What's this about?
Microsoft AI's leader argues that teaching models person-like interests complicates containment; this is a philosophical/training-policy argument, not an experimental proof about consciousness.
What exactly is model welfare?
The idea that possible model experiences or interests might deserve ethical consideration. It is different from whether a chatbot's first-person statements are evidence of experience.
What's the community saying?
HN 202 points, 537 comments.
- There is no agreed empirical test of machine consciousness.
- Training a persona can create circular evidence from its later answers.
- Others focus on user education, corporate rivalry and future personhood criteria.