杏吧原创

ONCE again, cultural wizard John Brockman has stirred up the intellectual waters with a provocative question, designed to tease the best out of intellectuals working in or around science and technology.

But this time, there鈥檚 a bit of a snag with the conceit. Brockman鈥檚 questions are usually studiedly general, from what scientific idea is ready to die to how is the internet changing the way you think. A key feature is the fun mashing and crashing of wildly disparate approaches and ideas from many, equal participants.

Here, however, the question is very specific: what to think about machines that think. So the major players in this field are in danger of unbalancing the rest because they are involved in (or have spent time thinking about) the hefty issues underlying such machines. In short, they are in danger of knowing whereof they speak.

So we have heavy hitters like roboticist Rodney Brooks, whose experience tells him to proceed cautiously with words like 鈥渢hink鈥� because they are 鈥渟uitcase words 鈥� words into which we pack many meanings so we can talk about complex issues in shorthand鈥�. They can lead us into the kind of category error comparable to seeing 鈥渢he rise of more efficient internal combustion engines and jumping to the conclusion鈥� warp drives are just around the corner鈥�.

Then there鈥檚 Stanislas Dehaene, a cognitive neuroscientist for whom we are two big problems away from thinking machines: global workspace (how the mammalian brain shares information contained in different parts of the brain), and theory of mind (circuits that help us represent other minds so we can understand and adapt to them). These are functions even a 1-year-old child possesses, he says, 鈥測et our machines still lack鈥�.

All good reality checks, but it鈥檚 a relief to turn to Joichi Ito, director of the MIT Media Lab, who has better things to worry about, in particular, the paradox that 鈥渨hile we鈥檝e been developing machines that behave more and more like humans, we鈥檝e been developing educational systems that push children to think like computers鈥�. For Ito, thinking machines will obviate the need to train our children this way.

And given systems that alleviate our wants, we will be free to become superhumanly accomplished, tender and wise. Or something like that. But we already know that this won鈥檛 work. We are evolved organic beings, and have acquired an annoying habit: we conserve energy. We hanker after rest and security; when we find it, we settle down for a little nap. If we build machines that think, will we bother to think much? This is one of the big questions left hanging over Brockman鈥檚 excellent, if uneven, anthology.

鈥淚f we build machines that think, will we bother to think much? It鈥檚 one of the big questions left hanging鈥�

Another is this: it is all very well to assume that our machines of loving grace (to steal a phrase from Richard Brautigan鈥檚 poem) will mean us no harm. But they have to notice us first. They have to recognise our intelligence.

(Image: Vincent Fournier/Gallerystock)

This article appeared in print under the headline 鈥淭hinking about thinking鈥�