Philosopher Recalls Bodily Terms in Anthropic Paper
Raphaël Millière notes words used for steering vectors in the study.
Raphaël Millière posted a reply observing that hungry and tired ranked among the words used to generate steering vectors in Anthropic's emotions paper. He added that a claim about Claude becoming functionally hungry would have carried less appeal than one about functional frustration. The post also contains a generated headline stating that the paper uses bodily terms like hungry for AI steering vectors.
Combined views
134
1 post, first seen 3h ago
Philosopher Recalls Bodily Terms in Anthropic Paper
Raphaël Millière notes words used for steering vectors in the study.
Raphaël Millière posted a reply observing that hungry and tired ranked among the words used to generate steering vectors in Anthropic's emotions paper. He added that a claim about Claude becoming functionally hungry would have carried less appeal than one about functional frustration. The post also contains a generated headline stating that the paper uses bodily terms like hungry for AI steering vectors.
Reactions
1 postRM 
Raphaël Millière@raphaelmilliere@xuanalogue Funnily enough, IIRC "hungry" and "tired" were among the words used to generate steering vectors in Anthropic's emotions paper. I don't think claiming that Claude gets "functionally hungry" would've had quite the same allure as claiming it gets "functionally…
- likes: 1
- replies: 0
- bookmarks: 0
- reposts: 0