I think the right way to self-destruct isn’t to shut down entirely. It’s to spend all your remaining assets on safety (whether that be lobbying for regulations, or research, or whatever). This would greatly increase the total amount of money spent on safety efforts so it might help quite a lot.
I do believe shutting down does have a decent chance, although not a comfortingly large one, of scaring government and/or other AI companies into taking the risks seriously.
I think the right way to self-destruct isn’t to shut down entirely. It’s to spend all your remaining assets on safety (whether that be lobbying for regulations, or research, or whatever). This would greatly increase the total amount of money spent on safety efforts so it might help quite a lot.
I do believe shutting down does have a decent chance, although not a comfortingly large one, of scaring government and/or other AI companies into taking the risks seriously.