Tech Brew Ride Home - (BNS) Big Technology Crossover OpenAI Teases GPT-5, Musk Raises $6B for xAI

Episode Date: June 1, 2024

OpenAI Teases GPT-5, Musk Raises $6B for xAI, Loneliness in Remote Work Era. Learn more about your ad choices. Visit megaphone.fm/adchoices

On April 4th, 2023, around 2 in the morning, a man was found stabbed multiple times on a sidewalk in downtown San Francisco. Hey, who did this to you? What happened next turned the story into a political firestorm. Reports have identified the victim as Bob Lee, the founder of Cash App. From Bloomberg Podcasts, this is Foundering, the Killing of Bob Lee, beginning April 16. Open AI says it started training its next frontier model. Saty and Adela calls Sam Altman to talk about his Apple partnership. Apple's big announcements, meanwhile, look clearer ahead of WWDC.

Elon Musk raised $6 billion for XAI, and the remote work era is spiking loneliness. All that and more is coming up right after this. Welcome to Big Technology Podcast, Friday edition, when we break down the news in our traditional cool-headed and nuanced format. We have Brian McCullough with us here in the house talking about everything from GPT5 potentially coming in to the Musk fundraising, and of course, what's going on with Apple and OpenAI and how Microsoft feels all about it.

Brian is the host of the TechMeme Ride Home show. It's a great show. You can find it in your podcast app of choice. He's also a partner at the Ride Home Fund. Brian, welcome back to the show. Great to see you. Great to see you again. I think, you know, this is what the fourth or fifth time we've done this.

But since Alex has been telling you about TechMeme recently, I believe, you know, FYI, if you don't want to go to TechMeme every 30 seconds like most of us do to read the articles, you can just have me regurgitate the articles to you every day for about 15 minutes, a really quick podcast. It goes well, as we always say, with big technology. That's right. Definitely go to TechMeme right home to check out your daily tech news. Come to us for, you know, on the weekends or on Wednesdays as we dive deep into the headlines. And let's just do that right now. So Open AI had a very interesting release this week, Brian. And by release, I mean press release, really.

They talked about how they were starting to invest in safety in a different way and made a big safety announcement. They have a new safety and security council. Second paragraph. They say, Open AI has recently begun training its next frontier model. And we anticipate the resulting system to bring us to the next level of capabilities on our path to AGI. I mean, talk about bearing the lead there. Like, why wouldn't Open AI just come out with a press release and say, hey, we're starting GPD? But I think maybe they've just taken such a hit publicity-wise with the leaving of Elias Sitzkever and the leaving of all this, like many of the folks in the super alignment team that they wanted to show that they were committed and sort of wrap the GPT-5 news and like a hot dog bun of safety.

I don't know. It's your perspective on what happened with this announcement. And what do you think about the fact that GPT-5 has started to train? So this could be a head fake by them, which we can talk about in a sense. second. But as you mentioned, I am a GP in two different early stage funds, one of which is AI focused. And one of the things that's been kind of unique about the spring so far is that people in the AI community and investors in AI have sort of been waiting on tender hooks for is, when is GPT5 going to come out? When GPT 40 came out, you saw a lot of, you know, there's

There's tons of startups that are doing like, what if there's a chat bot for customer service? What if there's a chat bot for you go through the Wendy's Drythrough or whatever? And that conversational step forward that GPT4O had obviated a lot of startups. So in the spring, a lot of folks have been sort of waiting to see if GPT5 was going to come out. A lot of people thought it could be out by now. I would say this is arguable, but I'd say maybe the majority of people thought it was coming this summer. So it's a bit of a surprise that essentially people were thinking that based on the cadence of when the previous models came out by saying that we've just started training. What they're essentially saying is that at a minimum, GPT5 is eight months out, right?

Can I give you a crazy counter here? And maybe this is completely wrong. But do you think that that might be GPT6 that they're training and they've got GPT5 ready to release if that was the expectation? People were so waiting for this that when GPT, remember there was that weird thing that was floating around and people like, this seems like a bot. It was GPT2 chat or something like that. And then it became GPT4. It turned out it was GPT40. People were thinking, is this, are they testing GPT5 in the wild?

So that's what I said is it could be a head fake in the sense that because the industry is sort of waiting. Because Open AI, as far as most people feel, although Anthropic. and others would argue with this is still the state of the art and people expect that whatever their next flagship model is going to be is going to leapfrog the existing stuff. So if if they're head faking us, what they could do is like we expected by the end of the summer released GPT5 because then everyone's like, oh, it's going to be a year out because the idea is it takes eight months to train and then you need months to like test and do the refinement on on the various weights and things like that. So,

What they're signaling to the industry is maybe it's a year out. Maybe it's a 2025 release, as I said. So then you could have Open AI competitors try to rush things and beat them to market. Because why else, I mean, it is mysterious that you're basically revealing your timeline, right? So yes, they could be head faking in that way. They could be saying, well, GPD5 is a, or signaling that it's a year away. and then release it earlier than people are anticipating. Or you're right, they've made a leapfrog and it's a second generation beyond.

But functionally, it doesn't matter because people in the AI space, there is concern out there that there is a ceiling to this technology, that the transformer technology that most of the current generation of AI is based on might have its limits. And those limits are around things like the data to train on and things like that. and energy, energy, accessibility to compute and things like that. So for so many reasons, people are watching Open AI to see where the state of the art is. Like I've said on my show, like if, if GPT4-0 had been GPT5, the narrative right now would be,

oh, we've reached our ceiling. There's a gating function to what this technology can do right now. And that would have reverberations towards everything, towards the investments that people are making to from Microsoft to VCs to whatever. And just what you mean when you dropped in our doc this sentence that you said, the degree, that basically waiting for GPT5 is freezing the market. Is it basically because people are like saying we have these certain capabilities? Now we can do things with those capabilities, cool things with those capabilities. But we don't know how much further to go in until we see that step change improvement. that is GPD5.

Exactly. Well, I mean that in a very basic level, which is there are a lot of startups out there that are essentially glorified wrappers around OpenAI APIs. And so when they release the new model, if all the work that you've done for your image creating bot, your chap, whatever, suddenly looks, if if GPD5 is a 10x or 100x in terms of capabilities, then all the work that you've been doing around the previous APIs might be obviously But it's also freezing the market because if Open AI continues to be the gold standard, people are waiting to see will this, remember the step change that we saw from GPT2 to GPD3 and then especially three to four. And so people are are trying to get a sense of will that happen again?

When GPT5 comes out, will it basically be like take the chessboard and throw it up in the air and start fresh? So I said it's freezing the market for investment, but also for people thinking about it strategically, because if Open AI is still, as most people believe, state of the art, cutting edge, the best practitioner of this with the best tech and the most money behind it, people are waiting to see what the leader can do. Now, conversely, let's go in the opposite direction. If this is right and they're a year out, then this is the window for an anthropic or somebody else to do. essentially release their version of GPT5 or what GPT5 could be. So while I'm saying that this is freezing the market, it also could be opening up a window of opportunity for OpenAI competitors. Yeah, all the breadcrumbs that we get also suggest that whatever's coming down the road for them, their next generation model, Open AI, they believe that it's going to be big. And this is from the information. Altman and his colleagues also believe that the newest large language model OpenAI is training.

will far surpass, far surpass the best LLM it sells today, putting to rest the growing questions in the industry about whether the technology has hit a plateau. What do you think when you hear that? I hear that as a good competitive stance, keeping everybody sort of on their toes, sort of uneven, because on the one hand, you know, Sam Altman will say to startup founders, like, don't do anything that is just a rapper because our technology is going to get so good that. it's going to blow anything like that away. Now, that is contradicted by the fact that a part of their business is selling their APIs to people to do things to it.

So if you are a founder in the AI space, if you're an investor in the AI space, imagine the fact that on one hand, you know, Open AI and Sam Altman have said from day one, we're going after AGI. We're going after a computer that can meet and exceed anything a human mind can do. In that case, we don't know what the business case is because, like, does that mean that the computer takes over the world and tells us how to run it better? Or do you sell that for people that?

You see what I'm saying? So they are right now sort of straddling the fence of, hey, we're allowing you to do all these things. Let a thousand flowers bloom, create your startup around our APIs and things like that. But also, meanwhile, stick with us because when the next version comes out, it's gonna blow everyone else out of the water

and maybe all these startups are meaningless. So bottom line, your prediction on GPD5, you think most likely, sometime in 2025. I'll give you 50-50 that it would happen before the end of this year, and then 50-50 that it would happen in Q1 of next year. I mean, come on, Brian, you can't go 50-50 that it's going to happen this year or next year. I'm giving you those are solid, tangible, those are tangible numbers.

It may, it's like basically your probability on that, on that line is that it will 100% happen sometime. 43 by the end of this year and then yeah that will accept okay okay so most likely sometime next year I think it'll be next year yeah yeah in the meantime there are already some near-term concerns that opening I is going to have to deal with and maybe near-term breakthroughs and big partnerships that we're going to see before we ever get to gpt5 and the first one of those is going to be this big partnership with apple that we're bound to see in wdc which is in I guess a week from Monday. So this is really coming close. And we're going to talk a little bit about what we know already about WWDC. But first, the very fact that OpenAI took $13 billion

from Microsoft and is now doing this partnership with Apple is already, I don't know if you want to say, raising alarm bells within Redmond and the Microsoft headquarters, but it certainly seems like Microsoft is curious about it, and I think they'd be within their rights to sort of want to know what's going on. And there's this information story. The headline is OpenAI CEO cements control as he secures the Apple deal, which is about Sam Altman. But there's a little detail buried kind of deep in the story that actually seems to be the most interesting news nugget of it all. Here it is. The Apple deal could complicate Altman's relationship with OpenAI's most important business partner, Microsoft.

Altman recently met with Microsoft CEO, Satya and Adela, to discuss Microsoft's concerns about how the Apple deal might affect the cloud software giant's own product ambitions. And the two executives discussed the servers Microsoft would need to handle Apple's use of open AI services. I mean, that is a very extraordinary type of meeting

where Satya is, I mean, obviously Microsoft and Apple are they're much more sympathetic than they've been in the past. where like the jobs and the Gates era where they were at each other's necks. But it sounds fascinating that Sadia is like, how many of my servers do I need to give to make Apple a more valuable company? And aren't you supposed to be our more aligned product team? So that's, that's sort of the original sin of this partnership.

Is that the right analogy? Like, the original sin of like OpenAI itself, the weirdness of Microsoft investing in OpenAI was that it was sort of a nonprofit that has moved to for-profit, and as we're hearing whispers, it's going to move 100% to for-profit. But from Microsoft's perspective, the original sin here is that they can't acquire open AI, right? So if in a different regulatory environment, or if this were 10 years ago, look at what Microsoft has done. They have essentially gone all in on AI.

They are putting AI into all of their. most important products from Excel to Teams to Windows 11 itself to you name it. And so they're basically making it a core component of all of their most important product, most important product lines. But there was a day-long Microsoft outage recently where you couldn't use things like teams and the AI within Word, the co-pilots and things like that because the Microsoft APIs are down. So think of how core.

these things have suddenly become to most of their most important product lines. Microsoft naturally would want to own that technology. If it is going to become, Microsoft is the most valuable company in the world now because essentially investors have given them credit for being AI first and maybe the most ahead of the Big Seven or whatever in terms of AI. And also because they're the most natural company to apply this technology and sort of this business sense. What I'm saying, though, is like, you know, meta's trying to apply AI inside their family of apps, right?

But they own their own AI technology. Google, obviously, is trying to integrate into everything, but they also own their own stuff. So the original sin is that Microsoft, if this is core technology to them, they want to own it. Now, we've seen them signaling that they have not put all of their eggs in the Sam Altman basket, right? They've invested in other Open AI competitors. They're, you know, in their earnings reports, they're talking about their CAPEX spending going up because, you know, we're theorizing that they're training their own. They're even, they are training their own.

Like, the news reports have them training huge models. Great. So my question to you is, no, and by the way, before we even get to that, don't forget the weird chaos around Open AI where they got blind. What weird chaos, Brian? That's in the past. They got blindsided by Sam Altman's ouster. I don't imagine that they're thrilled with some of the weirdness, like, around Scarlett Johansson and, like, the, you know, you went down the list of weird things that happened over the last two weeks. So it's clear that Microsoft is hedging their bets, right?

But what I'm also saying is, what would you, okay, you give me the odds on how tight this relationship between Microsoft and Open AI would be five years from now. I don't think Microsoft would ever divest themselves of their investment, but five years from now, will they not need Open AI? And Sam's not, he wasn't born yesterday, he knows that. In a way, the partnership between Microsoft and Open AI is sort of like what Apple's doing right now with Open AI, which is our own technology is not good enough. We know we need to launch AI chatbot something that is sophisticated enough now. And then three years from now, five years from now, when our tech is good enough, we don't need them.

Right. So that's essentially what Microsoft has done from day one. So you give me the odds of Microsoft and OpenAI five years from now being as tightly partnered as they are right now. Well, Brian, I'll have you know I'm rewatching The Wire. And what the guy's name, the detective Lester, tells you to follow the money and not the drugs. And so I'll just follow the money here. And we know that Microsoft gave OpenAI things. $13 billion. We think that a lot of that was either, you know, server credits or potentially

just plowed right back into Azure to train and deploy models. And we know that OpenAI is reported to be on schedule to make about $2 billion this year in revenue. Okay. So that's the money picture for Open AI. In comes Apple. Apple says Open AI, basically, we want to plug you into our operating system. And how much is that deal going to be worth to Open AI? It could be worth, this is from the information. It could be worth billions of dollars to the startup if it goes well. Not a billion, billions. Billions. That's their entire revenue. I mean, let's say it's even two billions. That's their entire revenue this year. And that's right now. What what Sam and Open AI want is they essentially want to maybe not replace the Google search deal that that Apple has had for 20 years

where Apple makes $20 billion. Obviously they don't want to pay Apple 20 billion. It's the the opposite. They're going to get paid by Apple to do this. Exactly. So in a way, if they could replace Google by being searched, Open AI is imagining five years from now, they're making $5, $10 billion a year from Apple because they are fulfilling all of these functions that include the sort of autonomous agents, the assistant, the searching and summarizing the web and all that stuff. Exactly. So I think that like it's good for, honestly, look, it's good for both opening and for Microsoft to diversify here. It's good for Open AI to have multiple partners

because it's more sustainable that way and it's sort of really able to chart its own future in a way that it wants if it's not reliant on just one company. And frankly, it's good for Microsoft because it allows it to both rely on a partnership with Open AI but also if it's developing and building internally.

We know that they have Mustafa Suleiman there working on consumer products now, right? Like it gives Microsoft greater control over how this technology integrates with its products and greater insulation from whatever, you know, potential chaos might ensue at a partner, although you would imagine that Open AI is a lot more stable now than it was previously. And one last thing, one last thing here. We also know that the priorities of Open AI and Microsoft might be different, right? Right. Open AI, I think at its heart, still a research house, right? And I was on air on CNBC this week. I was talking about how

chat GPT is still only at 100 million users. That's the number that it hit two months after launch. But ultimately, like, I don't think that that is like the live or die for them. It's about the technology. Whereas Microsoft was pushing open AI after the Sam Altman. This is a report we have in the financial times that Microsoft was pushing open AI after the Sam Altman blow up to be more focused on commercial products. So if that's what Microsoft wants by building these capabilities in-house,

Microsoft is able to get that. And then also, let's say, you know, Open AI comes up with like some massive research breakthrough. And let's say they actually are training GPT6 now. It's just like freaking human level intelligence. Like Microsoft will be able to benefit from that. So that's my perspective on this. Yeah, except that, um, I think that open AI is incentivized right now. Again, these are two extremely cany operators who this is not their first rodeo. Um, open AI is incentivized now because they know they can see the strategic angles here. Microsoft wants to have control over this core technology. They can't buy open AI, so they're going to try to build it in-house so no one can stop them from doing it.

So open AI is incentivized now to go product. You said, well, they're still more focused on the research. I don't think that's necessarily true. I think that the path for open AI to stay strong and independent is to productize themselves, which would run contrary to Microsoft running to productize themselves as well. This is extremely reductive. But in a sense, you can think of this as I've always said that AI is trying to make the Star Trek computer a product, which is you don't have to know anything.

It's like the shift from the command line to the GUI. Computing is not having to know, I got to go to this app, I got to go to this website, I got to type in these forms. The AI just does it for you. The first person to productize that, whether it be in Excel, whether it be inside of Instagram, whether it be inside Google Search or your Google Docs or whatever, that they imagine that that is people will pay, you know, tens and maybe hundreds of dollars a month for that, you know, per seat and enterprise per person and consumer. So open eye in that, if you look at it through that lens, is actually weak because they have to build their own brand.

You know, there was a story this week that I think that they just signed up Pricewaterhouse Coopers or PWC, as you're supposed to call them now, to the tune of thousands and thousands of seats. Well, so they have to do that from a standing start. Like if you take that reductive lens of who's going to be the first one to productize essentially having a butler that a computer butler that does everything for you, they would have to do that without a platform of their own and, you know, sign up for a chat GBT account and pay us per month and that sort of thing. Yes, but that's why these partnerships are so important to them. And that's why they really do win with these multiple partnerships because they can create that for Microsoft and they're going to create it for Apple at WWDC.

I'm saying for the long term, that doesn't work for anyone. Long term, five years out, Apple does not want to be using Open AI. They want to be using their own stuff. I'm suggesting that's also true for Microsoft, and I'm suggesting that's true for Open AI. They know that, and so they need to plow their own sort of path here to be competitive five years out. When Microsoft hired Suleiman. And that's why they might be a research house because they ultimately will have to keep advancing the status quo in order to be able to maintain these deals. Because if they just had stuff that was easy to replicate on par with everybody, then they would be out already.

when when Microsoft Aqua hired even though technically they didn't to get around again the competitive landscape or the the regulatory landscape hiring Suleiman was the biggest signal that hey by the way Sam we don't need you forever

but Suleiman is just consumer products and he was also building a failing company so true but they hired somebody that can be the vision like I'm not one of these people that denigrates or says that Sam Haltman hasn't achieved a lot of stuff, but he's not a product guy necessarily.

He's not an engineer. He's also not a researcher. I don't know, Brian. I mean, he, whatever you want to call him, he's the guy who's, whose company who is running underneath him took this technology that, you know, was built within Google and everybody had access to and actually built the product. And has the state, not only that, and has the state of the eye state of the art research underneath him. For sure. That nobody else can equal.

has, as we saw with the attempted coup, the backing of some of the most talented people in the AI field. So, again, I'm not denigrating Sam Elman at all. He's clearly the caniest operator in the space as it exists right now. And he brought the space to the forefront. So he deserves all the credit for that in the world. Yeah. But okay, so let's talk about a little bit about why Open AI is going to be in sort of why they're going to win in the near term for sure. Even though I agree with you, there's vulnerability I've been talking about on the show for a while. But I also think this Apple partnership might really change the game for Open AI in a way that I'm just starting to appreciate because it seems like Apple is going all the way in. And there was always a question of like, what was Apple going to do? So previous reports, this is from Mark German in Bloomberg.

So previous reports had Apple preparing a bunch of different features like voice memo transcriptions and summaries and recaps of websites and notifications and automated message replies. and advanced photo editing and AI generated emojis. Okay, all that to me, total snooze. But the report this week really brings into focus about how big Apple is going to go in on this, right? So this is from German again. Apple plans AI-based Siri overhaul to control individual app functions.

Now, the second part of that headline is really important. It's not about information retrieval like your traditional chat GPT. It's actually giving control of the app over to Siri. And here over, yeah, well, basically allowing you to control the app with Siri, not letting the AI control it for you. But here's the story from German. Apple Inc. is planning to overall Siri and move that will let users control individual app functions with their voice. The new system will allow Siri to take command of all the features within apps for the first time.

That change, okay, it required a revamp of Siri's underlying software. We know this. Okay, here is an example. At the start, Siri will handle one command at a time, but Apple has plans to allow users to chain commands together. For example, and I'm 100% certain this is going to be the hero demo at WWDC, but my certainty is probably just me being overconfident reading this, but it sounds likely. Okay, he says, for example, you can ask Siri to summarize a recorded meeting and then text it to a colleague and run request.

Or an iPhone could theoretically be asked to crop a picture and then email it to a friend. Right. And you don't have to say, and then open it, in photos and then crop it you just you just like like Picard says in Star Trek it's like you know computer enhance and it just does it right it is also what the rabbit R1 suggested was it was trying to do although whether it was successful doing that is obviously hugely debatable but imagine that you want to book a flight and you don't like once you give it access to your expedia account and your your your Gmail and your whatever you just say book a flight for next

Wednesday to Seattle. And that's it. It just does it and maybe you give it the price parameter, but you don't have to tell it. Open this app, do that. The access to the system level is what's key here. And that's what people and some of the papers that have been released by Apple Research suggest that they're going in this direction. The idea is that at the OS level, the AI will have the ability to do stuff for you, to execute commands. multi-level commands with things that would take different steps. That's why I say it's similar to the moving from the command line to the GUI, where as opposed to having to know how to, in a command way,

tell the computer what to do, you clicked icons and you did drop-down menus. If you need to plan a vacation now or do a doctor's appointment, you know that you as the human have to go into this account, go to that website, use this app, check. this account, you know, all that stuff is obviated by if the AI has the ability to execute at the OS level on your behalf, then it is Picard saying computer enhance without having to say, go to Adobe and then enhance 30%. You just say enhance more, enhance more. That's what we're, that's what Apple is suggesting. Okay. So let's just, as we end this segment, just kind of do like

a little scorecard here. So we think Open AI is going to come out of this this next year stronger i i'm gonna i'm gonna put a i'm gonna 50% stronger 50% weaker brian i still think that there is there's the there's a possibility that people are underwhelmed by w wdc what happens then if if the demos are not wow gee whiz enough um not only is apple's stock price in trouble but um like opening i could should still come out of this looking good. But more than, I was saying to someone just this past weekend, more than any other WWDC that I can remember for a decade, I don't know where this is going to land. So yeah, there's your 50-50 again. I don't know what the odds are that they're going to have an iPhone

moment where it's like, I didn't even know that was possible versus, oh, a slightly smarter Siri. I do think that WWDC is going to be very impressive, mostly because it has to be for both of these companies for Open AI to be able to continue to push forward. I mean, open AI will be fine if it releases GPT5 and it's like a godlike AI or not even that, but just like another step forward and what we were seeing. And I mean, Apple does need this. I think they really do. And I'm not rooting for this because I love to be wowed. I live for iPhone moments where the possibilities and the vistas are pushed forward. But if people come, if two weeks from now, people are underwhelmed and that's what you're talking about on the show. Remember I said that

there's a real possibility here that they don't hit the mark. Well, I mean, they were able to make the Vision Pro look cool in an era where like mixed reality. That has missed the mark. Yes. I mean, it looked cool and people talked about it for three weeks. Exactly. So I mean, these are two companies that really can do excellent marketing. And they often have products to back it up. But not all of them hit in the long run. Okay, I want to talk about Microsoft real quick, in terms of like wrapping up where things are going. What's your sense as to like the uptake on the co-pilots? Because that's some of the feedback I've been getting is that like,

okay, Microsoft, because you mentioned, okay, Microsoft, the most valuable company in the world. Yes, by a, by a good deal, actually. It's predicated on their ability to be first in AI and a lot of their ability to productize this stuff and put it into office and Excel and all these things. Are you getting a sense that people are actually using these co-pilots? Is this enthusiasm for Microsoft sustainable in terms of being able to be traced back to actual usage?

For Microsoft specifically, I can't speak to, but I'll use, I'll broaden it out to tell you what I, you know. In the end, the promise of AI and AGI is that the computers are smarter than us, so they make smarter decisions and we leave the decision making to them. where the state of the art is right now is essentially getting rid of the busy work of the you know instead of taking 90 minutes to create a a slideshow presentation just do it in five and you know tweak it um i'm seeing that uptake in a lot of places where there's busy work to be done uh law uh the the the uptake that i have seen anecdotally just from companies that i'm aware of startups and things like that. As an example, going through case files, depositions, things like that, things that you hire people to spend hours, weeks, months, years to do, like that sort of stuff.

So I would extrapolate onto that, that first of all, I hear all the time from developers that, yeah, I'll never develop without a co-pilot again. It is creeping into things like law, like healthcare, like education. If you're a teacher and you can use it to do lesson planning, do IEPs and things like that, and you've got troves of paperwork and things like that. So I'm answering your question by saying, I bet Microsoft is seeing uptake in it. We know this because they continue to double down on every product they're putting into. I'm saying anecdotally, I believe it too, because anything that is wrote mind-numbing work, can be replaced or not replaced with a co-pilot,

but made less terrible by a co-pilot. So, yes, I think that Microsoft is perfectly positioned for that. Before we go to break, we should talk a little bit about the fact that Helen Toner and Tasha Kali, who were the two OpenAI board members who sort of sparked the firing of Sam Altman. They actually said why they fired him this week. and again they basically said little so i didn't want to dedicate too much time here because it's been like the same sort of like he was not consistently candid like come on like actually like this it sort of seems like it was just a power struggle the more they talk um but this is just from the ft that toner uh

it's summarizing what she said on the ted ai show uh she said that he missed that sam misled the board of multiple occasions about its safety process for years she says sam had made it really difficult for the board to actually do its job by withholding information, misrepresenting things that were happening at the company, and in some cases outright lying to the board. I mean, one example that she gave was that the board didn't know chat GPT was coming until they found out about it on Twitter. How do you read this? Is this more concerning that I'm starting to think it is? Yeah, I think so, but this comes back to my disagreement with you that they are still research-based. I think that the fundamental disagreement within Open AI is there were a ton of people when it was founded in, you know, 2015, 2016.

I can't remember what year it was exactly. They basically got the best and the brightest of the entire academic space, right? And they essentially said, the whole premise was, we're going to do this independently because we, this is too important for any big tech behemoth to own. for any one company to own. So you have, I don't know what the percentage would be of the company at one point, maybe not now because a lot of those people have left, as we know, who believed in that as an academic and research mission. The Helen Toners of the world probably can't give us like more tangible examples of when

specifically Sam Altman was talking out of both sides of his mouth to people. That seems to be what they're suggesting. For various legal reasons, they probably can't give you tangible examples. My read is the fundamental problem here is that a certain percentage of the company bought into it being a research company and not a product company and that either Sam falls on the it's going to become a product company side or he's trying to play both sides as long as he can because it would help Open AI to go down parallel tracks as long as it can before. it has to make a decision. And so that suggests what essentially, if you read between the lines, the board has always said is you're telling me one thing

and then I heard you told somebody else the exact opposite. Right. All right. My point here is that it has no choice but to be both research and product. And I guess for us, it's like what leads? You're saying product leads.

I'm saying research leads. And then there are people that want to get off that train if it's not the train that they wanted to be on. That's true. Yeah. We also got an email from OpenAI last week talking about their alignment work. So they actually, I don't think they have a super alignment team, but they do have a They do have an alignment team.

They do have an alignment team. They do that John Shulman, who's their co-founder, is taking on an expanded profile, a portfolio as the head of alignment science. And they say it's a priority for Open AI and the company expects this investment to increase over time. So that's what they told me. We'll see. Okay, I'm sorry.

Before we go to break, one more thing, which is just an amazing detail with this Apple situation. This is from the information's report. In wooing Apple, Altman had to overcome skeptics within the iPhone maker. Some Apple executives have long had an aversion to chatbots, especially its head of machine learning, John Gendra, who by the way came over from Google.

Okay. In early 2023, after chat GPT exploded in popularity, he told staff in an internal all hands meeting. And this is after I assume, presumably he's tried chat chippy T. Chatchip T. He said, the last thing people needed was another chatbot. I mean, come on. Well, that maybe explains why Siri has been stuck in the mud for...

Lord Almighty. I mean, what, it was going on there. All right. There's somebody who thinks the world does need one more chatbot, and that's Elon Musk. We'll talk about his $6 billion fundraise and what he plans to do with XAI. on the other side of this break. Again, we're here with Brian McCullough.

He's the host of the TechMeme Ride Home podcast, which you can listen to every weekday, where Brian will give you a great recap of what's going on in the world of tech. Recommend it, one of our favorite shows. And also, before we go to break, this is going to be another one of our episodes where we're collaborating with TechMeme.

You'll be able to see it on the homepage of TechMeme. And I wanted to thank TechMeme for showcasing the podcast on its homepage. It's my go-to site for finding out what's happening in the tech, world as it's happening. Found a lot of stories for this week's show on TechMeme and then went in and read and I think you're going to find great value from it too. So that's Techmeme.com. All right. Elon Musk's $6 billion fundraising coming up right after the break. And we're back here on Big Technology podcast with Brian McCullough. He's the host of the TechMeme

Ride Home podcast and the partner of the TechMeme of the Ride Home Fund and another fund that's AI focused. What's the name of that fund, Brian? The Right Home AI Fund. And specifically I need to specify the funds are not related to tech meme the ride home fund is a separate brand brand right McCullough production it's a Brian McCullough production yes are you gonna invest or let's say this so Elon you know what I'm gonna ask already Elon just invested six billion I sorry raised six billion from Sequoia and Dreson Horwitz and others okay would you have invested in that musk xAI okay I have some real inside baseball for you which I have mentioned this on my

So I was offered that daily. So you often, when you're an investor, people learn you're an investor and you get offered things often secondaries from a company that's raising its B round or it's C round. And so somebody got a hold of an early employee that's trying to unload a million shares. And what they do is they spin up a special purpose vehicle. and they offer it to people that have funds like me. And they say, do you want X number of shares of this company that's going to do a C round or whatever? Very rarely has, I mean, it's not like this is the late stage round of XAI.

But I was offered this by a lot of people I've worked with in the past and people that I had never heard of. This was all over the place. and I'm not casting aspersions here. I don't necessarily know what that means. But Alex, I could have gotten you an allocation in this if you wanted. Like this was shopped around a lot. At the right home funds, we passed on it because we're an early stage fund. And so, you know, we investing above a $10 billion valuation is not necessarily the waters that we fish in.

Just for the how the math works out on our thing. Let's say you weren't. would you have put the money in? I mean, did it look like a good and investable opportunity for you? I think I still would have passed. Why? Because I think that the value proposition here is Elon Musk and then that he has Twitter. You know, everybody is concerned about where am I going to get high quality data?

This week, Vox and the Atlantic. signed deals. I think the Dow Jones signed a big, and these are big deals. And maybe we'll talk about that again in a second. But why did they go to the Atlantic? Because the Atlantic has been published since like 1843 or something like that, right? So you have high quality stuff going back forever. Is Twitter high quality? Now, Open AI has paid Reddit. Is Reddit high quality? I don't know. What I'm saying is is if the value proposition is Elon Musk. That's interesting to me, except for the fact that he seems to be distracted by so many things. I mean, the guy did also co-found Open AI. Did, yes. So also that makes me a little nervous. Is this a revenge job or whatever? But I would have passed on it because I don't know that just training it on

Twitter is enough. And then I would make the argument that people are behind. They are starting behind. Now, they have raised more money than Croesus to obviously catch up. They're going to buy as much as Nvidia will give them in terms of the compute that they need to do this. But I just would have passed for that reason. Now, forget about me. What do you think that the larger move here is? Because obviously, you know, Andreessen and Saudi Arabia and like big blue chip names clearly believe in this as a play. do you think that if Microsoft is going to do it,

Open AI's here, Google's doing it, Meta's doing it, Anthropic, this one, that one, is there room for another one? Like I've discussed before, is there a commoditization in the sense that there were 20 different search engines before Google came in and killed them all? Yeah, I mean, why not, right?

Like, if there's room for two, there's room for, you know, seven, I guess. Maybe not, but. Maybe not. But you're right. Look, I think that here's my perspective on it. So first of all, we do have some information about how they're going to play it. So first of all, they're going to spend a lot of this money on compute. And Musk has said he's, and this is from the information, he wants to build a gigafactory of compute.

And that XAI is going to need 100,000 specialized semiconductors to train and run the next version of AI GROC. He is saying that, okay, in the May presentation to investors, Musk said he wants to get the supercomputer running by the fall of 2025 and will hold himself personally responsible for delivering it on time when completed the connected groups of chips and Vida's H100 units will be at least four times the size of the biggest GPU clusters that exist today. So I think the question is like if improvement is going to come from brute forcing these models with more compute and more data, Elon Musk can get that compute, he can get the data, right, from Twitter or wherever he figures it out and common crawl, wherever

might be. So, so it seems like he's, he might, he will probably be able to catch up if his computer is going to be Forex size. Then the question comes down to will, will you end up with technique that helps you get better beyond size? And everybody's going to make a bet on technique. And I think that like venture capitalists, you know, they could put money in and lose. And that's okay. Right. but they bet on the chance that this can win. And is there a chance that they can have like the right technique that helps push this forward? I think it's at least a probability. I don't know if it's likely, but it's possible.

And also like the one wildcard is, and I know that it's probably not going to happen, but Elon Musk was able to recruit Ilius at Skever to open AI once. Can he recruit him again? Can he recruit somebody with his level of talent to come do it and become, like the chief scientist at XAI, I think it's possible. And if that happens, all bets are off. I had Nat Friedman, the AI investing extraordinaire. There's nobody that is a more prominent personal investor

in AI right now than that Freedman, ex-CEO of GitHub. And he very strongly believes, because we asked him, like, can the biggest model, will the biggest model always win, which is a little, there's fudging around the edges. But essentially the question is, the most compute, the most data, will that always win? And he's, he was like, so far, I haven't seen anything that suggests otherwise. The more compute you throw at it, the more high quality data you throw at it, the better it can be. So yes, there is a, there's a philosophical debate right now versus what you call technique, what we call terroir, which is like, yeah, but do you,

just want GPT6 to be reading x-ray scans or do you want a specialized model that is trained on medical stuff and x-ray scans because maybe that would be better? Possibly Elon agrees with Nat and is like, look, just go big because until we see proof otherwise, big is continuing to win here. It would be interesting, you know, speaking of the compute, like, you know, Sam Altman, know, the story was he wants to raise $8 trillion or whatever it was to create. Seven trillion, man. Yeah. I mean, come on.

Well, listen, I dream bigger than even. I know. No, he joked why not eight, I think. Right, right. So, like, let's say that the pitch that I got for XAI was that, which maybe it is, then that's more interesting to me because if they're right and then building these superclusters is what's going to matter. then that's more interesting. But I would have passed because I don't know where it's going to go.

Other people smarter than me are fine with, listen, it's Elon, so we're in and we'll see. Yeah, well, Larry Ellison is certainly happy because he's going to be, yeah, he's going to be making some money licensing his Oracle servers out to make this happen. we got an interesting comment actually here so general foundational models will be commoditized and the value will be captured and vertical specific models trained on private data and Mike uh Mikkel or Michael whatever um that is the debate because on the one hand that is part of the thesis that I'm investing in which is um verticals and what I'm calling terroir in in the wine sense where what you train it on and how you do the the knobs and levers of the of the of the of the way and things like that will make things better. But again, like Nat Friedman, 100% believes that that's not true. And that the bigger, beefier model has until proven otherwise continues to beat even the most specialized models. Yeah. And maybe he's making a bet on both with this giga, what did he call it, a giga computer?

Gigas. Anyway. Gigac. A gigafactory of compute. It's actually excellent branding. He's nothing if not consistent. He held onto the X.com domain name for 25 years before he could finally. And everything is gig up.

Yes. Yeah. All right. Before we leave, I want to talk quickly about this. Oh, so first of all, I should tease. We talked a little bit about these deals that Vox and the Atlantic have made with open AI and whether the news industry can survive AI and whether these are advised or not. Big story this week, obviously, like in the media sense.

on Wednesday, Ben Smith and Naima Raza, who I just started this new podcast called Mix Signals. They're coming on with Joe Marquesi, who's a venture capitalist, and we're going to talk all about whether these deals are advisable or not. And it's going to be fun because Ben and I were both at BuzzFeed. And obviously, BuzzFeed made some deals with Facebook and Twitter and thinking it would be good to, you know, distribute the content. And that didn't work out exactly as planned. So that's coming Wednesday on the podcast. But let's, before we end, talk just briefly about this loneliness article because it was very interesting. So the Wall Street Journal had this story talking about how in the age of remote work, American workers, really workers around the world, I think, are getting way more lonely.

And they had some wild stats. So they said, I couldn't believe this, Brian. More than 40% of fully remote workers pulled in a 2023 survey of working parents said they go days without leaving the house. Those who work in office spend nearly a quarter of their time in virtual meetings and face-to-face meetings account for only 8% of their time. Also, this is crazy. Americans in particular have tripled the time spent in meetings since 2020. And let's see, 68% said they knew their coworkers on a personal level down from 79% five years ago. Are we under appreciating the impact that this era of remote work is going to have on people in terms of just making them

lonelier than ever before. And we're both, like, I think we're both remote entrepreneurs. Like, do you feel this at all in your life? I'm curious what you think about this. I have never worked in an office. Like, I have an office, but it's a we work like situation. So it's just me. I've never, you have, you worked in newsrooms and things like that. I've, never had that. My first company, I found it in 1999 was all of the work. workers were remote distributed. So what I would say about this is it's not for everyone, but that's the point maybe that this data is revealing,

is that if you were someone who went to college to get a certain degree to do a certain type of work, and I'm not saying that people pursued careers to work in offices and cubicles, that you know, people have have bemoan that for as long as cubicles and things like that have been around. And then like look at the movie The Apartment from the, is it the late 50s or early 60s? Like the office has always kind of seemed a little dehumanizing to certain people.

But if you're not prepared for the fact that like you can hear my dog flapping his ears over there, if you work from home and you're not used to that as being your sort of uh, yeah, that's why I'm right by a window. That's why like it's a different skill set. So let's say that you were someone that worked in an office for 10 years. And especially in Silicon Valley where you're used to the perks and the the, the, um, the, the gym membership and all that stuff. Like, I can see that this is you would be like, this is not what I signed up for. I think what I'm saying is, um, it's, it's not for everyone. There are a lot of people that I know that that prefer it. You mean I don't have

to give up two hours of my day for a commute. Awesome. Yeah. But yeah, I can see how people would feel a little rug pulled by being like, this is not what I signed up for. And what's with all the meetings, Brian? Can we end this podcast getting together and finding a place of agreement that the meetings are becoming too much? Like I said, the meetings, let's see, I've, okay, here it is. Paradoxically, meetings can make people feel lonelier. And even so if the meetings are virtual. A 2023 survey by employees, by employee experience and analytics company perceptics found that people who described themselves as very lonely, very lonely tend to have heavier meeting loads than they're less lonely staffers. So more than 40% of these people spend more than half their work

hours in meetings. And right, this is the stat again, that Americans have tripled the time they spent in meetings. Can't this meeting just be an email? I mean, you're right. Maybe AI can't solve it for, by having our like co-pilot show up well meetings but it's absolutely astounding to me how meetings have multiplied the way that they have since 2020 the the the hotness of the moment in the AI space is autonomous agents the amount of startups that i've seen that are like have our AI agent attend the meeting for you um it can answer questions on your behalf because it's been trained on your your email so it knows what your action items are you can also prompt ahead of time. And then so you don't have to attend and it will give you a readout afterwards so that

it'll be like sort of like getting the inbox zero at the end of the day. All of those eight meetings that you had today, you didn't actually have to attend. We can also get to a place where you train an avatar on yourself. So maybe people don't know that Brian didn't attend. So maybe AI is going to solve that for us. I think that people hated meetings and the joke was before COVID Times, like, couldn't this have been an email? But, like, scheduling an in-person meeting was always harder than being, like, just click on Zoom. So I can see that COVID times and remote work has just made. Maybe calendarly a little bit has caused that as well. Calendar links. Maybe they're causing more meetings. Well, Brian. It's taken away the friction of doing meetings is only made

meetings multiply like cockroaches. I know, but I do like Calumably. Anyway, I'm with you. Less meetings, more time in person. If you're working remotely and you've been in the house for multiple days on end, please walk out of the house now. Please take these headphones off and walk out of the house and go feel the sun on your face and potentially go to argue the opposite. I was going to argue the opposite. This is what keeps podcasters in business is the fact that since we're all by ourselves all the time, that we have these asynchronous friends that are our podcasting buddies that we feel like we know, even though we've never met.

Let me try that again, Brian. If you are at home and you haven't gone out of the house in multiple days, keep your headphones in. Go through the back catalog of... Feel the sound on your face. Listen to every one of these podcasts. Listen to all the tech meme right home podcasts. Go to a local coffee shop and say, hey, I can't believe I haven't told you about this podcast that I love. And then you will feel happy.

Yeah. And that is our message. So kumbaya, less meetings, more in person interaction. Lots of fun AI news coming up in the next few weeks. We'll be here to cover it. Brian is going to be there to cover it. You can listen to his show TechMeme Ride a Home podcast on your podcast app of choice. You can listen to our show, Big Technology podcast on your podcast app of choice, indoors, outdoors, it's your decision. On Wednesday, we'll be back with Ben Smith, Naima Raza and Joe Markezy talking about the implications of all of these AI companies and news organizations doing deals. with each other. Until then, I hope you have a great couple of days and we'll see you next time on Big Technology Podcast.