Shawn Wang: uncertainty

Transcript context

…Interesting. Because, because you basically have shown proof of concept that a non O1 model can beat O1. And their whole L1 marketing is, don't bother trying. Like, don't bother stitching together multiple chain of thought calls. We did something special, secret sauce, you don't know anything about it. And somehow, you know, your 4. 0 chain of thought reasoning as a software engineer is still better. Maybe it doesn't last. Maybe they're going to run L1 for five hours instead of five minutes, and then suddenly it works. So, I don't know. It's hard to know. I mean, one of the things that we just want to do out of sheer curiosity is do something like fine tune 405B on the same dataset. Like, same context window length, right? So, it should be fairly easy. We haven't done it yet. Truthfully, we have been so swamped with the waitlist, shipping product, you know, dev day, like, you know, onboarding customers from our waitlist. All these different things have gotten in the way, but it is definitely something out of more curiosity than anything else I'd like to try out. But also It opens up a new vector of like, if someone has a VPC where they can't deploy an OpenAI model, but they might be able to deploy an open source model, it opens that up for us as well from a customer perspective. So it'll probably be quite useful. I'd be very keen to see what the results are though. I suspect the answer is yes,…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.