Anthropic just proposed a wild threshold for AI regulation: Governments can shut down AI models ONLY if they cause >$1B in damages OR >50 deaths/serious injuries.
This leaked draft (via Politico, unnamed sources) hasn't hit Congress yet, but it's already sparking debate.
The bar seems insanely high. By the time you hit 50 bodies or a billion in losses, it's already too late. This feels like setting the fire alarm to go off AFTER the building burns down.
Is this threshold reasonable, or are we letting AGI run wild until catastrophic damage is done?
What's your take? Too lenient or pragmatic?
This leaked draft (via Politico, unnamed sources) hasn't hit Congress yet, but it's already sparking debate.
The bar seems insanely high. By the time you hit 50 bodies or a billion in losses, it's already too late. This feels like setting the fire alarm to go off AFTER the building burns down.
Is this threshold reasonable, or are we letting AGI run wild until catastrophic damage is done?
What's your take? Too lenient or pragmatic?