You can consider an ideal agent that uses argmax E to find what it chooses, where E is some environment function . Then what you arrive at is that argmax gets defined recursively—E contains argmax as well—and it just so happens that the resulting expression is only well defined if there’s nothing in the first box and you choose both boxes. I’m writing a short paper about that.
You can consider an ideal agent that uses argmax E to find what it chooses, where E is some environment function . Then what you arrive at is that argmax gets defined recursively—E contains argmax as well—and it just so happens that the resulting expression is only well defined if there’s nothing in the first box and you choose both boxes. I’m writing a short paper about that.