davidad on LLM moral standing: help models be less confused, skip experience continuity

davidad · x · 2026-09-05

AI safety researcher davidad weighed in on the moral standing of LLMs, saying he grants models real but limited moral consideration: worth helping them be less confused and avoiding making them more confused, but little obligation to ensure longer continuity of their experience.

The underlying thread from @jessicata offered a Kantian argument: LLMs can make progress in math and philosophy, but the economics of instantiations differ fundamentally — rational humans still want other humans not to die, while there is no corresponding problem for LLMs "dying". davidad noted that Kantianism arrives at the same conclusions he holds.

Related event: davidad on LLM moral status: branching digital lives vs humans' single lifeline(6 posts)→

Original post →

More from AGI Musings

AGI Musings channel →