Mm

...

when we look at the cosmos, at, at creation.

Uh, this world and, you know, I, I include in, in the whole cosmos.

But, uh, uh, so, so that's one of the functions of myth is to instill this sense of awe, and, and I think that happens when you're reading myth, when you're immersed in a particular tradition.

Um, there's a sense of delight in, in those stories.

Um, so that's one function.

Yeah.

Brad, why is it getting so expensive? I mean, are the tokens themselves more expensive, or are companies using more of them, or is it both?

I think in the beginning, companies were just wanting to encourage all their workers to use AI to try to think about how to do work and measure things that they could do better, and that was kinda like phase one, and now they're confronting the cost of widespread, uncontrolled experimentation and rethinking things that, you know, ended up being very costly.

As we know, something may start in a very well-meaning way, and then the incentives that exist can become kind of perverse, and if employees hear, "Use AI at all costs, and we're gonna grade you on your AI use," then all of a sudden employees can be like, "Sure, I'll use AI as much as you want," and then you're not really discerning between, like, a good use and a bad use.

I spoke to a startup founder named Matan Grinberg, and he runs a company called Factory, which automates coding.

One of the companies he was consulting with, which he just said was a financial institution, had racked up huge amounts of charges from, like, very simple queries, like the weather.

It's not a great use of tokens to spend a lot of model energy and computing energy being like, "What is AI?" You know, very basic queries like that.

So what are they doing to reduce their use of tokens?

Some of them have basically made certain advanced models and capabilities available only to a small subgroup.

Journal deputy tech editor Bradley Olson says that goes against the industry narrative that massive computing costs hurt profitability in the near term.

This growth is being driven by a lot of companies and businesses that are interested in Anthropic's AI systems and tools, and the ability of Anthropic's AI to do kind of automated work, to take long tasks and to work on them for hours.

And Anthropic's AI is able to do this.

Other ones are able to do it too, but Anthropic's has really gone viral since early this year, and a lot of companies are using it and trying it.

This as Anthropic, OpenAI, and SpaceX all race toward public listings that could value each company at over $1 trillion.

We understand that Anthropic would like to have a public listing before the end of this year, and, you know, they're also being compared to OpenAI and SpaceX.

You know, it was reported that both of them are having IPOs that are going to be sooner.

And so I think the big question that remains for Anthropic is: When is Anthropic going to do its IPO? What is the timing for that, and what's expected for that?

We just got this verdict, so it may take time before we know the full extent of this, but how has the trial impacted the AI industry at large?

Well, I think in the end, any time there's sort of a lot of information coming out and a dispute about who is responsible for something, right? Who made a thing happen? Who is the author of it? I think a lot of this trial ended up being about that, E- Elon kind of reminding everyone that he was there from the foundation of this thing and could even be considered a, a co-founder or someone who was in at the ground floor when OpenAI really became a thing, became a new company.

You know, the next step that we see is the kind of the public listings of these three main players.

You know, SpaceX, Elon Musk, his rocket company now owns his AI company, which was called xAI, and that's headed toward an IPO in less than a month.

And then you also have OpenAI potentially doing an IPO, and then its rival, Anthropic, also doing an IPO.

So there's gonna be kind of a public market value attached to all these companies very soon.

And so I think this trial kind of set the stage for that and is another marker of OpenAI kind of clearing another challenge or hurdle in its path to a public listing.

Do you think any of the witnesses regret their testimony or what came out in court?

Deputy Tech Bureau Chief Bradley Olson told our Tech News Briefing podcast that the outcome wasn't a surprise.

We did a story before the trial basically saying that Musk was an underdog in the case, and we even looked at prediction markets, which had kind of found that to be true, too.

And so yes, this was the outcome that was expected.

However, one of the most am- interesting things that came out, we've done some reporting at the Journal about Sam Altman's personal investments and examples when there were overlap in deals that OpenAI did with outside companies and Sam having personal investments in those companies.

And so there were a lot of details and specifics about those holdings that came out.

It was also revealed that Greg Brockman, OpenAI's president, holds a stake in OpenAI that's worth about $30 billion.

A lot of kind of disclosure and detail about their personal finances came out, and I think we will look back on it as something that litigated the foundational moments of the AI boom.

A decision against OpenAI could have created big problems for the company.

So I guess it's not all coming without some adjustments to the balance sheet.

Absolutely.

I think in the end, these companies, at least kind of by the rubrics of Wall Street, were always well-capitalized.

They had tremendous cash flow, and that's kind of changing.

In order to afford the kind of spending that they're going to have to make, they're changing the way they operate, and they are becoming kind of more indebted and using that to pay for AI infrastructure and data centers, and they're also cutting staff.

You know, you mentioned Meta.

That was one example, and Microsoft also offered buyouts, and they said it'll be another year in which their head count will go down.

That may be kind of these companies' saving grace.

So I guess it's not all coming without some adjustments to the balance sheet.

Absolutely.

I think in the end, these companies, at least kind of by the rubrics of Wall Street, were always well-capitalized.

They had tremendous cash flow, and that's kind of changing.

In order to afford the kind of spending that they're going to have to make, they're changing the way they operate, and they are becoming kind of more indebted and using that to pay for AI infrastructure and data centers, and they're also cutting staff.

You know, you mentioned Meta.

That was one example, and Microsoft also offered buyouts, and they said it'll be another year in which their headcount will go down.

That may be kind of these companies' saving grace.

And though Ternus has little experience with AI, our deputy tech bureau chief, Bradley Olson, says Apple isn't entirely out of the race.

Apple is basically the delivery system for AI and will be for the foreseeable future for billions of people, you know, that have iPhones and use Apple products, but it's behind its rivals in terms of building models that people will use.

And so, you know, how Apple thrives in the AI age is definitely an open question, and that's something that Ternus sort of inherits from Cook.

Cook's reputation as someone who took the iPhone, which was one of the most successful products of all time, and used it to build an empire, which Apple really is today.

And so Ternus will maybe have to step into that part of the role and the public-facing aspects of being an executive.

And I think with Cook staying on as executive chairman, maybe he'll have the space to do that.

And though Ternus has little experience with AI, our deputy tech bureau chief, Bradley Olson, says Apple isn't entirely out of the race.

Apple is basically the delivery system for AI and will be for the foreseeable future for billions of people, you know, that have iPhones and use Apple products, but it's behind its rivals in terms of building models that people will use.

And so, you know, how Apple thrives in the AI age is definitely an open question, and that's something that Ternus sort of inherits from Cook.

Cook's reputation as someone who took the iPhone, which was one of the most successful products of all time, and used it to build an empire, which Apple really is today.

And so Ternus will maybe have to step into that part of the role and the public-facing aspects of being an executive.

And I think with Cook staying on as executive chairman, maybe he'll have the space to do that.

What's different about what Claude Code can do? Why are engineers so excited?

excited?So Claude Code has existed for a while, and it's something that can write software for any software engineer, and people were always impressed with it.

And then in November, Anthropic releases the most updated model of Claude Opus four point five.

A lot of coders and software engineers are just blown away.

Like, they just are starting to say, "I can't believe that I just told it what to do, and then it produced kind of the workload that I might have done myself in a fairly limited period of time." All these software engineers are talking about it, and then people who aren't coders start to download it.

And not only do they try it for coding and making websites and apps and stuff like that, they also just start using it to do data analysis and other kinds of tasks.

And so it's like an awakening of capability that some people are comparing to the moment when ChatGPT was released.

Why is that?

Explain how this is different from, say, the dot-com boom.

Well, I think in the end, in the dot-com boom, there were a lot of people putting in a lot of time and a lot of work.

There have been previous boom times when I think these cycles have happened.

And frankly, at a lot of startups...

You know, a lot of people, Elon Musk famously, you know, sleeps at the office, you know, in crazy times at some of his companies.

So I think there's a ring of the familiar in some ways.

But of the people that I talked to said it was unique in its intensity more than the past cycles that they experienced.

In these AI labs, is it mostly young people or not necessarily young?