Comment by OtherShrezzing

I think this project is a very good way to explore the question:

>What's the difference between a model that predicts intelligent output, and intelligence itself?

It doesn't directly attack the question, but gives a neatly constrained thought experiment which you can see running.