I am curious about how artificial intelligence models really work when they have to make real decisions instead of just answering questions. Most of the time I see artificial intelligence models being tested on how they answer questions or solve puzzles.. I think that making decisions when things are not certain and you do not get feedback right away is a totally different thing. This is a skill that is not really tested anywhere.
I mean decision making where you are not sure if you made the right choice until a long time later. The information you have is not complete. It is messy. There is no right or wrong answer.
I think this is a way to see if artificial intelligence models can really reason and think. It is more honest, than most of the tests that are done now. Has anyone seen any tests that really try to measure this of just using the same old questions and answers?
[link] [comments]