Krystal Ball

Welcome to the show, Ed.

Great to have you.

Yeah, our pleasure.

So we were just discussing the forced, basically, pulling of Fable 5, which is Anthropic's latest model, based on Mythos, which they themselves saw as sort of too dangerous to release to the public.

Fable 5 was meant to be Mythos with some guardrails put in place for

Apparently, the Amazon CEO went to the Trump administration and said, oh, we figured out some ways to jailbreak this.

You should pull it from the public.

And through a national security directive, they effectuated that outcome.

I wonder what you make of this whole chain of events and what it says about where AI development is at this point.

So I think what they would say is that AI cannot do all the things that have been promised yet.

But we do see, I mean, this is a genuine progression in terms of technological capability.

And I'm far from the people who have used this to code extensively to be deep in the details of what it can do and what it can't do.

Certainly it hallucinates, although the hallucinations have been diminished.

But it's very able to go in and do a lot of what is, you know, entry level type work, spreadsheet jockeying type work, take a big data set, analyze it, put it in some sort of comprehensible format.

It does that relatively well.

It obviously does the coding relatively well.

We're now getting into the era of agentic AI that can actually go out and do stuff in the world remotely.

You know, the proponents argue that that is a significant development forward.

And then in terms of just sort of regurgitating pre-existing human knowledge, you did recently have AI solve, for example, a mathematical problem that humans had been unable to solve and trying to solve for 80 years.

So I do think that there is some significant technological progress here that may not live up to all the hype of we're going to replace every human, you know, all human endeavors and all human labor within the next five years.