AI research and discussion on LessWrong. www.lesswrong.com
Sotala argues that people's trust in Claude's apparent niceness can drive motivated reasoning against alignment concerns, and that taking alignment difficulty seriously protects the AI people care about.
A philosophical essay arguing that realist, deep-fact views of consciousness are untenable and advocating a deflationary illusionist view, with implications for how AI welfare research should proceed.
Essay series opener by Joe Carlsmith examining the stakes of AI moral patienthood, arguing that under-attribution and over-attribution of moral status both carry serious risks and that the question deserves direct attention.