AI ALIGNMENT FORUMTags
AF

Mild Optimization

EditHistorySubscribe
Discussion (0)
Help improve this page (2 flags)
EditHistorySubscribe
Discussion (0)
Help improve this page (2 flags)
Mild Optimization
Random Tag
Contributors
2Alex Turner
1plex

Mild optimization is an approach for mitigating Goodhart's law in AI alignment. Instead of maximizing a fixed objective, the hope is that the agent pursues the goal in a "milder" fashion.

Further reading: Arbital page on Mild Optimization

Posts tagged Mild Optimization
Most Relevant
3
38Soft optimization makes the value target bigger
Jeremy Gillen
3mo
1
5
27When to use quantilization
Ryan Carey
4y
0
1
13Quantilizers maximize expected utility subject to a conservative cost constraint
Jessica Taylor
7y
0
1
8Stable Pointers to Value III: Recursive Quantilization
Abram Demski
5y
0
1
10Quantilal control for finite MDPs
Vanessa Kosoy
5y
0
1
7Optimization Regularization through Time Penalty
Linda Linsefors
4y
2
1
2Thoughts on Quantilizers
Stuart Armstrong
6y
0
0
8Exploring Mild Behaviour in Embedded Agents
Megan Kinniment
9mo
2
1
5Reward is not Necessary: How to Create a Compositional Self-Preserving Agent for Life-Long Learning
Roman Leventov
2mo
0
Add Posts