Shawn Wang: commitment

Transcript context

…to solve benchmarks first before solving long contacts. We have your benchmark. No, no, no, no, not like benchmarks for long contacts. OK, yeah. because the needle in haystack is basically like an MNIST, or like a unit test for these sort of things, right? But I think there's one But about, like, hitting the context line and the other part about, like, actually Utilizing. Utilizing, right. I think Gemini's long context is surely, like, amazing. Right, but I think, like, for the community to move forward in this, then it comes to a problem of, like, How do we evaluate this? I think I've seen some long context benchmarks, like, coding one, like, And stuff like that. Like, I think Making those are important, and for the community to heal crime, but I think long context is important, it's just that you don't have a very good way to measure them properly now, and yeah, I mean, I think long context is definitely the future, rather than RAC, but I mean, they could be used in conjunction. Definitely, okay. Yeah, yeah, yeah. That's a hot take. Which part of the Long context is the future rather than RAG. Like, you would, they will coexist, but you are very positive on long context. I will put myself on the other, so your mirror image, which is like, long context is good for prototyping, but any production system will just move to RAG. There are a lot of application use cases where you want a model to take the time and then come up with the right answer, right? Sure. Because RAG is like…

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