LESSWRONGTags
LW

Quantilization

EditHistory
Discussion (0)
Help improve this page
EditHistory
Discussion (0)
Help improve this page
Quantilization
See also
Random Tag
Contributors
4plex

A Quantilizer is a proposed AI design which aims to reduce the harms from Goodhart's law and specification gaming by selecting reasonably effective actions from a distribution of human-like actions, rather than maximizing over actions. It it more of a theoretical tool for exploring ways around these problems than a practical buildable design.

See also

  • Rob Miles's Quantilizers: AI That Doesn't Try Too Hard
  • Arbital page on Quantilizers
Posts tagged Quantilization
5
106Soft optimization makes the value target biggerΩ
Jeremy Gillen
5mo
Ω
20
5
33Quantilizers maximize expected utility subject to a conservative cost constraintΩ
jessicata
8y
Ω
3
5
26Another view of quantilizers: avoiding Goodhart's LawΩ
jessicata
7y
Ω
2
4
65When to use quantilizationΩ
RyanCarey
4y
Ω
5
4
24Quantilizers and Generative ModelsΩ
Adam Jermyn
1y
Ω
5
2
25Why don't quantilizers also cut off the upper end of the distribution?QΩ
Alex_Altair, Jeremy Gillen
1mo
QΩ
2
2
11Quantilizer ≡ Optimizer with a Bounded Amount of OutputΩ
itaibn0
2y
Ω
4
1
30Recursive Quantilizers IIΩ
abramdemski
3y
Ω
15
1
19Stable Pointers to Value III: Recursive QuantilizationΩ
abramdemski
5y
Ω
4
Add Posts