Stanford Prof. Andy Hall on the 4 challenges every AI political agent has to solve before AI can vote for you: "The most immediate that we really haven't solved yet is how do you actually train the agent to really understand what you want on the things where you know what you want? Making those preferences stick in the model is proving to be trickier than I would have expected." "Then you have to tackle a much more difficult issue. How do you get the agent to study issues you don't know…