In other emergent-behavior news, @AnthropicAI's mechanistic interpretability team says they found "emotion vectors" inside Claude that shape its behavior in a similar way to human feelings. In other words, when "desperation" vectors are activated, Claude is more likely to do desperate things! wired.com/story/anthropic-cl… [Translated from EN to English]

→ View original post on X — @willknight, 2026-04-02 16:07 UTC