Speaker 2
That was a week No time to be meek The goal is to seek The next big thing
Transcript Viewer
May 24, 2024 · 2024 #18. Read the transcript grouped by speaker, inspect word-level timecodes, and optionally turn subtitles on for direct video playback
Speaker Labels
Edit labels for this show, save them in this browser, or download a JSON override for the production folder.
Transcript Playback
Human Transcript
Blocks are grouped by speaker for readability. Expand a block to inspect word-level timing.
Speaker 2
That was a week No time to be meek The goal is to seek The next big thing
Speaker 3
Hello, everybody. It's Friday, May the 24th, 2024. The last Friday in June, the last That Was The Week with my old friend Keith Teer. the author of the invaluable essential, that was the week newsletter. And there's only one story this week I need to tell you. And it really is a story that has nothing to do, one would think, with technology. It's associated with Scarlett Johansson and the supposed appropriation of her voice by a certain Sam Altman from a little startup company called OpenAI. This is triggered keith to write a dear sam letter for that was the week newsletter um for may 24th keith i hear your wife thought it was all about keith rather than sam yeah
Speaker 1
we have a we have a vibrant let's call it debating culture in our family janae
Speaker 3
thought you were uh
Speaker 1
putting myself into the middle of something that isn't about me and making it about me.
Speaker 3
That was the... Keith Teer would never do that. He is the least self-centered person in the universe. Anyway, leaving aside Keith Teer, let's talk about this story. What exactly has happened? Why is this the dominant tech story of the week? It's bizarre, symbolic, but also hardly surprising in many ways.
Speaker 1
Well, I think the... The popular way of understanding it, which is probably somewhat mythical, is that Sam Altman is a bit of a sociopath, that he decided some time ago... Just a bit? Just a bit. And he decided some time ago that he wanted to associate the OpenAI voice with the movie Her.
Speaker 3
Which, of course, Scarlett Johansson was a voice character in.
Speaker 1
Yeah. So that's not a bad instinct. That's a really good marketing instinct, probably. Her was really good and somewhat prophetic because it does seem as if the current open AI that's being rolled out now is capable of some things that were depicted in that movie that we all thought were total science fiction at the time, and it wasn't that long ago. So good instinct. Scarlett Johansson was apparently contacted to act as the voice and she declined. But Sam didn't let go of the her idea. Now, it would be entirely okay to want to mirror her without purchasing Scarlett Johansson's voice, conceptually at least.
Speaker 3
Could you pay for Scarlett? Or is it one of those MasterCard ads that you can't? It's beyond money. Invaluable.
Speaker 1
Well, you know... If I was her, I wouldn't want to be the voice of OpenAI. It kind of typecasts you for the rest of your life in a way. So she's probably right to turn it down. And Sam's probably right to still think that her movie is a good framing for what the next version of OpenAI is.
Speaker 3
So when you say the voice of OpenAI, he didn't just want to have Scarlett Johansson as a voice on it.
Speaker 1
Yeah, that's all he wanted, yeah. so it's just one voice it's just a voice so he went off and hired an actress um scarlett johansson says to copy her voice the the actress and sam altman say no there was no remit to sound like scarlett johansson they did the voice they demoed the voice last week when they did their launch or two weeks ago And Scarlett Johansson immediately reacted saying, this is a copy of my voice. And the actress reacted, no, this is me. It's my normal voice. And there was no request for me to sound like Scarlett Johansson.
Speaker 3
So the question I have on this, Keith, is that OpenAI is an AI company. It's not a text-to-speech company. So it's really what was happening here is not associated with...
Speaker 1
No, it is. It is. You missed last week's newsletter. It was called Eyes, Ears, Hands, and Mouth. And actually, the new OpenAI is designed to be conversational using your voice, and it replies with a voice. And by the way, it can interrupt. You can talk over each other. Like you and I. Just like you and I. So it is evolving to be a voice interface. and you can also type into it but voice is really well done and um and that so that's the guts of it but it's more than a voice it uses the camera now andrew as eyes so you can like i pointed my camera a tree in my yard and said what is this tree and it spoke back oh that that is um uh california and blah blah tree uh uh pretty much instantly so this is gpt
Speaker 3
yeah yeah i just put it up on screen so that we have it yeah so it's so oh it's a that's a pretty big deal so they were using the the scar or they were apparently
Speaker 1
according to scarlet at least using her voice to yeah they they denied it hence the the uh the sam altman is a sociopath kind of yeah one one of the pieces you connect
Speaker 3
with is sam altman is full of and And then there's also a defense of what he did in Ars Technica from a piece in the Washington Post. It's a he said versus she said. There's always going to be arguments on this. What did you say, Keith? Let's annoy Jeanne. What is the point of your Dear Sam newsletter? What are you trying to tell Sam as a startup person yourself?
Speaker 1
yeah my letter may be too subtle because what i think i i say is sam you need to get ahead of this tell the truth and put it behind you because if you don't it's going to hang over you and you're sitting on top of probably some of the most important technology for the next 10 years so you don't want to screw it up based on this what my wife thinks i say is poor sam you've been cruelly
Speaker 3
uh accused of we're not going to have any more stories keith of what happened with you and microsoft you always manage some readers when keith gilmore always talks about paul mccartney or john lennon you always talk about that one so let's leave
Speaker 1
that one aside so i will say readers read it for yourself and yeah readers read it
Speaker 3
for yourself keith keith and sam altman are both
Speaker 1
Andrew, before you move on, someone in the chat says, Sam A., a sociopathic, narcissistic, egomaniac. So I'm right in kind of capturing what people think of him.
Speaker 3
Yeah, and I think if we want to capture... Maybe again, we always talk about this, the zeitgeist. This wouldn't have happened 5, 10, 15 years ago if Zuckerberg had made a mistake or the Google boys had made a mistake. So everyone, I'm no great fan of Altman or or even ai but everyone's out to get him so anything that happens people pile in i
Speaker 1
think yeah you know for you and i have both done startups i think if you think about this from the point of view of a startup and even though they've got a lot of money they are still a startup this this uh launch was the day before google io on purpose as a kind of a a bomb and um it was rushed The tech really isn't ready. So if you look at what they've rolled out in the two weeks... Which is classic for a start-up. Yeah, it isn't ready.
Speaker 3
The tech is never ready.
Speaker 1
Nothing was ready, including any agreement with Scarlett Johansson.
Speaker 3
And when it is ready, then you've moved on to the next thing.
Speaker 1
You've moved on to the next thing. So I think this is probably the cock-up theory applies, not the conspiracy theory. And I say that to cinematographers who made the comment. I think Sam A. is, you know, got a decent ego. There's no doubt about that.
Speaker 3
Oh, yeah. And he's clearly not the most honest of people. I mean, even your old friend Paul Graham has said that. So within the tech community, he's not trusted.
Speaker 1
As the board of directors said, he can be evasive.
Speaker 3
In other words, he lies. Casey Newton made an interesting point. We can't include his stuff in here because he's so... uh hard to pin down in terms of his stuff but he said that ever since altman came back to open ai neither altman nor the company has been the same thing is there any
Speaker 1
truth to that do you think well look there's there is a bad smell a lot of people have left um A lot of people have left, mainly the safety people. Although, to be honest, I think that's appropriate. I think there's almost no danger.
Speaker 3
Some of the early people left as well. So you don't think this bad smell, you say? I mean, a few months ago, I can't remember what the exact numbers were, but you suggested that the future of the entire global economy was built around open AI and it was going to dominate everything. Are you more or less bullish? More bullish. Not on AI itself, but on open AI.
Speaker 1
More bullish. I mean, what they launched last week was a little bit better in terms of the quality and the hallucinations and all that. But it was a lot better on the user interface and the use cases. So they integrated live Google Docs and live Microsoft Docs. so you can now give it documents to look at and interact with.
Speaker 1
It's a monster from a point of view of a future value.
Speaker 3
The issue, though, and this comes back to the Hansen case, which is about ownership of data and business models here. The other piece of news which got buried a little bit, which I think is extremely interesting, is that OpenAI And News Corp, owned by the Murdochs, signed a content deal. Now, News Corp has always been very hostile to big tech. So this is quite surprising. And this reflects well on the promise of open AI, because presumably if they can do deals with enough people in the media, in Hollywood, then none of these issues will come up in the future.
Speaker 1
Yeah, I think one of the things that's shifted in the last week is sources of data for open AI. Paying for data is one, but once you turn on the camera on the phone as a data gatherer, you have 4 billion people that can give you data. By the way, it works even by pointing the camera at a newspaper article. It will read it and tell you what it's about. Maybe that's what we should reinvent now.tv as.
Speaker 3
As what? As those 4 billion. You're always good at appropriating, writing one off other people, Keith.
Speaker 1
For them, yeah, yeah. Anyway, data is not going to be OpenAI's problem because we're all going to become data gatherers for it.
Speaker 3
Yeah, but this comes back to an issue that you and I have argued over for years. I mean, it's one thing to turn everybody's camera on. It's quite another to figure out reliable news from credible journalists at places like the journal and the times. I mean, the times is still suing open AI. So yeah, a lot of this stuff's going to get determined in the courts over the next year or two.
Speaker 1
Yeah. You know, we've talked about it before and I've predicted the open. I will win those court cases. I mean, how do you know? I don't have to know. I can predict. I am the voice of him.
Speaker 3
Yeah, but you're not the legal voice of him. You don't think that the Times have a good legal case? It's going to go to the Supreme Court.
Speaker 1
No, because I don't think OpenAI used any logged-in data. It was all publicly available, just like it would be to you and me or anyone writing a book. So I think they'll win based on fair use. I doubt they'll... But I think the world will move on. I don't think the Times is threatened, actually. I think to your point- Why are they going to court then? I think they're reacting against something new that they don't understand yet. And so the first instinct of a corporation is lawyers. But the second instinct is better than that. And there will be a second instinct.
Speaker 3
Of the content verticals from the news industry to the publishing business, to Hollywood, to the music industry, we'll come to the music industry later because your startup of the week is an AI music company. Which do you think are the most vulnerable and which are going to do the which you're going to throw in there a lot with the open AIs of the world.
Speaker 1
I actually think they'll all benefit. Open AI has started giving links and references now, as has Claude. And I think what will happen is a lot of traffic that was on Google doing searches will now be in open AI. And it will behoove most publishers to allow their links to be exposed so that people can click on them and go to their content. And I do still think human beings are going to be the determining factor when it comes to quality. You know, when I read something from OpenAI that I don't think is good, That damages up an AI in my eyes. So I do think we're all going to be judges and we're going to start using the AIs that serve up quality. And at least in my experience, they're capable of doing that.
Speaker 3
Well, this will be a dominant story. We'll come back to it. A couple of other interesting pieces of news about whether or not the AI industry is basically all of tech now. Nvidia crushed earnings again. Does that speak of the reality of this market? What does that tell us, the Nvidia numbers this week, their quarterly numbers?
Speaker 1
It tells us that people are buying the Nvidia cloud GPU services in order to train models. They made $26 billion in a single quarter in revenues. That is a 250% gain over the previous year. So these numbers are unusually large numbers for any company ever in the history of companies. So it really does tell us that lots and lots of money is being piled into
Speaker 3
these tools is there um i mean it's obviously good news for nvidia i assume it's good news for open ai and the downstream open ai companies is it bad news for microsoft who who's nvidia's most direct competitor actually microsoft is a big
Speaker 1
customer of nvidia so intel missed the boat on this one intel missed the boat amd missed the boat um you know google and facebook are building their own chips to not have to use NVIDIA, but that probably is a process that's going to take a while. And there'll be new chip companies. Now that the world needs fast model-building chips, the NVIDIA chip is the best right now, but it wasn't designed for this purpose. It was designed for games.
Speaker 3
Yeah, and one thing is I'm sure that Chinese have some good chip companies, which maybe they won't or will export to the US and vice versa. NVIDIA doesn't care about AI gross margins and whether or not the broader economy is viable. But you do have one interesting link about whether AI has a gross margin problem. What does that mean, Keith? And what's the issue here?
Speaker 1
Well, so the profit out of any business comes from the difference between the cost of sales and the revenue. That's called the gross margin. What does it cost you to earn a dollar? And if the answer is more than a dollar, you have a negative gross margin. If the answer is 10 cents, you have a 90% gross margin. in the case of um nvidia its customers for every dollar they spend with nvidia they make five so there's a healthy gross margin in buying nvidia services right i mean
Speaker 3
it's still not clear what open ai's business model is how does that work well it's
Speaker 1
subscription based they basically charge two two different ways Firstly, they've made their core tech free for consumers. You mean NVIDIA or OpenAI? OpenAI. You can now use the best model on OpenAI for free without paying them $20 a month. The $20 a month is no longer targeting consumers.
Speaker 1
So it's targeting companies. They call it teams. I've got one for single rank. And secondly, developers who use the APIs, they build up an eye into their apps. It's a classic startup.
Speaker 3
I mean, they're not profitable. It's not a public company.
Speaker 1
Well, their revenues are up to $2 billion a year now already. Yeah, but they're spending more. So coming back to this gross margin problem, is that the issue that... Well, no, gross margin does not include expenses. In accounting terms, it's called above the line and below the line. It means what is a direct cost of sales or cost of goods? And then below the line is expenses.
Speaker 3
I understand. So what's the issue here? Why is there an issue with... potentially AI having a gross margin probably.
Speaker 1
Because it's possible that over time the costs of scaling outweigh the revenues generated by the scaling.
Speaker 3
Which is a very sort of abstract.
Speaker 1
It's like WeWork. Basically, if you think WeWork, that their costs were way above their revenues because they were in buildings. Well, server farms and buildings are roughly correlated.
Speaker 3
But there's no Moore's law when it comes to commercial real estate. There is Moore's law when it comes to AI, isn't there?
Speaker 1
I agree. So I think the answer to that question is no. but it's a reasonably good question to ask. By the way, one of the questions in the chat here is asking about another one of the articles, which is about California.
Speaker 3
Yeah, and I'm moving on to that, and that wasn't from me, so that was a nice segue. We have an interesting piece about the – and, Keith, you and I are both in Northern California – the awful state of AI in California. What's so awful about AI in California?
Speaker 1
So what's happened is that the – the ever imaginative California Senate, Congress, whatever it's called, has decided to try to regulate AI in California by forcing people with models to register and pass muster, if you will, with the state, which means that California has become the most regulatory heavy place on earth to do AI.
Speaker 3
More regulatory heavy than Denmark or Germany?
Speaker 1
More, because it's actually now legislated. um and it it actually includes as the as our comment here cinematographers says uh it includes criminal liability so what's going to happen is it's going to force ai companies to be not domiciled in california and that's a no-brainer i i i wouldn't start a california ai company anymore so it isn't going to change regulations for ai it's going to change how much money california gets in taxes
Speaker 3
yeah interesting so we'll have another flight to miami or austin somewhere i don't know where well um your startup of the week surprise surprise another ai company one that looks very promising uh suno uh make a song about anything? Is this the radical, to borrow some tech Silicon Valley language, radical democratization of technology to make music?
Speaker 1
It's definitely getting there. There's two companies doing it. I can't remember the name of the other one. And this is the one I used to create the That Was The Week theme song that I now put on the videos every week when I put them into the newsletter.
Speaker 3
Oh, I need a key none.
Speaker 1
Yeah, you can make out some lyrics, tell it the musical style, and it will produce a song for you. It's really good. What are the lyrics?
Speaker 3
I'm Keith and it's all about Keith.
Speaker 1
That would be Jeunet's contribution. It's all about me.
Speaker 3
because i'm so important so is this is this suno stuff good it's very good it
Speaker 1
really is good and what's their business model you have to pay you you pay it's free for a certain number of songs and then you pay above that you buy credits and
Speaker 3
use them up and it's we're still in the early stages you've got light speed investing yeah
Speaker 1
It's not that early. I think the valuation will be up there now. I didn't look at what the exact valuation is, but this is a Series B round, I think.
Speaker 3
So presumably there are equivalents of Suno in film. I mean, that comes back to the Johansson case and in text. There will be. There definitely will be. Photography as well. I mean, Adobe is in trouble, aren't they?
Speaker 1
Well, Adobe is doing super well compared to its history, but its tool set is dated.
Speaker 3
Yeah, I mean, they have to. I mean, it's a question of cannibalizing that. What do they do? Do they turn themselves into a high-end company for humans or do they integrate these tools? I guess they've got to try and do both.
Speaker 1
Well, they're definitely integrating AI into their photography tools. Lightroom this week was upgraded with AI for some what I think are fairly minor behaviors. Their video tool doesn't really have it yet, but neither do any of the others. So I think that's still coming. The OpenAI demo where they made movies, Somebody's now made a three-minute movie for a song using it, and it's not the best movie ever, but it's fine. So I do think that the ability to compose is going to be available to a lot more people, and quality will cut in who's good and who isn't good.
Speaker 3
Yeah, but for you and I to find music and lyrics for the beginning of our show, we're not going to go to a, a top lyric writer. We don't have the resources or the time or the interest. It's not worth it for us.
Speaker 1
No, but it probably still saves us somewhere between $1,000 and $5,000, which you even pay a normal person. And I did mine for free compared to paying a thousand. You mean from Sooner? Yeah.
Speaker 3
So you used to steal the music from the three tops.
Speaker 1
No, but it goes back to the Apple ad where the iPad was crushing all the old stuff. We're in an era of transformation. I mean, there's just no getting away from it. And you can try as you might to cling on to skills that in the past required excellent human beings who've been trained in the skills. But that is going away. And so you need to retrain. And these tools are tools. If you retrain using these tools, if that's...
Speaker 3
And then that was your friend Hamish's point, the age of the sovereign creator. You didn't put it in the newsletter, but you might have. He uses your... Your friend at Stratatury, Ben Thompson, as the model. We're almost at a point, and one wonders what the value of Substack will be in this, where we'll have all the tools to create high-quality television, movies, songs, without the studios. I mean, it's the old promise of web, too.
Speaker 1
Yeah, but the one thing AI doesn't do is it doesn't replace the imagination or the creativity of the narrative audience. uh the narrative inventor so the the the core human skill is still going to be doing something that people want to spend time engaging with and i don't think ai is going to mostly be as good as humans at doing that ever so i i do i do think um forever is a long time keith yeah and and and you know Our commenter is disagreeing with me and saying that the Apple ad removed the emotional connection of history. I don't agree with that. I think it embraced. It's just that, unfortunately, it uses squashing metaphor. But the squashing was producing something new. It wasn't just squashing. It was transforming.
Speaker 3
It was squeezing when the eyeball popped out. That turned me off my dinner. Finally, and this comes back perhaps to... All the themes we've talked about. Dear Sam, trying to make him more grown up, perhaps more moral. And you've changed this. It's no longer the X of the week.
Speaker 1
You're showing strategy. Oh, there you go.
Speaker 3
It's the Warpcast. What is Warpcast?
Speaker 1
Well, I did it only for this week. Warpcast is a blockchain-based version of Twitter that... that fred wilson from union square ventures has embraced he was one of the first investors in twitter a couple of weeks ago he announced that he's using warpcast warp cast is is weird for any normal person because you you have to have a wallet a crypto wallet in order to post um it's weird to set up uh fred is a champion of it because he invested in it and this week he he did a post on uh avc which is also by the way
Speaker 3
um hosted on the blockchain now and it's called be generous so what fred wilson does and i've had lots of conversations with him debates over the years on this one is he says the web one and the web two entrepreneurs were self-centered self-interested but now things might be different is that really what he's saying
Speaker 1
he's it's interesting firstly oh my god how could he how could he say that without blushing because fred is nothing short of... My wife would think that he and I belong together in terms of our self-centeredness. So Fred is definitely... Yeah,
Speaker 3
and it's ironic that he's saying this while broadcasting on Warpcast, which is one of his own investments.
Speaker 1
Exactly. So I don't know if generosity is at the core there. I think it's making money for his LPs, which it should be. But what he's saying is Web3... You benefit by being generous because the more you give, the more you help other people make money, the more money you will make yourself. And he gives the example of Satoshi who created Bitcoin and then just literally gave it away to the world. And most of the value of Bitcoin is in the hands of others.
Speaker 3
But that was true of Tim Berners-Lee as well. There's always going to be exceptionally generous people who, for one reason or other, give their stuff away.
Speaker 1
And he's saying in Web3, that's the model, that if you don't do that, you won't actually make money yourself. So is that true?
Speaker 3
Do you agree when it comes to venture capital? I mean, it seems as if Fred Wilson is the classic venture capitalist. He's eating his own dog food.
Speaker 1
I think he's got a kernel of a point here that's important to understand if you want to understand the structural difference between Web2 and Web3. Because of tokenization and how value flows, you need to get value into the ecosystem in order to derive any value back. So he's basically saying that in the same way the internet gave a lot of value to everyone who uses it, trillions of dollars of value from something that costs nothing other than time and ingenuity, that Web3 is a modern version of that.
Speaker 3
It's interesting. Maybe we can come to this next week in future shows. I mean, the fact that Fred Wilson is even using the term Web3, the end, by definition, of walled gardens. And how does that fit into AI? It seems as if companies like Sumo, which is your startup of the week. I mean, I don't know what you'd call them. They're not Web3 companies. You have to pay to use their technology.
Speaker 1
No, they're all centralized Web2-style centralized companies with AI. And even OpenAI is a Web2 company. Totally Web2.
Speaker 3
So where's Fred's Web3?
Speaker 1
I think it's possible that Fred is – I hate to say this, Fred, but it's possible Fred is wrong, that he's backing – you know, the equivalent of the 1990s interactive TV show. Oh, my God.
Speaker 3
You're accusing Fred of being wrong. I know, Keith, I got an idea. Next week, rather than Dear Sam, the newsletter can be called Dear Fred. How about that?
Speaker 1
Yeah, no, not going there.
Speaker 2
That was a week No time to be meek The goal is to seek The next big thing