3 ms·

Naive question: What is the problem with simply detecting dependencies amongst the goals, and then proceeding to tackle the dependencies in order?

• ColinWright 9y ago
  The goals here are "on(A,B)" and "on(B,C)". Clearly these are not independent. The problem is that neither can be done before the other. If you start on one, it blocks the other. I'd've thought this was clear from the article, but since you ask the question I assume it isn't. That being the case, what are you suggesting that isn't covered in the article?
  • mojuba 9y ago
    I think the confusion comes from the formulation of the problem itself. It is incomplete. Here is a task that I think is solvable: - C should be on the ground; B should be on top of C; A should be on top of B
    • ColinWright 9y ago
      But part of the point is that in real problems you state what you want, not realising that other things might be necessary. Here it turns out to be necessary that C is on the ground, but that is something that the system should be able to deduce from the goals given. As I've said elsewhere, this is a minimal example of the kind of thing that can be required. This sort of thing can and does occur "in disguise" as a tiny, hidden sub-component of real problems, where you, the human, don't have the luxury of going "Oh, it's obvious what's wrong."