Beth Lyons

as a strategy, right?

They're subsidized and it's in China's interest to lead the US companies, right?

I mean, to a certain extent, everybody's racing toward that specific goal.

That's fascinating because prior to this, the main thing that you were seeing

in terms of comparing cost was, is it worth the $200 a month lovable?

Or is it worth the $200 a month max subscription to Anthropic?

Or the pro subscription to OpenAI, which gives you those

those higher rates, right, those higher limits, so that you could conceivably use codex more, or you could, or you could use max at the, you get 20x more, 10x more, I don't remember what their, what their extreme max was.

That's fascinating.

And

Uh, yeah, no, that, that's, that's wild.

So now it's not, um, is this worth $200 versus a, uh, hiring somebody now is it worth a buck 30?

Yeah.

And you are no longer, did you put yourself on mute with your mic?

And there is that theory that is not fully playing out, but maybe playing out in China, that we'll

Will cost to train models come down if there is a corresponding knowledge that starts to increase about the kind of knowledge that is most important, right?

Again, we're sort of shoving everything we have, we.

Yes, me and all the research labs that are training models are showing all of the content possible to create and spending time, right?

It's the increase in data and increase in amount of time that can spend growing or training or any of those things that seems to lead to the increase in intelligence.

right what the model is capable of there's a little more secret sauce in terms of reasoning and those kinds of things but that seems to be it there is a theory that that it is likely that within that giant corpus of data there is data that's really important and a bunch of noise so if the noise if there's a system that allows you to identify noise and pull out those costs may come down but