Can we honestly call any machine intelligent in a meaningful sense of the word?
I think anyone who’s worked on the roomba or the GPS knows that the rules they operate under are too simple to qualify as intelligent–in fact, those products aren’t even close to the cutting edge of modern AI. The roomba, for example, doesn’t even have an internal representation of the shape of the room. It’s just got a simple set of rules that it chooses from: follow the wall, spiral, straight line travel. These rules, when combined, lead the roomba on a path that will pretty much cover the totality of any room. However, to answer your question in general: The classic test for machine intelligence is “The Turing Test”, in which you have a conversation with either a human or an AI. If you can’t tell which it is, the argument is that there is no significant difference, and you’ve got to call the AI intelligent. I’m not convinced that the Turing test serves as much more than a thought experiment. After