“Interesting. So you feel like the— so yeah, I guess how would you characterize the gap between American API models, Claude perhaps specifically, and a DeepSeek? We hear, we're, I think we kind of go through these cycles, right? Where it's like, oh, the gap is closed. Oh, maybe...”

ConceptResearch

“Hop. Fine-tuning as a dimension in this whole model situation. This is another thing, if I go back in time, I'm like, I definitely expected a lot more fine-tuning than we are getting.”

ConceptResearch

“I think the LoRA thing is going to happen soon in the next 6 months. I think even like one of the frontier labs may do it.”

ConceptResearch

“AI in the benchmarks tends to be very different than AI in the real world”

ConceptResearch

“I take any release benchmarks not just with a grain of salt, but with an entire bowlful.”

ConceptResearch

“there's the risk that this is just benchmark maxing and actual performance will feel a little flat.”

ConceptResearch

“what does it even mean to be a frontier lab if DeepMind is not a frontier lab? I would have just assumed that they are, but like, what does that even mean? I don't know.”

ConceptResearch

“They're trying to argue distillation, though. Yeah, they're— because they see similarities model to model.”

ConceptResearch

“you do not need a lot of data if you— the novelty of the paper in this work, how to incorporate everything into a reinforcement learning framework that would make it possible and training, fine-tuning such models would be also kind of very affordable.”

ConceptResearch

“It's exciting to see it being incorporated into practical algorithms, training algorithms nowadays.”

ConceptResearch

“You know, Sam, which something I like to also add, because it is related to what we are talking about now. Optimization, of course, is right objective optimization at the right data is definitely very critical. That is what this paper is talking about, but we may also kind of ...”

ConceptResearch

“Scientific problems have a vast breadth, and I don't think current AI models have as much coverage as many think.”

ConceptResearch

“I see. So the tokens go in, the transformer establishes the relationship between the different tokens, and then the diffusion model turns that represent— that latent representation into a 3D structure.”

ConceptResearch

“people kind of, like, gamed these benchmarks for a while”

ConceptResearch

“we're getting to the point now where these structure prediction models are within, you know, a few angstroms or less of the actual atomic positions that you validate.”

ConceptResearch