Monday, May 28, 2012

[PHI 1000] Can Machines Think Like Us?

John Searle argues that computing machines cannot think like us. His argument is based on a thought experiment and can be stated as follows:
  1. Computing machines can think like us only if a machine can run a program for Chinese and thereby come to understand Chinese.
  2. A machine can run a program for Chinese without thereby coming to understand Chinese.
  3. Therefore, machines cannot think like us. (That is, machines can follow syntactic rules to manipulate symbols, but they do not have an understanding of meaning or semantics.)
Now, meet Watson.

Watson is able to answer Jeopardy questions by testing hypotheses. Isn't that how we think (at least sometimes)? If so, does that mean that Watson can think like us? In other words, if we can't tell whether it is Watson or a human being who is answering Jeopardy questions, does that mean that Watson has a mind?


  1. This is a fascinating issue.

    I don't know if you meant the question about whether Watson can think like us to be read with an emphasis on the *like us*. But this is how I'm reading it for the purposes of this comment.

    It seems to me that Watson can think (in some sense of the word), but I'm reluctant to say that it has thoughts like you and me. I doubt that it feels surprised when it gets the wrong answer, has images come to its mind as its calculating which answer to spit out, that it regrets not opting for an alternative answer, or has other phenomenal experiences. This might be an important difference between organisms like us and the machines we make. It seems to preclude such machines from having minds like ours. Perhaps computer scientists, cognitive scientists, and philosophers will work together to make a machine that can have these other kinds of mental states. If they could pull it off, then I'd be tempted to say that such a computer thinks like us.

    There are a host of questions that your post raises: What is it to think? What sorts of behaviors/experiences are required for thought? Is simply behaving as if one is thinking enough to show that one is actually thinking? I wonder if you have any answers to these questions/what you think about them.

    1. Hi Jesse,

      Thanks for your comment. I agree that my post raises many questions for which I wish I had (good) answers. So let me say a couple of things that I hope will be illuminating.

      First, I think that it would still be correct to say that machines can think like us, even if they think like us only in some respect but not every respect. So, Watson may not be able to have thoughts about regret, for example, but that doesn’t mean that he cannot think like us in some respect. Watson seems to be able to think (i.e., reason) like us insofar as he can form hypotheses and test them. Of course, we don’t reason by testing hypotheses all the time, but we do reason in this way sometimes. And since Watson seems to be able to reason in this way as well, it seems to suggest that, to that extent, Watson is able to think (i.e., reason) like us.

      Second, suppose that Watson is playing Jeopardy with two other human beings. But we cannot see the participants. We can only hear their answers and determine whether the answers are correct or incorrect. Would we be able to tell who is answering the questions (i.e., if it is Watson or one of the human participants)? If the answer is “no,” and if Turing’s imitation game is a test for the presence of a mind, then it seems to follow that Watson has a mind.

  2. Don't let the title fool you, this is about you topic:


This is an academic blog about critical thinking, logic, and philosophy. So please refrain from making insulting, disparaging, and otherwise inappropriate comments. Also, if I publish your comment, that does not mean I agree with it. Thanks for reading and commenting on my blog.