19:28I think there's a UI solution to this. And Kevin Weil's a pretty damn good product designer and product creator, so he should be able to figure this one out. You know who does this well? DeepSeek. DeepSeek. Not to be that guy, but you could just click reasoning on DeepSeek, otherwise you get the traditional model. The Chinese company got the English words right in the UI button. Come on. So there's another bit of news that's been buried kind of in this new reshuffling. And that is that OpenAI did have a O3 reasoning model, right, which is going to be their latest reasoning model. They've already released O3 Mini.

20:10And now that was supposed to come out. So this is from TechCrunch. The company postpones that model in favor of this unified release. So the company originally said in December that it aimed to release O3 sometime early in the year, which would kind of be now. And just a few weeks ago, Kevin Wheel, the chief product officer of OpenAI, said that O3 was on track for a February or March launch. Again, now. Now that is released. Now that is pushed. We don't know when that's coming. That's going to come in GPT-5. And by the way, GPT-4, which we're supposed to see in weeks. This is, again, from the TechCrunch article.

20:47So that's the Project Orion that we've been talking about. I think that was probably going to be called GPT-5, except it hasn't lived up to expectations. So GPT-5 or Orion is said to have suffered a number of performance-related challenges and technical setbacks. Bloomberg, the information in the Wall Street Journal, have independently reported that Orion has shown less of an improvement over its predecessor GPT-40 than GPT-4 did over GPT-3. So let's just like very quickly parse this bit of nuance out here. Does this mean like OpenAI basically has kind of hit a wall? Or is this like pretty concerning that we have both an underperforming model and a reasoning release that's been pushed back, especially after DeepSeek?

21:34I mean, what do you read into all this? Yeah, but I think this goes back to that distinction between model and product that on one hand, deep research is probably the most impressive product that has been released in any of the generative AI space, at least in the last few months, which is a lifetime in the space. But then the models themselves are not scaling at the step change level that they were before between GPT-3 and 4, obviously, where anyone who just touched those things saw how much more impressive 4 was than 3. So I think that actually shows that exactly they could be having that kind of trouble in the background.

22:19It didn't affect the product as of today. So to kind of be betting the company or putting all of your focus on the next big model, I think is the wrong one. Okay, I want to give you one more data point that goes to my side on the pick your model or the dynamic model switching. The other data point is that Anthropic is also going to switch to this. So Anthropic has also not released a reasoning model. This is from the information Anthropic strikes back. Anthropic is taking a slightly different approach to reasoning. It has developed a hybrid AI model that includes reasoning capabilities, which basically means the model uses more computational resources to calculate answers to hard questions.

23:01But the model can also handle simpler tasks quickly without the extra work by acting like a traditional large language model. So the company plans to release it in the coming weeks. Oh, we're going to get a new model from Anthropic in the coming weeks. the interesting thing is that developers can adjust the amount of reasoning versus traditional llm on a sliding scale you slide it to zero and then you can use it as a run-of-the-mill non-reasoning ai i guess you slide it to 10 and it's a reasoning model actually you know maybe this is kind of a hybrid in terms of what what we both want this is what i'm saying give me a sliding scale that's actually a beautiful ui trick right there how hard is this problem you're trying to solve and honestly if it means I don't hit my clod rate limit after just like 30 minutes of use, I will be the happiest person ever.

23:49If you actually let me, you know, kind of help guide the type of work I'm doing. But yeah, so I'm going to put this more in the, this is the UI approach, the user interface approach rather than the model is smart enough to understand intent. Give me a sliding scale. Okay, so as long as we're talking about product and positioning of all these AI companies, what did you think of the AI Super Bowl ads? Okay, I'm going to take a moment here. And if we had cameras here, I'd be looking over to camera three right now. Last week, we said AI has a branding problem. I think that extends both a lot of the figureheads of the industry to the general public are not the most palatable folks.

24:38I think from an expectations management standpoint, the industry keeps promising more and more while not delivering exactly that today. And the ads I thought were terrible, like absolutely terrible. And just to go through some of the ads, and I'm guessing most of our listeners probably watched a number of them, and trying to describe them at least very quickly. OpenAI's ad, I'm assuming you saw? Yes, I did. It's a series of dots going through kind of the history of creation over time from spears and stones and wheels to fire to eventually rockets and this very cool pixelated dot format. that when you're trying to reach just the everyday consumer and get them excited about ChatGPT, it did not make any sense to me.

25:33What does ChatGPT do? How can I use it in my life? It made no sense. It felt like it was built by an agent and created by an agency geared towards the most power users of power users already of ChatGPT. What did you think? I hated it too. Yeah. I mean, it just said chat GPT at the very end. Yeah. I mean, this is where then Google the Gemini ad, very heartwarming. It's about a dad, like, and it's kind of like a play on speaking in very corporate language, preparing for a job interview, talking to Gemini, very, very heartwarming, emotional journey, him and his daughter, the most important job he's ever had.

26:18But he only talks to Gemini at the beginning and asks, or Gemini asks, what's the most important job you've ever done? And at the end, Gemini tells him, great answer. So what Gemini has to do with any of this is not clear. I think Meta, we'll get into their ad, was actually pretty good, I thought, out of all of them. But But I actually had the chance to watch the Super Bowl with my parents and my 76-year-old mom after watching a lot of these ads goes, so AI, what should I do? Like that was literally the question. What should I do? Like I got to do something. And none of the ads geared anyone more towards anything that they can do, how it's going to bring value to their life.

27:09And I get Super Bowl ads. you shell out, what was it, like$5 or$10 million for that 30-second spot? $8 million, I think. Yeah,$8 million, you hire an agency, you're paying them a lot of money. You're not going to put forth some kind of like feature-oriented ad, but I still think something a bit more explanatory or creative in terms of here's how AI can help your life would have gone so much further rather than just the most kind of like agency-ish ads imaginable. that's my that's my rant no i fully agree with you look at us it's a moment of agreement here on the show yeah we'll take this one we'll take it i i will say that these ads that try to do this heartwarming moment showing people like bonding with ai it's very interesting to watch this very real thing that's happening in the ai world where people are really forming insane relationships with their AI.

28:08And extremely personal, like weirdly personal, like we, again, we had the replica CEO talking about how people are marrying their replicas. I spoke about this with Ratliff. People are asking their chat GPT instance for love advice, all these things. You send that to an ad agency and like the brief is show how people are falling in love with AIs and becoming best friends with them. And the ad agency spits out this very sanitized thing because it is kind of creepy for a company to tell you this will be akin to your friend right um and that's sort of where i think these heartwarming ads fall short is they want to show it but if they really showed what was happening it would be pretty freaky and creepy no but but but i disagree in in terms of uh like even the gemini ad between the father and his daughter show them using technology together like any of these things show someone's life being improved there was also the agent force ad with Matthew McConaughey of course which again he's running through an airport saying like if my AI agent had given me better advice as though the idea that Salesforce is going to deliver some consumer facing AI agent to you they apparently their Heathrow airport is a customer and somehow Salesforce's agent force will help Heathrow deliver a better customer experience which may Maybe it can.

29:31But again, Apple intelligence can't even find our flight information with Siri. So like the idea that our but meanwhile, I am planning an entire March vacation to go skiing using a combination of chat GPT of perplexity. That's where I'm doing all my searching. So I'm using it because I know how to show that to people somehow in a creative way, rather than these vague promises of agents will help Matthew McConaughey run through Heathrow in a more efficient way. Yeah, I agree with you. There's so many interesting consumer applications that you could just show them, right? If the technology, we both agree the technology is powerful, just show it.