Jyunmi

So that's a bigger investigation to go on, you know, to ultimately be revealed on what happened there that made that happen.

Beth, do you know anything about that?

Well, this was an exploit, a major exploit.

That suggests that they're just using the model to see what it can generate as an inference in response to an input that they create.

And I don't know that that would be if that's a thing.

and you can actually generate a model that surpasses the training model, there's more secret sauce in there than I imagine is available.

So I'm skeptical about that one, but that approach being the explanation of why DeepSeek is being accused of purloining.

But in that term...

Okay.

Anyway.

I'd like to pivot to some AI and science.

I have a couple of interesting things I want to share.

Science.

First of all,

Yeah, go ahead.

Well, I hope I'm not stealing a science thunder, you know, for tomorrow.

Anyway, models are now solving problems that humans have not.

GPT-5.2 recently delivered a novel result in theoretical physics, proving that certain gluon tree-level amplitudes are non-zero.

OK, that's that's where that's the level of mathematics and physics that, you know, these models are operating at and some humans are operating as well.

I have no idea what that means.