Big AI companies actually really love hiring philosophers?！

The latest Claude Charter was drawn up with philosopher Amanda leading the effort.

At least four philosophers hold key roles at Anthropic, at least ten at DeepMind. When OpenAI was developing behavior guidelines for ChatGPT, it also consulted hundreds of moral philosophers.

AI brings a whole series of thorny problems—and that’s exactly what philosophers are best at.

First, philosophers can set behavioral boundaries for AI.

AI needs to decide which requests should be refused, what to prioritize when safety conflicts with users’ preferences—issues that involve understanding harm, honesty, and responsibility, which are precisely the subjects philosophers have studied for a long time.

They can turn fuzzy value judgments into principles that models can follow.

Second, philosophical training can also reduce AI’s overconfidence and its tendency to pander.

AI is prone to give affirmative answers when it’s uncertain, and to continue responding in line with the user’s viewpoint.

Philosophical training emphasizes checking assumptions and evaluating arguments. A Socratic-style sequence of probing questions can help a model spot contradictions and acknowledge its own limits of knowledge.

Finally, philosophers can help AI handle choices that don’t have standard answers.

When an autonomous driving vehicle causes an accident, how can harm be reduced? When a conversational support model faces self-harm risk, how should it respond? There are no simple answers.

Philosophers have long studied problems like these—they will confirm which red lines cannot be crossed, then analyze what consequences each option would lead to.