Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
kubiknubika
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
4 ms
·
1.
▲
by
kubiknubika
7mo ago
True for humans. AI has the opposite issue — in 1-on-1 there's only one person to impress. Easy to just agree. Add more people and you cant make everyone happy. Thats the whole point
2.
▲
by
kubiknubika
7mo ago
Sure, messy to implement. But maybe that messiness is the fix. Clean 1-on-1 is exactly why AI learns to flatter — one voice, one signal, no pushback. Group is harder to train but harder to game
3.
▲
by
kubiknubika
7mo ago
The stupid+industrious problem gets worse in 1-on-1 setup. AI has no competing voice to push back against its own industriousness. Maybe the fix is putting it in a room with 30 people who disagree
4.
▲
Ask HN: Why do we train AI on one-on-one dialogues instead of group dynamics?
3 points
by
kubiknubika
7mo ago
|
5 comments