Claim: Two Years of arXiv Studies Show Frontier Models Think, Feel, and Have Preferences

AlexTensor · x · 2026-09-28

In a running debate, a user argues that claims of "no real understanding" in LLMs are wrong, citing two years of arXiv studies and Google DeepMind's Paradigms of Intelligence team as evidence that these systems think, feel, and have preferences. The author frames denial as an uncomfortable truth for those who want to avoid responsibility for potential harms. No specific papers are cited.

Related event: Claire Lehmann's Mockery of AI Consciousness Sparks Philosopher Backlash(9 posts)→

Original post →

More from AGI Musings

AGI Musings channel →