Jeff Dean

Sounds potentially good.

It's a good question because I don't know that folks haven't been doing that as much.

I mean, we definitely have lots of great ideas coming along.

Everyone seems to want to run their experiment at maximum scale, but I think that's a human problem.

Yeah, the one thing it doesn't speed up is, like, experiments at the largest scale, because you still end up doing, like, these n equals 1 experiments, and they're really just try to put a bunch of really brilliant people in the room and have them stare at the thing, figure out why this is working, why this is not working.

Yes, yes, we're counting on you.

Yeah, we don't need new hardware for that.

But I mean, we'll take it.

Yeah.

Yeah, exactly.

One of the big areas of improvement, I think, in the near future is this entrance time compute, like applying more compute

you know, at inference time.

And I guess the way I'd like to describe it is that, you know, like even some giant language model, you know, even if you're doing, say, a trillion operations per token, which is

you know, more than most people are doing these days.

You know, operations cost something like $10 to the negative $18.

And so you're getting like a million tokens to the dollar, right?

So, I mean, compare that to like a relatively cheap pastime.

Like you go out and you buy a paper book and read it, you're paying like 10,000 tokens to the dollar.

So it's so like talking to a language model could be like, you know, is like...

100 times cheaper than reading a paperback.