Haseeb

This is not the first time that this has happened.

It's been well known actually that Opus, you know, Opus was previously, or sorry, the predecessor to Sonnet

which is the midsize model for Anthropic, that was at one point their strongest publicly released model.

And the reason why they did that was that they created a much more capable model that was very, very large and they distilled that down into a smaller model that's more economical for inference.

So it's not uncommon that the labs will do this.

They'll create these huge models and they'll distill them down into something that's more economical.

Exactly.

And that's mostly for economic reasons.

It's not because they're like, oh, this is too scary or powerful.

Maybe there's some of that.

But mostly it's just that it's bad business to do it that way.

It's just better for your margins to do so.

I think this is the first time that they've had this level of alarm in what a phase shift this is.

And this alarm is not, oh, you know, what if scammers use this?

Oh, what if people create deepfakes?

This alarm is...

This will definitely cause huge amounts of damage to the world if you allow this stuff to happen, to continue unperturbed.

And there were some tweets that I was reading from folks who I thought were very, very aptly stated, which is that this level of capabilities...

is stronger than the NSA.

It's stronger than Mossad.