Machines are smart, but can they think about thinking?

20 Mar

Computers can play chess, Jeopardy and now they can do crossword puzzles. In a competition last weekend with 600 of the country’s top puzzlers, a machine named Dr. Fill came in 141.

The term “artificial intelligence” is usually associated with computers that do such things. I have my doubts. I suspect smart humans are simply slipping their own knowledge into code which then exits a box that does not think. When IBM’s Big Blue was losing at chess, a small army of programmers quickly updated the code.

That’s almost like cheating at cards.

I believe a box is capable of thinking, but before it can someone will have to unlock a secret … uncover a currently unknown approach that will allow a human process to be trans-synthesized by something nonhuman. When this happens, machine intelligence really won’t be artificial. The will have to change the term.

In AI there is debate over whether a machine designed to think should do so like a person or like a machine. The former, once popular, is losing out to the latter. It’s because the former didn’t work very well.

I say it has got to be both. Consider this:

A human – me, for example – is asked a question. The question is: “How many rhymes are in the song Moonlight in Vermont?” The human, me, chooses not to answer immediately but to think: Why such a silly, out-of-context question?

Would the machine do that? Doubtful.

After mulling it over, I would agree with myself that the person asking the question did not do so to delight in telling me that my answer of eight is wrong and instead the correct answer is 10. There would be no satisfaction in that. No, the questioner wants me to say eight and then find joy in telling me the answer is either really high – say 500 – or really low – say one. Maybe it is zero.

In other words, I would use my intelligence to determine that “How many rhymes are in the song Moonlight in Vermont?” is a trick question. Does a computer know what a trick question is?

Only if it can think like a human.

Of course, it doesn’t have to think at all if it has access to every song ever written. All it has to do then is get the song and count. That is not AI.

So what is AI?

And how many rhymes really ARE in Moonlight in Vermont?

One Response to “Machines are smart, but can they think about thinking?”

  1. MBREBO March 22, 2012 at 2:58 pm #

    Making a machine think too much like a human would forfeit most if not all of the proposed advantages of having such a machine in the first place.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: