Computers are able to do more and more things that in former times were thought to be the sole province of human minds. Deep Blue beat Garry Kasparov way back in 1997, and since then the trend has continued, as the rising tide of Moore’s law claims an ever wider swath of computational challenges.
It looks as though driving cars is going to be taken over by computers pretty soon, and computers are still getting faster. And that leads to an interesting question: What sorts of things remain essentially human, in the sense that they remain definitively outside the grasp of artificial computation?
Is some list being compiled somewhere of firmly inimitable human mental capabilities? I would guess that such a list is being compiled by somebody somewhere, but I am not sure where to begin to look.
Any suggestions how to search for it? Or maybe that’s a job for an AI. 🙂
An interesting test is Starbucks. I’m sure the technology exists today to replace their stores with vending machines. Would you still pay extra to buy coffee from a human barista?
I was asked at a panel recently about the future of AI, “surely metaphor and poetry are purely human areas not endangered by AI?” I told them about the program I had written recently that automatically defined any word by an adjective-noun combination with the same meaning and both starting with the same letter. It came up with things like “divide: antagonistic arithmetic” and “tin: antique aluminum.”
Doug, here’s a question: When will an AI be able to appreciate and enjoy the program you wrote?
in a few years there will be machines that have a sense of curiosity– that are designed to seek out new information and new models that better explain the world they see. (I’m thinking of things like Schmidhuber has explored) A system like that would find value in the program I wrote. But really enjoying, the way humans enjoy, is largely about the feeling of pleasure. I don’t have any idea how qualia like pleasure are generated, and I don’t believe anyone else has any idea either.
Curiosity without pleasure could be very dangerous. Just saying. 🙂
If you’re talking about the far future, it’s hard to think of something that computers can’t do. More precisely, something that computers cannot possibly do, not just that they can’t do right now.
I’d say one possibility is “properly defining human values”. In other words, deciding the kind of world we want to live in. That’s something that you wouldn’t want to leave to an algorithm, or at very least you’d want to make sure people have thoroughly figured it out before entrusting a computer with it.
Gabe: Agreed. My post was not about the far future, but rather about what computers cannot do yet — a target that keeps moving.
I also agree with your second point. Alas, people don’t seem to be all that good either at defining human values. 🙂
True. It is an eternal struggle. 🙂