QuantilizersSaferAlternative.pdf
intelligence.org · 7,077 words · saved by 1 readers
N/A
Quantilizers: A Safer Alternative to Maximizers for Limited Optimization Jessica Taylor Machine Intelligence Research Institute jessica@intelligence.org Abstract utility function, with U (o) being the utility of outcome o. Then an expected utility maximizer is an agent that In the field of AI, expected…
related reading
- Why The Focus on Expected Utility Maximisers? — LessWronglesswrong.com
- Why Tool AIs Want to Be Agent AIs · Gwern.netgwern.net
- Towards a scale-free theory of intelligent agencymindthefuture.info
- Optimality is the tiger, and agents are its teeth — LessWronglesswrong.com
- Pascal's mugging - Wikipediaen.wikipedia.org
- Utility Maximization = Description Length Minimization — LessWronglesswrong.com
- Towards a scale-free theory of intelligent agency — AI Alignment Forumalignmentforum.org
- Dreams of Friendliness — LessWronglesswrong.com
- Skyscrapers and madmen - Joe Carlsmithjoecarlsmith.com
- The Model You Audit Is Not the Model You Shiptechpolicy.press
- A positive case for how we might succeed at prosaic AI alignment — AI Alignment Forumalignmentforum.org
- Pascal's Mugging: Tiny Probabilities of Vast Utilities — LessWronglesswrong.com