Debate: Human Taste Can Be Gamed, Not Verified

Commenters debate whether human taste can be optimized like a reward signal. @torchcompiled argues that optimized 'tasteslop' is a hollow supernormal stimulus, noting preferences are continuous, multi-axial, and often intransitive—making taste far from verifiable.

2026-08-25 ~ 2026-08-25 · 2 related posts

• Human taste can be gamed into optimized slop; preference is not a verifiable reward — torchcompiled · 2026-08-25
• Human Taste Can Be Gamed: The Challenges of Verifiable Preferences — zacharynado · 2026-08-25