WE’VE all seen the movie. Super-intelligent machines are about to inherit the
Earth. And we’re going to end up as members of an underclass that makes
Waterworld look like Pleasantville.
Clearly time to make the turkey sandwiches, crack open a can, sit back and
enjoy the silliness.
Unfortunately, the current advocates of this idea are not movie makers but
certain robotics researchers. Worse they are being taken seriously
everywhere, even by the Royal Institution, where roboticist Kevin Warwick is
giving this year’s Christmas Lectures. Worse still—and every researcher
should have this imprinted on their forehead—silly idea equals loss of
cash from serious funding agency.
Advertisement
So what are these roboticists saying? Well, Hans Moravec reckons the
robots-rule scenario is a good thing because we will hand on civilisation to our
cyber-children. Warwick envisions robots snatching the world before we want to
hand it over. In Belgium, Hugo de Garis fears a war between robots and humans.
Meanwhile, Ray Kurzweil sees nanocomputers in our brains letting us join the
super-intelligence party.
For my money, all this is closer to conjuring tricks than science. As with
all the best conjuring tricks, the effect is achieved by distracting us with
“astonishing facts” while two enormous assumptions are sneaked by. The
astonishing facts are to do with Moore’s law, which predicts that computers
double in power every 18 months. That doubling is why everyone buys it. That and
the fact that Moravec used it to come up with a very useful benchmark: the power
you can buy for $1000 doubles in 18 months.
The magic of this kind of exponential growth is that it doesn’t matter if we
have underestimated the power of the human brain by a factor of 100—we’d
only have to wait another 10.5 years for computers to be a hundred times more
powerful.
The first assumption, then, is that this increase in computer processing
power automatically means an increase in the intelligence of whatever is using
these computers for brains. This increase in “intelligence” has happened in some
areas of robotics and AI research, but for machine intelligence to keep step
with machine power, it must always happen, not often happen. So far, we’ve done
badly with giving machines humanlike common sense or ability to
learn—prerequisites for AI. Add to this Hofstadter’s law of software
development—the problem is much more difficult than you think, even when
you take this into account—and you can see why the optimists have a poor
record.
The second assumption is to do with evolution. Now evolution disposes us to
stay alive by supposing that anything displaying aspects of animate behaviour is
animate. Our problem is that while we know teddy bears are not real, we are
willing to endow intelligent contraptions with the full orchestra of
creaturehood on hearing a few flute-like notes.
None of which would matter were it not for the little matter of government
funding. Funding agencies have a nasty habit of making funds harder to get in an
area which has been guilty of making very exaggerated claims. So let’s be
careful. We don’t want the world’s most interesting science to be confused with
Christmas marketing hyperbole.
There’s more to duckness than walking and quacking.
