Oct 5, 2024 ยท 2024 #35. Read the transcript grouped by speaker, inspect word-level timecodes, and optionally turn subtitles on for direct video playback
Edit labels for this show, save them in this browser, or download a JSON override for the production folder.
Transcript Playback
AI and Venture Capital
Human Transcript
Timed transcript
Blocks are grouped by speaker for readability. Expand a block to inspect word-level timing.
Speaker 3
Time for technology. That was the week. with my old friend Keith Teer. The big news this week is that OpenAI, which I'm sure you've all heard of, if you haven't, you probably don't deserve to be watching or listening to this show, raised $6.6 billion in the largest VC round ever, and Keith is focusing on that in this week's newsletter, including his editorial. How big a deal, Keith, is this $6.6 billion latest raise by OpenAI?
Obviously, it's big numerically. It's the biggest ever, as they say, although Microsoft did put $10 billion into OpenAI earlier in its life, but that wasn't a VC round. So it's big numerically. It's big in what it tells you about venture capital. This did have venture investors in it, like Binod Khosla, for example, but it really can't be described as a venture capital round because The round itself is bigger than almost every venture capital fund. So there's something new happening here, which is a bit of a throwback to the days before 2021, when very large investors were writing very large checks. For example, SoftBank put half a billion dollars into this round. But they're not really venture capitalists. They're late stage, following the money, hoping to make maybe three times their money on the low side, a lot more on the high side. Which is a lot.
If you invest a billion dollars, then you make 3 billion. I remember Keith, we haven't talked about Tiger Global for a while. You used to talk about them all the time and the changing nature of venture. Are we going back to firms like Tiger? I'm not sure Tiger wasn't part of this round.
Well, We're both going back and going forwards at the same time. The bit that is like going back is these big checks. And Tiger is in this round, by the way. They are an investor in this round. But the bit that's different is in the past, these companies were discovered at the seed stage by people like Aileen Lee, who we talk about this week, and then went through a middle phase being invested in by funds like Sequoia or Accel or Benchmark. and then went through a late stage where money like this would flow. But the relay race between the stages has disappeared. OpenAI seemed to go almost from nothing, a few seed investors, to these huge rounds without going through the middle. And so that raises the question, what happens to the early stage funds that don't initiate these large companies and are therefore not in the value flow these large companies and and i i think that that's what's different this is high high stakes you know risk um being played by very large check riders very early in the life of a company yeah the
The narrative has been turned on its head. It's almost as if the principles of quantum computing have been applied to today's world. We'll talk about quantum later. You mentioned the Eileen Lee's response. One of the meatiest pieces in this week's newsletter is a piece from Lee when she asked, do startups have a chance versus big tech? In the age of AI, and I guess you could also read that, do VCs have a chance versus big tech in the age of AI? History says yes in due time. Patience, Eileen Lee is calling for. What's her argument, Keith?
Yeah. Well, it's easy and convenient. And, um, you could even argue a little bit lazy to do that. Um, cause it, it means you don't have to think, but her argument is that, uh, you know, the web, Tim Berners-Lee didn't own all of the value from inventing the web and Apple and Android didn't own all the value from inventing smartphones on mobile. And therefore, um, Same with Web 2.0. The companies that created the infrastructure for Web 2.0 are not the same companies that benefited like Netflix or YouTube or the SaaS companies. So she's arguing that AI is going to be like that, that the foundation level companies, of which there are maybe 10, are going to create a platform on which lots of value is created that they will not participate in. The timing for that is not yet. Therefore, the right thing to do is wait until the layers above the foundation level become investable. And so I think that's a debatable point of view. It could be right.
Well, I mean, everything's debatable when it's speculative. No one can be sure. She uses history. It's a nice piece. She talks about platform shifts. She talks about previous platforms which have benefited, like Netflix and TikTok. I mean, isn't there some... I mean, I take your point. Maybe she's slightly lazy in her thinking. But isn't it inevitable that the foundations of AI are being built? and that there will be winners in the future that we can't even imagine. And in a sense, the web is the equivalent of open AI or what once was the web is now open AI.
Well, here's the, here's the reason she might be wrong. And I don't know that she's wrong, by the way, she could easily be right, but, uh, there, there's at least some, and she, she herself, um, has a paragraph towards the end saying, AI is different from the others, and she spells out why. The difference is AI isn't not really a platform. It's more of a replacement technology for knowledge. And once you have a replacement technology for knowledge, especially as we're now beginning to get agents, that is to say mini AIs that can use the platforms, it's at least conceivable that the foundation layer companies will fill up every other layer themselves.
And they want to. I mean, the whole reason why OpenAI has such an enormous valuation or that Google stock is worth so much or Microsoft or even Apple is that it's taken for granted that these companies will participate in the other layers.
Yeah. I mean, for the more technical audience... The very concept of a led stack, which previous innovations have had, may not be an appropriate concept for AI. It may be such a replacement, a more profound platform shift that you can't recreate the old apps that lived on top of the prior platforms because there's no requirement for them anymore. Like Khan Academy, for example. Take Khan Academy, which was a layer on top of the web and on mobile. It's conceivable that AI can transform education all by itself from the current foundation level using agents and expertise. We talked about expertise a couple of years ago. But expertise in different subjects. Sorry, a couple of weeks ago.
You might not acknowledge this, but I sense a hint of pessimism, unusual, un-Keith-like, un-Tier-like pessimism in your thinking, because You've always been Mr. Democratize technology and the web and opportunities. But if what you're saying is true, then we're going to just see stronger, more and more dominant monopolists or oligarchs, whatever you want to call Google and open AI. Do you fear this?
To be honest, I don't. I haven't conceptualized my ideas in an optimistic, pessimistic framework. It's more realistic. One of the recommendations from Aileen Lee, and my colleague Rob Hodgkinson wrote a piece this week which echoes that, is be patient and wait. And based on my analysis, the right thing to do is to be aggressive and move fast. Because if you don't, the spaces may shrink. And so I'm not really arguing for passivity. I'm arguing the opposite.
No, I didn't say that. But what I'm saying is that you've always believed that new technology would enable the democratization of opportunity and wealth.
Well, let's take it step by step. I mean, what price do you pay for free is a separate question, let's say. But I do think ChatGPT4 is already free. They're rolling out a new thing this week called Canvas into the paid version that costs $20 a month. But most of their revenue, by the way, which is going from $3 billion this year to $11 billion next year, according to their projections, is coming from API payments made by third party developers, enterprises and the like, which is how it they're seeping into everything through API use.
I mean, that would support Lee's position that it's a foundation and they monetize through APIs and then you get new companies, new developers, new opportunities, new product.
I don't think that's their plan, but it happens by default because what they've built is so good and the APIs are so available and relatively cheap that enterprise IT departments are building directly with it, not hiring other companies to do so. So enterprise sales is a You know, Mark Benioff has just announced what he calls a hard pivot to agent-based AI. Why? Because a website with forms to fill in doesn't look like the future anymore.
So what would happen to, in this world, if it's straight from open AI to the corporation, what happens to Benioff and Salesforce? Are they part of the... the obliteration of this middle layer?
There's a narrative that suggests that. It's called systems of record. Systems of record are what we think of as databases with real-time data that feeds the enterprise. And systems of record are what most third-party companies focus on for management tools and workflow and all those things, of which Salesforce is a big one. If agents, if AI agents can access data and interpret it and act on it, yes, Salesforce is at risk.
And I always think the word pivot, particularly radical pivot, is just a euphemism for recognizing that whatever you're doing is not working, so you have to do something else. So Benioff's a smart guy, and he must understand exactly what's happening.
Yeah, he was a guest on the All In podcast. podcast summit and they tried to get him to talk about this and he said he couldn't and wouldn't. And they did a show immediately afterwards extrapolating from why he wouldn't, what the risks are for him. And it's quite an intelligent conversation worth listening to.
It's a big deal. I mean, it might sound slightly abstract to a non-technical audience, but it changes the very nature of the tech business. You had a couple of other interesting pieces, five qualities of AI apps by Max Reed, and then something from Ethan Mollick on AI and organizations. Did Mollick and Reed touch on this?
Max Reed is really trying to home in on use cases and at an abstract level saying, here's the things you need to be good at if you want people to use AI that you deliver. And I think he kind of concludes that OpenAI is already doing most of those itself.
implicit in his narrative, they are. The other one, which is more about enterprise, is a little bit different. It's basically saying what you need to do to be good in the enterprise, integrating it. And that's the point about speaking directly to the APIs and building it yourself. And it's kind of a roadmap.
What about small companies, though? I mean, most small companies don't have AI departments or AI experts? Aren't people still going to be dependent on easy-to-use products?
Well, remember, there's a huge distinction between an AI model and then using an AI model. The foundation guys are building models, that is to say, core intelligence. You could argue whether it really should be called intelligence, but anyway, core capability, let's call it. And training those models is super expensive and uses terabytes of data. But the end product from the training is something that is only a few gigabytes big. That's the model. And it can act all by itself without ever being required to refer to that data ever again. Now, you can put those models on a smartphone and they'll function. So. implementing the models of the big guys is super cheap and easy, almost free. So there's the democratization piece. Well, it's not,
but it's democratization in a sense, but it's also knocking out the middle layer and empowering, making open AI, maybe Google, one or two of the other core companies, more and more powerful. What about Apple in all this? There was an interesting piece. You didn't put it in the newsletter. I think you forgot. It's on Apple's AI is landing soon on iPhones. I was really struck by the fact that you, Keith, who is the ultimate... Apple head, if that's the right word, or fanboy, didn't even buy the new phone. Because if you're not buying it, then God knows who is.
Yeah, but I'm not criticizing. It's just the reality. Every year, I've known you for the last 15 or 20 years. Well, I guess there haven't been 20. In the last 15 iPhones, you bought the new one. This time, you've chosen not to. Again, it speaks to the way to borrow some language from Marc Andreessen. that software, at least in the form of AI, is really eating the world and that the hardware is... It's not irrelevant, but it's trailing behind.
Yeah. Well, obviously, we all use AI through hardware, but existing hardware is good enough for that. So, you know, especially when it's a chat interface, like chat GPT is, you could run it on a Raspberry Pi, which costs... $199. So hardware is not a block. Although, of course, on training model side, it is. But on the user side, no, it isn't. And Apple, by the way, hasn't shipped Apple intelligence yet. It's promising it now that it's launched iOS 18 and is it Sequoia, the new OS on the Mac? I think it is.
Yeah, because they've got lots of images of big trees. Apple's always done a great job making new technology accessible for regular consumers like myself. I'm not the kind of person to go out and buy a raspberry pie, Keith. Do you think, and I'm still not convinced by ChatGPT4 or 5 or 6's relevance to myself, do you think Apple's integration of AI may be a breakthrough, maybe in a sense a Netscape moment for most mainstream consumers?
I'm going to say no, because I think Apple's constrained the use case so narrowly down to things like summarizing emails or spell checking stuff you write, all of which is using AI and is perfectly good. But it's such a narrow scope. I think already youngsters, I spoke to a class of 20 students when I was in Brussels, and they all use AI. And they use AI for a myriad of different use cases, but they don't have to switch apps to do that. They can use the same AI app for many, many different use cases, which tells you that the scope of AI is pretty much unlimited. And the idea that you're going to need an app for every use case is ridiculous because the AI can do all those.
Maybe we're at the end and we're always looking for ends of ages. Are we at the end, in a sense, perhaps of the app age? I mean, there's nothing inevitable about apps.
I remember a time when there were no apps. We certainly should be. I mean, why do you need an app for knowing the weather? Why do you need an app for knowing the time? You know, why do you need an app as a calculator? it literally doesn't make any sense anymore. So you can assume that Apple's trajectory is that the phone is more and more minimalist. Well, the phone is the app. Well, the phone is just a container for processing.
Right. Now, put that in the context of Facebook announced its glasses this week that are truly AR. They're not very good yet, but they're as good as the Apple $3,500 thing. Which you bought. Which I did buy. And you combine... earphones with glasses with a microphone, and you've got the interface to AI that can deliver anything you want. And so we're definitely in a moment where how we access knowledge and information, how we communicate, we're at a tipping point of a new set of experiences.
Yeah, and the tipping point, I think I actually probably a little bit more optimistic than you. I think I'm not sure if Eileen Lee is right in her perhaps over-reliance on history, but every tipping point creates new opportunities. Maybe we just can't imagine them.
By the way, I think she is great because no one else is writing. You're being lazy, Keith. Can she be lazy and great? Well, at least she's addressing the topic and realizes there's something to answer. And by the way, has enough caveats, just like I do, because there's no truth here. We're all guessing.
Right. And everyone's threatened from VCs like Lee to entrepreneurs like Benioff. Even the owners, I mean, if you're right and we're at the end of the age of the app, then Google and Apple should be particularly worried because much of the value of their businesses are based on their app stores.
You know, it's funny that Mark Zuckerberg for two decades now, at least since 2007, has been really angry that the hardware that sits between us and his content isn't owned by him. Is that why he changed his hairstyle? He's changed everything, hasn't he? He looks like, I don't know.
But he's been focused, laser focused, really. Remember, he once tried to do his own phone on owning the interface to content and information and communications. And the glasses is his latest attempt. It definitely is the case that Apple and Google could lose the audience if they lose ownership of the devices we leverage to get access to content, communications, information.
And fascinating stuff. I mean, these are huge issues. Once in a decade or perhaps in a century, we get these kind of things. We'll come back to that. You mentioned earlier, Keith, that you gave a speech at a school in Brussels. We missed last week's show because you were in Brussels, I was in London, and you... wrote about unleashing AI for Europe. Can Europe produce world-class AI innovation? You've always been pessimistic about that. Did you come back from Brussels more or less pessimistic about Europe's ability? I mean, and in terms of this open AI issue, it's bad news for Europe. I mean, they're doubly now in jeopardy. I mean, they couldn't keep up with Google and Apple. They're certainly not going to be able to keep up with open AI.
Yeah, but they're fighting, speaking about fighting yesterday's battle. I mean, worrying about monopolies in the App Store from Google or Apple is profoundly counterproductive in this new environment.
Yeah, which we talked about a couple of weeks ago.
Words and timings
Yeah,whichwetalkedaboutacoupleofweeksago.
Speaker 1
Yeah, and now it's interesting, Draghi puts Europe's weakness down to the fragmentation of languages and rules-based regimes, and argues for a stronger centralized EU. I think the opposite is needed, which is bottoms up innovation from a culture of risk. That's the missing ingredient there. It's interesting. Europe only spends $55 per head on innovation in venture. The US spends $500 per head per year, and the UK even $320 per head per year. So Europe just isn't
isn't a bottoms-up innovation culture. Worse still, of that $55, more than 60% comes from government organizations, whereas 100% of the US comes from private money. So you've really got a top-down, stultified, risk-averse culture in an era of innovation. It just can't work.
Yeah, and all the value in European economic terms increasingly is in its it's museum-like qualities. So people go and visit literally its museums, its beaches, its beautiful cities. But by definition, that seems to make the economies increasingly, if not marginal, certainly unable to compete with the US and presumably also with China. I mentioned quantum earlier. You were nice enough to make my interview with Martin Schmidt, who's the president of Rensselaer Polytechnic Institute, RPI, which is up there with MIT as a top tech school about quantum. RPI is the first university to acquire a quantum computer from IBM. I'm not sure if you looked at the interview, but... Are we at the beginnings, Keith, of the quantum age, just as AI obviously is changing everything? Is quantum existing in parallel or could it conceivably power AI?
You may have to help me out here, Andrew, because I am not even a good amateur when it comes to quantum. It's a strange world to me. I don't really understand how qubits, which is the unit of storage in quantum, work. I do notice in tests and coming out of labs more and more proof that quantum can be very good at certain things. I'd be interested in what he says about whether it can be good at all things.
Well, he doesn't say that, and you'll have to watch it. You made it the interview of the week. Everyone should, of course, watch it, especially that was the weak viewers. Keen on people will probably have already seen it. He makes an interesting case. He doesn't overpromise. One of the takeaways that I thought was particularly interesting, actually, from the Schmidt piece was how he believes that the biden chips initiative massive investment is actually beginning to change uh upstate new york and this represents the beginning of a fundamental shift in the u.s economy so it's not just always about open ai and google and apple it's about new factories new companies new kinds of hardware initiatives so it's it's an encouraging piece and he's uh He's an interesting man. Let's move on, Keith, to startup of the week. You didn't have the cheek to make it OpenAI. Instead, you made it a company called Cerebras. Tell us about Cerebras. Why are they interesting?
Well, firstly, they filed for an IPO. Secondly, they build AI chips that are specifically built for AI. They're called wafer scale chips. So they're much bigger than the NVIDIA chips that people use today.
And almost all of it's coming from one customer that is leapfrogging current generations and using their chips for AI. It's a Middle East-based AI platform. And they claim that their chips are 20 times faster than Intel, 40 times faster than the latest version. Sorry.
No wonder Intel is rumored to be on the acquisition block.
Words and timings
NowonderIntelisrumoredtobeontheacquisitionblock.
Speaker 1
Yeah. And by the way, I meant to say NVIDIA, I meant, not Intel. And so they're probably the first semiconductor company saying we're better than NVIDIA. And if they're right, obviously it's going to be super valuable in future. Now, it did come out this morning that their CEO 20 years ago was convicted of accounting fraud.
Oh, dear. It also speaks to the broader conversation this week about the stack. And maybe we're still in the earliest stages of this. We talked before about how Cisco drove the original Web 1 boom of the 90s. So maybe in a sense, we're in such early stages of all this that things are going to dramatically change. We haven't even got to Web 2 or 3 with AI. Yep.
Entirely possible. I know I have a friend who works at Cerebrus. He's like the general manager that does all the office bookkeeping, stuff like that. And he speaks very highly of the team and the CEO. Well, he would do if he works there. No, you've got to know my friend. He's a contrarian. So if there was anything to be said negative, he'd be saying it.
Well, finally, you've changed the name of... of our final feature on the show to post of the week from uh uh what is it elene uh mitchell and i don't quite know why this is the post of the week and i'm reading it out three richest people in the world elon musk 256 billion zuckerberg 206
Bezos 205, they're all not that far away from being trillionaires. There have been some people who suggest that Musk will be the world's first trillionaire, which doesn't actually sound that unrealistic. Why did you choose this post, Keith?
But I also thought it was quite topical. Who is it from? Noel Ellen Lee Mitchell? I actually don't know. I'm in the same boat as you. I have no idea who this person is. I don't know where the spaces would be if it was broken out. That said, I thought it's quite topical because Billionaires are like the new devils, aren't they? In the popular narrative, there's nothing as bad as being a billionaire. It comes with suggested devil horns.
Yeah, he's wealthier now than Bezos, which is quite...
Words and timings
Yeah,he'swealthiernowthanBezos,whichisquite...
Speaker 1
Yeah.
Words and timings
Yeah.
Speaker 3
I never would have guessed that.
Words and timings
Ineverwouldhaveguessedthat.
Speaker 1
And Bill Gates and and the Berkshire Hathaway guy as well. Not not. So it's a symptomatic that back to the theme of open air, large pools of money are getting larger and dominating. And that begs the question, how do you democratize the benefits of what are obviously great human scale achievements? And I don't have the answer, but I thought this tweet of the week or X of the week or post of the week actually would be interesting. would put that question right in the scope.
Yeah, it's interesting in the context of the OpenAI news about raising $6.6 billion. I know that Sam Altman has renegotiated whatever business arrangement he had and presumably will end up owning more stock. Could you imagine, because you've always been very bullish on AI, could we imagine Sam Altman at some point getting onto this list?
Hard to tell because we don't know the structure yet, but it's entirely... Certainly, OpenAI, I think, is going to become the world's most valuable company.
As always. If everyone agreed with me, I wouldn't be me.
Words and timings
Asalways.Ifeveryoneagreedwithme,Iwouldn'tbeme.
Speaker 3
Yeah, although it's not a particularly outrageous thing to say. And... I mean, what happens, we need to end here. I don't want to make this into another Elon Musk conversation, but what happens if one of these people, Musk, Zuckerberg or Altman becomes a trillionaire? What will be the implications? We already see Musk in some ways taking on foreign governments, sometimes even beating them. Is this something, you've always been a great critic of the nation state, Keith, but is this something that you will welcome, that you will celebrate the world's first trillionaire?
No, because I don't think billionaires are better than nation states. I do think the human race is going to be challenged to answer the question, what could be better than nation states? But the answer will not be billionaires. I do think in the absence of an answer, billionaires fill the void. And so if you really don't want billionaires as global billionaires, you know, globalists to have a lot of power, it's incumbent to come up with an answer of what's next for humanity. And, you know, I don't have all the answers to that, but I still- You don't have any of the answers.
None of us do, and we know that. Yeah, but I have the question. Most people don't have the question. Well, it's easy to have the question, what's next, but none of us know. And it seems as if we have companies like OpenAI, or the Musk companies, which are increasingly powerful and wealthy, making them increasingly powerful and wealthy. And there's very little evidence of anything else.
Yeah. I mean, my intuition is that it's some combination of highly local democracy and global frameworks that allocate resources to agreed human goals. But that's a bit vague and abstract.
That is very vague and abstract. You sound like some UN guy or a mid-level academic, Keith, in a business school. We will come back to this subject. You are not alone in not understanding the future. That's why we have. That was the week. And we will be back next week to talk more big tech with Big Keith Teer. Thank you so much, Keith.