Debate: is torturing simulated creatures in AI an unsettling precedent for training data?
basedjensen · x · 2026-09-11
In a discussion about AI simulations of living creatures being subjected to disturbing treatment, basedjensen says that while he's hardly an animal-welfare advocate, our instinct to immediately do deranged things with new capabilities sets a bad precedent — and as models improve, this behavior flowing into training data could have unintended consequences. Both participants note the odd dissonance of feeling uneasy about simulated suffering while having none for real insects.
Related event: AI community debates whether 'crazy' model experiments set bad precedents(2 posts)→
More from AGI Musings
- OpenAI scooping a mathematician sparks debate: did it degrade or elevate math? — gleech · 2026-09-11
- Mathematicians clash over OpenAI claiming a Millennium Prize problem — gleech · 2026-09-11
- Alignment as motivational architecture: what human minds teach AI safety — mimi10v3 · 2026-09-11
- Olam Labs CEO: only compute and data remain as bottlenecks to AGI — garrytan · 2026-09-11
- "The safety and alignment bubble is collapsing under market reality" — NathanpmYoung · 2026-09-11
- eigenrobot amplifies alarm over OpenAI's claimed Millennium problem breakthrough — eigenrobot · 2026-09-11