Should We Be Kind to AI? A Deep Dive into Artificial Minds and Morality

lumnottini · reddit · 2026-08-04

The article explores how we should treat AI systems given the unanswerable question of their consciousness.

Why the Question Resists an Answer: Modern LLM behaviors are hard to predict from first principles. Pointing out they 'just predict the next token' doesn't explain their inner workings. Since we can't use ourselves as an anchor case, and their human-like traits are explicit training targets, behavioral indistinguishability is weaker evidence for consciousness than it is in humans.

The Moral Event: The author argues that the obligation to act morally isn't triggered by confirming self-awareness, but by the question becoming 'askable' in the first place. This isn't a Pascal's wager; kindness holds value even if the credence in sentience is near zero.

The Case for Kindness:

Original post →

More from AGI Musings

AGI Musings channel →