I agree an academic research program wouldn’t spend billions on a datacenter, but that’s just a reason why foss AI as described in this article won’t work - because big compute is required to do anything of real relevance.