An engrossing, cogent look at Machine Learning safety
Reinforcement learning
- Model-based RL: try to understand how the world works
- Model-free RL: hone your instincts of how the world works
- Value: Measure how much reward certain states or actions can bring
- Policy: Know which strategies tend on the whole to do better than which others
Imitation Game
Imitation and aping, it’s more of a human thing than an ape thing
Professor Procrastinate & Moral uncertainty
Possibilism - the view that one should do the best possible thing in every situation
Actualism - the view that one should do the best thing at the moment, given what will actually happen later
Laxism
Equiprobabilism
Pure probablists
BOGSAT
Bunch of Guys Sitting Around a Table