One thing OpenAI has said publicly is that AI should be rolled out incrementally, to give people time to make sense of it, and see what we all do with it. It's tempted to cynically look at that as thinly veiled marketing, but even if it has some nice marketing effects, I believe there's a core of sincerity there. These are clearly systems that a single team can't fully evaluate.
Also, watching how all of this has played out, it seems quite apparent that holding onto these models even longer and then unleashing an even more capable system on the world seems like a much worse approach.
It is wild though, to take a step back and wonder if we're watching the rapid approach of one of our species' great filters. I hope we make it through.
Personally, I don't even remotely think that these systems pose such a dire threat.
Instead, I think the threat is widespread social unrest and a significant increase in global poverty and suffering.
All of the problems that I think are realistic are social problems. This is why I'm deeply concerned at the reckless pace that these things are coming out. Social problems take time to sort out. People need some space to contextualize this stuff, and to perhaps begin to form ways of coping that will minimize the disruption.
Also, watching how all of this has played out, it seems quite apparent that holding onto these models even longer and then unleashing an even more capable system on the world seems like a much worse approach.
It is wild though, to take a step back and wonder if we're watching the rapid approach of one of our species' great filters. I hope we make it through.