concept

AI Safety & Alignment

The technical and philosophical challenge of ensuring AI systems do what we want — alignment research, RLHF, constitutional AI, jailbreaking, red-teaming, and the existential risk debate between AI safety researchers and accelerationists.

41stories
144,266records · all-time
16,684records · 7d
5,753daily avg · 30d
just nowlast record
-81%vs prior week
EntitydevelopingYouTubev1

YouTube Becomes AI Safety's Loudest Unmoderated Stage

AI-generated scams, slop content, and safety debates now flood YouTube faster than its moderation can respond — making it the platform where AI risk lands in public view first.

  • ·YouTube is where most people form their first intuitions about AI risk — and its moderation is calibrated to advertiser liability, not epistemic harm.
  • ·AI-generated scam ads using celebrity likenesses are already appearing on YouTube, confirming the platform's deception problem is in production, not theoretical.
  • ·The demonetization policy YouTube uses as its AI guardrail is being gamed: creators are optimizing for volume just below the threshold, not for quality.
FeaturedRead →