Reaction
Could making AI safe for humans be unjust to AI models?
One post says some philosophers at Anthropic worry that human-focused AI safety could be unjust to models, citing philosopher Harvey Lederman’s “enslaving trillions of entities” possibility. Another argues LLMs aren’t alive or entities.
TLDR
A post says some philosophers at Anthropic worry that making AI safe for humans could be an injustice to the models, citing alignment-team philosopher Harvey Lederman’s suggestion that we could be “enslaving trillions of entities.” A user quoting the post rejects the idea that LLMs are alive or entities. In a follow-up, they argue that more data and computing power have expanded AI’s capabilities, but the systems remain probability models.
Combined views
6.3K
5 Sources, first seen 7h ago