Jul 7, 2024 ยท 2024 #24. 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
Rise of The Algorithms
Human Transcript
Timed transcript
Blocks are grouped by speaker for readability. Expand a block to inspect word-level timing.
Speaker 1
hello everyone we are back with technology and my friend keith tear the editor of that was the week the tech newsletter where he aggregates all the content last week we talked about uh whether or not we asked whether silicon valley has an ai bubble problem that was the title of keith's A newsletter from last week, is there an AI bubble? I answered the, which of course is a yes. I think there is a massive bubble. And this week, though, Keith, you have the rise of the algorithms. So what is it, an AI bubble or the rise of the algorithms?
I mean, there's an AI bubble and then we have the rise of the algorithms. It's got to be one or the other because AI is algorithms. So either the algorithms are rising, it's a genuine rise or there's a bubble or maybe both. I mean, I know you're looking at this from an investment point of view, too.
Well, actually, I said yes and no. And on the yes side, I said there is one, but it's a good thing. And the rise of the algorithms is a good thing as well. Well, the rise of the algorithms is all about the main news story this week, actually, which is that BlackRock, which is the largest investment company in the world, has spent $3 billion plus buying a company called Prequin. Prequin is a company formed in London quite a number of years ago, 20 or 30 years ago, that is one of the companies in the space of data for private company investing, a little bit like Crunchbase. And $3 billion is a big price tag. And the CEO of BlackRock announcing why they're buying Prequin said that they believe that there's going to be a trend towards, in quotes, indexing private markets. And, of course, that made my ears pick up because my company, SignalRank, indexes private markets, specifically Series B round venture-backed companies. So I decided that BlackRock, which, by the way, in its previous history, bought Barclays Global Capital, which owned the iShares brand of... public market funds, which are also indexes. BlackRock is the world's leading indexing company. If they believe it's true, then all those people have been saying to me for the last couple of years, Keith, what you're doing doesn't make sense. Now I have to eat humble pie.
No, more than that. It's also based on the thing we've discussed on previous shows, which is Destiny's Tech 100 index having listed on the New York Stock Exchange and doing very well, trading at four times the underlying asset value, which indicates that investors like private company indexes. That was the second thing.
Yeah, it's a good question, Andrew, because there's a lot of subtlety here. The first thing is to acknowledge that private markets typically are not indexed. What does that mean? It means that private markets are where venture capital and private equity plays. And typically, you have firms picking single companies to back. And an index is the opposite of that. An index is when you buy all the companies in a space to create, to, if you will, capture the average performance of the entire asset class. So the S&P 500 would be the one most people would be familiar with. It's 500 companies. When you buy the S&P 500 index, you get a little piece of each of those 500 companies weighted by the companies. And in private markets, that hasn't existed. The reason it hasn't existed is nobody can invest in every venture-backed company in an asset class, for example, every seed company or every A-round company or every B-round company. So up until now, indexing has not really been possible. With the rise of algorithms, it's now possible at least to claim that you can pick a subset of the companies that will outperform the average and that that subset takes on the form of an index. You invest in every company in the subset picked by the algorithm. So there's a great article this week by Jared from Rebel Fund. He picks Y Combinator companies. and he shows the stats to show that the algorithms he uses to do that outperform the average Y Combinator company. Now, he is not indexing in the sense that he's not going to list publicly and let you and me buy shares, but he kind of is indexing when it comes to using algorithms to select a subset that will outperform the average.
Okay, so I get your point, but coming back to the bubble issue, the rise of the algorithms you're describing in this week's, uh, newsletter, the private market indexes is just, uh, in a sense is the froth on the AI bubble. It's not proof that the, there is, or there is, or there isn't an AI bubble. Is that fair?
Yeah. It doesn't really address that narrative at all. Um, because I will argue that, um, BlackRock's $3 billion. It's buying a company with revenue. And Prequin is not an AI company. But BlackRock is an AI company. And it will plan to use that data to enhance its AI selection process and build indexes from it. So it's fairly complex. I don't think it directly addresses the bubble question. But that's a big check. So you could say that tangentially, at least, it does. SignalRank, by contrast, we find it super hard to raise money because the world doesn't yet understand how powerful indexing is, how powerful the algorithms are. So we struggle. I think we're going to struggle less in the aftermath of the last few weeks. But it's certainly early in the adoption curve of this idea of private market indexes driven by AI. and therefore no bubble there, at least not yet.
Well, let's go back to the bubble issue. A lot of the articles this week that you link to, to me at least, and maybe I'm a natural pessimist, but to me at least prove or certainly convince me more and more that we are living in an AI bubble. You You quote O'Malick. He's one of your favorite writers. He talks about the future, appropriately enough, of writing, how AI will shape or reshape our tools. And he's touting a product called pseudowrite.com. And I tried it out, and it was so bad that it was almost absurd. Now, maybe my reaction is anecdotal, but how many of these companies, Keith, are just raising money because they claim to be AI companies. So in the writing space, it's companies like Pseudo, right? Maybe they're better than my experience was. But I tried them out, and it just seemed to be so bad, so much worse than what I myself could write.
Yeah. So let's separate a few things out. I mean, firstly, I don't think Home is especially touting those companies. Lex as well is one he mentioned. I think he's just using examples. And more important in what he said is this concept of building AI tools into writing tools. He wrote the piece because you'll remember WordPress. I certainly do. The founder of WordPress passed away last week.
Well, so Grammarly is a word and sentence checker, a meaning checker. It's got a thesaurus. It's very good. It's nowhere near as good as OpenAI or Claude. And if I copy and paste stuff into OpenAI or Claude, it will do a better job than Grammarly. Now, OpenAI and Claude are not yet built into any word processor. If they were...
I think I'd be interested to see what would happen. I think something good would happen as long as you were writing and it was helping and you weren't kind of trying to delegate your writing to an AI because I don't think the AIs are good enough yet to do writing better than humans.
I mean, they're so far away. One of the other articles you linked to this week is another one about boom times or the bubble from Spyglass, lowering the boom on the new boom times, and it talks about how much money is flowing. How much of the current stock market boom, Keith, is being driven by AI optimism?
in particular. And it's more than a lot. What numbers are we talking about? I mean, if you took Nvidia, Google and Microsoft out of the NASDAQ or indeed the broader economy, you wouldn't have a crash, but you would certainly have a severe recession.
Well, you actually have had one. It's just disguised by those companies having gone in the opposite direction. Most companies have lost value in the last couple of years. So the same with the S&P 500. Apple is a huge part of that, for example. So I do think the index is, and that's, by the way, the beauty of an index. Lots of companies can be losing, but the index still gains if it's led by winners.
It doesn't work. Indexing didn't work during the Wall Street crash. I mean, the economy has to be underpinned by something real, with or without indexing, with or without algorithmic indexing. This has to be real. And at the moment, it just strikes me as being increasingly fictional.
Well, I think that's because you don't pay attention. I mean, if you look at the GDP growth in the US, it's very healthy. So there is something real underneath it.
That's a fair point. I accept that. But coming back to how much of the GDP growth in the US is being driven by AI? Well, do you think Tesla is AI? Well, Tesla's a strange word.
It's part of the GDP. It's the biggest car sales in the world, and it's American. I do think that the answer is yes. AI is on the production lines. AI is in customer outreach, reducing costs. I think AI is Very early, by the way. I mean, one would want to overstate the impact yet. I think the future impact is a lot bigger than the current. But yeah, AI is there. And you also have to not, don't separate the idea of an index from a company. An index is made up of companies and companies have to have profits. And if they don't have profits and aren't growing, their stock price won't go up. And so the index won't go up either. So there is a, you can trace a line from an index to lithium being mined. It is real underneath. And there's no way an index can outperform real. I agree with you on that. You're entirely right. But the world economy only really grows when two things happen. One is productivity goes up. And two, consumption doesn't fall. And it seems to me right now productivity is going up and consumption is not falling, at least in the States. There are some parts of the world where it is. I think there are some dangers. The biggest danger isn't AI. The biggest danger is aging and whether the declining working population can be compensated for by automation. Thank you.
And you had one interesting piece in this week's newsletter from the CTO of OpenAI acknowledging that creative jobs can be harmed by AI. Although, again, given how bad a product like PseudoWrite is, I'm not actually sure that's true.
Yeah, I wouldn't judge AI by PseudoWrite. I would probably judge it... Look, I don't want to be all... I know a lot about AI. And I know enough to know that it isn't finished yet. And sometimes it's disappointing.
But it's not just let's leave the we pummeled or I pummeled pseudo right. Probably unfairly. But there are lots of other companies. You have a link this week, Keith, to an end gadget piece about now you can get. AI Judy Garland or James Dean to read you the news through 11 Labs, which actually isn't a bad product, much better than pseudo, right? But I'm wondering, is there really a business model? Are people really going to pay to get AI Judy Garland or James Dean or whoever to read you the news? It just seems it's like we're back in the 90s and there are all these people desperate for a business model using technology to chase it.
Yeah, yeah. I put it in because it made me smile. It made me think that the product manager at Eleven Labs, which, by the way, is a great company technology-wise.
They would have qualified for the single rank index, but we didn't invest because we didn't get access to it. But I was going to say, he must, or she, must be my age. Because why would you choose Julie Garland? I mean, even older than me, actually.
Why would you choose Julie Garland and James Dean, which most young people who might want to have a book read to them have never heard of? It should be P. Diddy and, you know, Tracy Almond.
Whether it's P. Diddy or Tracy Almond or... Judy Garland or James Dean or Marlon Brando. Is there really a market for these sorts of things? People are going to pay for this. I mean, it's all fun. And it was always fun in the 90s. We both did startups back then. All sorts of creative ideas were thrown out. Very few of them actually made money. There wasn't really much of a business model for most of them.
yeah, if you copy and paste text, it will, and I give you access to my voice, because it's, I own my own voice, but if I give you access to my voice, yeah, you can, you could do, read the, you could make, you could write something, let me speak it, it's only audio though, no video, so what we really get to, we'll get to it later in the show, but it's time to be the case that you can do the same with video and voices. And you've probably seen some of those.
Right. And then, and again, and I keep on saying basically the same thing, but your startup of the week is a company called runway, which, um, apparently, according to some online media, is in talks with General Atlantic for a $4 billion valuation fundraise. So having seen that, I went, I think I may have even emailed you the link in the first place. So I signed up for Runway. And again, it's, I mean, maybe not be quite as bad as pseudo, right? But it's not that impressive.
I looked at their front page video, which supposedly is their strength. I didn't use it at all. And it was just this music and lots of weird video that didn't strike me as being particularly original or worthy of payment.
I asked it to make me a scene of people serenading on the Scarborough seafront. That's the town I was born in. And... I didn't give it a picture of Scarborough, and it produced its own video with some weird hybrid Scarborough. But it was good enough that I recognized that it was Scarborough, and it wasn't that bad.
Well, not that bad, and yet General Atlantic believes it's a $4 billion company. I actually did... play around with it. On our street in San Francisco, it's a narrow street, a bus got stuck last weekend. So we took some photos and I put the photo into Runway and said, make me a video showing what happened and explaining that it was the driver who screwed up because the bus took the wrong turn, shouldn't have been on our street in the first place. And the video was just an utter mess. It didn't understand what I was asking. And the product was worthless. Less than worthless.
The demos in the piece accompanying the new version claim that it's better than Sora. Sora is already very good, although it only produces one-minute videos. I don't know how long these ones are.
Yes. But if you ask the question, why would a sophisticated shop like General Atlantic, which is a super intelligent investor... Why would they do this? Then you've got to, let's start by assuming that they're not stupid. So, which they may well be stupid, but I'm going to guess they're not.
But they weren't any less stupid in the 90s when, it may not have been General Atlantic, but when lots of highly credible VCs poured ridiculous amounts of money into crappy startups.
Well, if you want to go back to the 20s, I mean, maybe this would or wouldn't create a stock market crisis. But if, as I suggested earlier, the big tech companies like NVIDIA and Google and Microsoft are being radically overvalued because of this AI boom, ultimately the correction won't just affect the startup market. environment, but everybody.
I've always said it's a bubble, but I think it's a good thing. I think bubbles are good. Bubbles indicate opportunity. If there's no bubble, then there's no excitement.
So was the tulip boom in a bubble? Was that a good bubble? Or was the bubble in the 20s around the American automotive industry, was that a good bubble?
If you take the really long term, it's the Keynes argument or joke that in the long term, we're all dead anyway. But the automotive one, Keith, sure, in the long run, probably you've got the roads and the railways. But leaving aside the railways, let's say the automotive... bubble in the in the 20s which resulted in the great crash um i'm not saying that the 2020s are the same but there are some ominous eerie similarities so the car industry was good and ultimately car technology radically revolutionized everything about america i still had to go through the great depression yeah but i i don't
think it's it's not intellectually honest to focus on a crash unless you draw the whole graph before and after. And if you look at the stock market over any period of time, really, it only goes up. So yes, it goes up like this. And when it goes down, it's a crash. And we recently have seen three of them, right? You and I lived through the internet bubble. We lived through 2008. And there's one happening now. Yes, there are crashes, but those crashes are crashes in the context of human progress.
I mean, that's a tautology. Coming back to the current situation, it's not equivalent to the 20s because there's many more protections against a stock market crash. But what's the worst that can happen, do you think, in the current environment? I mean, I know... the Google numbers because my wife works for them. I mean, their stock has gone up almost 200% in the last couple of years. And it's basically the same company.
Yeah, but if you do the math on the growth in revenue... and profit, and you look at the multiplier effect, how much does your stock gain for a certain amount of profit? I don't think it's disconnected and got out of whack.
So what's the worst that could happen if there is a really, really sharp correction later this year or next year in the stock market, in the public stock market?
Which has happened in the past. I mean, you know, that means a million dollars becomes half a million, right? So I don't think any of these things are life-threatening. I think it's part of capitalism. I'm not really sure what you want rather than this. But to me, it's part of capitalism that you speculate to accumulate. I'm okay with that. There's many things about capitalism I don't like. But speculating to accumulate, I'm okay with. And venture capital... is where the most extreme speculation happens because that's how it's designed.
Let's go back to the algorithms and the AI revolution. There's clearly massive froth. What is real? Open AI is real. The Google and Microsoft applications are probably real. What else is real, Keith?
Well, I think I can tell you one thing that isn't real. It isn't real that lots of money is going into AI. That's a myth. Lots of money is going into these companies you just described as real. But actually, small AI companies, I find it almost impossible to raise money.
You mean the pseudo? Well, what about runway? That's why they're your startup of the week. They've just got a $4 billion valuation. God knows what their revenue is.
Then why put it in the newsletter if it's not a real piece? I mean, what kind of revenue would a company like Runway have? No one's paying for this. Actually,
It's really about risk and reward. It's what bed are you prepared to place on the likelihood that this thing will become a thousand times more valuable than it is today? And General Atlantic have a lot of money, so they can take a big risk. That $4 billion valuation, probably they're putting in $100 million. And they manage several tens of billions. So it's a smallish bet for them. And if it turns out that that $4 billion becomes worth $40 billion or even $400 billion, which is not too idealistic these days, then that will have proven to be a great bet.
Would you make that investment, though, in runway?
Words and timings
Wouldyoumakethatinvestment,though,inrunway?
Speaker 3
If I had General Atlantic's capital, I might. The valuation is an irrelevant data point in an early-stage company. The only real data point is what will be the future value and how fast will it get there. So you don't really care about the short-term valuation as long as you believe it's going to go up and go up fast.
Let me revise the question. Coming back to these, what I think at least, very unimpressive AI companies like PseudoWrite, maybe Runway, maybe 11 Labs. Is it conceivable that this whole AI economy won't take off? Is it conceivable that even the open AI CTO is wrong and that none of this is really going to make any difference? And in 20 years, you're still going to have writers and filmmakers and musicians basically using Similar tools that they're using today.
Yeah, I Find that hard to believe just because I use the stuff every day and there's enough That is very valuable that I don't believe that can ever go away now one it's like once you've seen it it can't go away and A lot of money is flowing to those good things now if you ask me another question, which is I is it going to be like the rest of venture capital in that 90% of these investments will fail? The answer is yes, absolutely. That's guaranteed. That's the whole point of indexing. 90% of these investments will fail, but the 10% that don't will carry the whole thing up so that the total amount of money made will be vastly bigger than the total amount of money.
The X of the week, I think, touches on it as well. It's from... Cloudflare to help preserve a safe internet for content creators, we've just launched a brand new easy button to block all AI bots. It's available for all customers, including those on our free tier. And that's T-I-E-R rather than T-E-A-R-E, Keith's tier. Read our blog post for more details. Does that show, Keith, I mean, maybe Cloudflare is doing all right. Does that show it's like back to the The gold rush of the 19th century, the only people really making money here in the 19th century was Levi's selling the miners their clothing and picks. These days, it's companies like Cloudflare and probably Nvidia.
Yeah, I mean, Cloudflare is a great example of some of the things we've been talking about. Cloudflare today is valued at $30 billion, $30 billion. Its annual revenue is $1 billion, 1.3. So it's valued at 30 times its revenue.
Matthew Prince and Michelle Zeitling, who were the two founders, I was their mentor at Sequoia Capital when they did TechCrunch Disrupt in, I'm going to guess it was 2012-ish, when they got their first investment and people said, how could you value that company?
I did invest, actually. Yeah, I did. They eventually IPO'd. Now they're worth $30 billion. And what they do is they run your network and they protect you from all kinds of bad stuff. So this is and they weren't very well understood because in those days people didn't really understood how networks worked. So they were very technical and it was hard to get people to understand them. But now they're huge. Now, what they've done is they've added blocking bots from using your site for AI training. and it's part of then services they offer if you use them. By the way, it's free for small companies.
So New York Times, I know they've got a lawsuit with OpenAI claiming they're stealing their content. If New York Times used Cloudflare software, then the OpenAI crawlers couldn't access their data. Is that right?
Correct. As long as they respect it. I mean, they could.
Words and timings
Correct.Aslongastheyrespectit.Imean,theycould.
Speaker 1
I mean, respect it.
Words and timings
Imean,respectit.
Speaker 3
Well, it's a set of rules. They're standardized. Technically speaking, you put a file on your server called robots.txt and you put rules in there. And everyone usually agrees to play by the rules, which is if it says, no, you, Mr. OpenAI cannot, you don't have permission, then the OpenAI crawler won't look. That's the way it works. I don't know if that's exactly what, Cloudflare have introduced. It sounds like they've added...
There's a real... I mean, to be fair, I mean, I've been rather negative today. Cloudflare's a real product. It's not like PseudoRite or Runway or Eleven Labs. It's not speculative. There's a real demand for what they have. But again, it goes back to the success of a company like Levi's in the 19th century selling clothing to the miners.
And you've often pooh-poohed this, but it seems as if the real interesting model for what's going on was the success of Cisco in the 90s, who, in a sense, I guess, it was a network company, much more powerful than Cloudflare, that was driving the 90s boom in the same way as NVIDIA now is driving the 2020s boom. And when you look now, I don't know what I don't know what they're doing now.
Not a lot. I mean, your startup's worth more than that, isn't it, Keith? I wish. That would be nice. But at one point, Cisco was the most valuable company in the world in the 90s.
Well, what's happened actually, Andrew, is the value of software has grown relative to the value of hardware. Cisco is a hardware company and the cloud has virtualized hardware so that the biggest companies in the world now are software companies because most money is spent on software, not on hardware. Hardware has become commoditized and sold You know, the service providers buy it and provide the cloud service.
A good example would be, well, not really. It's still that most of its revenue comes from hardware. But if I bought an NVIDIA GPU today for $30,000, which is roughly what they cost, if I can put it into a computer and run it, I can get that $30,000 back in three months. and the rest of the year are making a profit because the software running on it is more valuable than the hardware.
You sound like an NVIDIA salesman, Keith. We are having a break for the next four weeks because Keith and I need a break. It's summer. It's summer, and summer is not a time to worry about technology. But to end, Keith, because we won't talk again until August, a couple of questions. It'll be a month's time. What's the chances in a month that the entire world economy will have crashed?
Well, those are a reflection of my concerns. What about for you? I'm always asking you the questions. Any reverse questions? What should we be looking for in the next month? What's interesting?
Whereas in a month, that's long term. Isn't that a century, a month in Silicon Valley for most people? I think... Maybe not the summer, because everyone goes on holiday in July.
Well, that's how we'll start, Keith. In a month's time, we will address whether or not the global economy has crashed. It's not supposed to be a show about politics, but it will be interesting to see. My sense, by the way... is that Joe Biden won't be running this time next month. But we will see. Have a good break.