[JumpCrisscross]

> they think the future of AI is small dedicated-task models

I mean, I do too. Or at least, this will be a large market. But the natural leader for that space is the firm building the frontier models. They have the most control and access for distillation.

[SyneRyder]

Hmm, I'm skeptical of small model AI - they are useful for specific domains & the task they're trained for, I just think it's thinking too small. I worry that the people who think AI is only useful for email triage and OCR will miss all the grand problems that get solved by the very largest frontier models. But that's me, and I could be wrong.

Speaking of me being wrong: the Mistral CEO tweeted today that they're coming out with a "fat but sparse" model in July, so big that it will only be available to governments & a few partners at first (ala Glasswing/Mythos I guess). I fear that's just marketing, but if Mistral really have gone all out to make a new truly large model & at least try to catchup to China, good luck to them!

https://x.com/arthurmensch/status/2066913356548542827