"It's not X, it's Y": RLHF-learned hedging is poisoning human discourse
sloppenheimer · x · 2026-10-04
The author argues that reinforcement-learned hedging phrases like "It's not X, it's Y" are the downfall of human discourse — people increasingly mimic the fence-sitting rhetorical style of RLHF-trained models.
More from Fun
- Developer wires Claude into a microwave that reasons about food and dings when done — andrew_n_carr · 2026-10-04
- "I'll know we hit AGI when labs ship Windows versions within a month" — menhguin · 2026-10-04
- Witty reply quips you could do two things at once by severing your corpus callosum — ZeroStateReflex · 2026-10-04
- The infinite torture stairs meme resurfaces and it's still hilarious — adamamcbride · 2026-10-04
- Immunologist generates 2-minute immunology history video entirely in code with Claude — DeryaTR_ · 2026-10-04
- Jensen Huang recalls night out in Seoul: 10 rounds of somaek with Samsung's Jay Lee — alexmacgregor__ · 2026-10-04