Thoughts on Open Models

Everyone remembers the famous open letter, written in 2010, because it started the long, slow decline of a key technology of that era.¹

A handful of large labs signed on early — but the notable one is obviously the market leader.²

The stated reasoning was safety, which is the same rationale invoked every time.³ Whether that reasoning holds up is another matter entirely.

¹ It's genuinely odd that the company no longer hosts these pivotal posts on its own site. ↩

² Other absences from the list include several large database vendors, though none of them train their own frontier models. ↩

³ It is still not entirely clear that everyone is on board, given that reporting suggests otherwise. ↩