Krystal Ball

But I do also think that they are genuinely on the cusp of what's called recursive self-improvement.

The reason that matters is because the expectation is that once these models are able to train themselves, so you have the most advanced model training a new, more advanced model, and on and on and on, where human beings aren't really required anymore, and this can go for 24 hours a day, and experience exponential explosion of growth and capability,

that we could really be on the precipice of something that is quite different than the already extraordinary capability that Mythos and Fable 5 and the other leading frontier models have.

The expectation is that once you get to that recursive self-improvement,

that you can have an intelligence explosion that would lead not only to AGI, which is artificial general intelligence, which means the models can do basically every human cognitive task as well as a human, that you could quickly speed run that to super intelligence, where they can do, you know, genius level, they can far surpass even the smartest human in all cognitive fields.

We don't know what it's like to live in a world where we are no longer the peak of cognitive abilities, and there is basically no one who believes we have done the cognitive work required to anticipate all of the dangers and risks

of creating such a model could create.

So I think that should be making people very nervous.

And Sagar, I know you played a little bit with Fable 5.

I did as well before it got pulled.

I don't use it nearly as extensively as people who are like coding things with it.

I just did some queries with them and get a feel for this thing.

Does it feel different than the other cloud models?

And even just with a few sort of basic analyses and queries,

You could tell it was a significant leap forward from the previous Claude model, which already was quite advanced.

So I do think we may be on the precipice of something here that could be quite transformational.

I don't necessarily say that in a good way.

Yay.

For more on the entire AI landscape, including his views on Anthropic and the new SpaceX IPO, Elon Musk becoming the world's first trillionaire, we're very fortunate to have joining us this morning, Ed Zitrone.

He's a fantastic journalist, has been sounding the alarm about what he sees as a massive AI bubble, and is host of the Better Offline podcast.