I think your pessimism of org culture is pretty relevant for the question of big decisions that GDM may make, but I think there is absolutely still a case to be made for the value of alignment research conducted wherever. If the research ends up published, then the origin shouldn’t be held too much against it.
So yes, having a few more researchers at GDM doesn’t solve the corporate race problem, but I don’t think it worsens it either.
It might be “fine” to do research at GDM (depending on how free you are to actually pursue good research directions, or how good a mentor you have). But, part of the schema in Mark’s post is “where should one go for actively good second-order effects?”.
I think your pessimism of org culture is pretty relevant for the question of big decisions that GDM may make, but I think there is absolutely still a case to be made for the value of alignment research conducted wherever. If the research ends up published, then the origin shouldn’t be held too much against it.
So yes, having a few more researchers at GDM doesn’t solve the corporate race problem, but I don’t think it worsens it either.
As for pausing, I think it’s a terrible idea. I’m pretty confident that any sort of large scale pause would be compute threshold focused, and would be worse than not pausing because it would shift research pressure towards algorithmic efficiency. More on that here: https://www.lesswrong.com/posts/Kobbt3nQgv3yn29pr/my-theory-of-change-for-working-in-ai-healthtech?commentId=qwixG4xYeFdELb2GJ
It might be “fine” to do research at GDM (depending on how free you are to actually pursue good research directions, or how good a mentor you have). But, part of the schema in Mark’s post is “where should one go for actively good second-order effects?”.