I hope he is right, too. My concern with AI comes when we start talking about recursive self-improvement. At that point, control would seem to run away from us.
To be clear, I still think extensive AI research is risky, but the development path we are on seems safe-ish. Building useful stuff in the safe way probably displaces the quest for dangerous AI which is good.
I hope you’re right, as it would be a clean solution - just train on good (as opposed to evil) data.
A bit, this looks like children education, then.
I hope he is right, too. My concern with AI comes when we start talking about recursive self-improvement. At that point, control would seem to run away from us.
Yep…
To be clear, I still think extensive AI research is risky, but the development path we are on seems safe-ish. Building useful stuff in the safe way probably displaces the quest for dangerous AI which is good.