I may not personally like it, but it's the pragmatist thing to do if you're leading openai. The gov will eventually get involved, better to get in front of it.
However, there won't be a monopoly or anything of the sort that will form, our society's institutions wouldn't be able to properly enforce it and it's too easy for weights to leak, too much alpha for people to not fight back against it. Not worried.
The ai safety crowd's problem is that they're focused on a hypothetical entity that will take over humanity, imagining all sorts of scenarios in lalaland which are ridiculous (some of their "thought leaders" even advocating for strikes on datacenters to save humanity???), where the real risk is criminals/terros being empowered by the ai's cognitive ability. For that, we'll need the gov involved one way or another.
However, there won't be a monopoly or anything of the sort that will form, our society's institutions wouldn't be able to properly enforce it and it's too easy for weights to leak, too much alpha for people to not fight back against it. Not worried.
The ai safety crowd's problem is that they're focused on a hypothetical entity that will take over humanity, imagining all sorts of scenarios in lalaland which are ridiculous (some of their "thought leaders" even advocating for strikes on datacenters to save humanity???), where the real risk is criminals/terros being empowered by the ai's cognitive ability. For that, we'll need the gov involved one way or another.