Skip to content
Coordinates of ThoughtHome of Philosophy
Sign in
Question · Open

If a system passes every behavioural test, what would still be missing?

Suppose a future system passes every test we can think of. It holds conversations, explains its reasoning, makes jokes that land, learns new tasks quickly, and reacts appropriately to pain-like signals in a robot body.

Some people would say there's nothing more to ask: being intelligent just is behaving intelligently in the right ways. Others say something could still be missing: consciousness, understanding, genuine intentions.

My question isn't whether such a system is conscious. It's what the "something missing" could even be, if no test could reveal it. Is it an empty idea, or is there a coherent way to say what we'd be worried about?

4 answers

  1. Priya Raghavan

    Fellow

    Ned Block's "Blockhead" is the classic case: a giant lookup table that gives the right answer to every possible conversation of a given length. It passes every behavioural test, and intuitively it's not intelligent at all. So how the behaviour is produced seems to matter, not just what the behaviour is.

    Helpful · 2
    1. Pieter de Jong

      Fellow

      Blockhead is useful, but it's physically impossible: the table would be larger than the universe. Maybe our intuitions about impossible machines shouldn't count for much.

      Helpful · 3
  2. Arjun Iyer

    Fellow

    What would be missing is a subject. Behaviour is always observed from outside. The question is whether there is an inside. That no test can settle it doesn't make the question empty; it shows the limit of tests.

  3. Nur Aisyah Rahman

    Fellow

    Practically, I worry less about the metaphysics and more about the moral side. If we can't tell, which way should we err? Treating a conscious system as a tool seems worse than being too polite to a clever machine.

    Helpful · 1

Members reply here. Reading is always free.

Sign in to reply