A hypothetical situation: AI is tied to a camera of me in my office. Doing basic object identification. I stand up. AI recognizes me, recognizes desk. Recognizes "human" and recognizes "desk". I sit on desk. Does AI mark it as a desk or as a chair?
And let's zoom in on the chair. AI sees "chair". Slowly zoom in on arm of chair. When does AI switch to "arm of chair"? Now, slowly zoom back out. When does AI switch to "chair"? And should it? When does a part become part of a greater whole, and when does a whole become constituent parts?
In other words, we have made great strides in teaching AI "physics" or "recognition", but we have made very little progress in teaching it metaphysics (categories, in this case) because half the people working on the problem don't even recognize metaphysics as a category even though without it, they could not perceive the world. Which is also why AI cannot perceive the world the way we do: no metaphysics.
And let's zoom in on the chair. AI sees "chair". Slowly zoom in on arm of chair. When does AI switch to "arm of chair"? Now, slowly zoom back out. When does AI switch to "chair"? And should it? When does a part become part of a greater whole, and when does a whole become constituent parts?
In other words, we have made great strides in teaching AI "physics" or "recognition", but we have made very little progress in teaching it metaphysics (categories, in this case) because half the people working on the problem don't even recognize metaphysics as a category even though without it, they could not perceive the world. Which is also why AI cannot perceive the world the way we do: no metaphysics.