PHI 13 Lecture Notes - Lecture 5: Artificial General Intelligence, Grammar Checker
Document Summary
True ai will be general, it will be artificially general intelligence. That means it will not be domain restricted (like a chess playing machine, or a grammar checker), but will be able to handle an infinite variety of novel situations, just like human do. But how can an engineer guarantee and ethical agi, when the engineer cannot possibly predict all the situations the agi might agi encounter, let alone how it would react. The problem is that this is notoriously difficult to do. In fact, the robot might even figure that it must lock. Since a robot can derive incorrect answers like this one from ashmore laws. If they see a kidnapped people, who has no physical harm, no order given, so robot do not have to do anything. It is easy to sometimes get correctness and completeness confused. We don"t know what the right ethical role is or what result is.
