Smarter than Humans
We've all heard about glorious potential upsides and horrifying potential downsides. But one way or another, we know that this development is going to happen -- if ever -- a long time from now.
Or possibly sooner:
'Genius' computer with an IQ of 150 is 'more intelligent' than 96 per cent of humansFirst off, computers outscoring humans on IQ tests is no more evidence that they have exceeded human intelligence than computers beating humans on Jeopardy. These kinds of developments show that we have programmed computers to solve increasingly complex problems. They won't truly exceed our intelligence until they're the ones deciding what they want to think about and subjecting us to tests.
A computer has become the first to be classed as a 'genius' after scoring 150 in an IQ test.
The average score for people is 100. A score of 150 ranks the artificial intelligence programme among the top four per cent of humans.
The software was designed by a team led by researcher Claes Strannegård at the University of Gothenburg. His aim was to make a programme that 'thinks' like a person.
'We're trying to make programmes that can discover the same types of patterns that humans can see,' he says.
The sentient artificial intelligence with a will of its own is a frightening idea to many. If its desires and goals don't align with ours (and why should they?) it represents a real threat. However, the methodology that Strannegård and his team have used to make their computer a star at scoring big on IQ tests suggests the right direction for getting us to greater-than-human machine intelligence:
Strannegård says 'One, two - what comes next? Most people would say 3, but it could also be a repeating sequence like 1, 2, 1 or a doubling sequence like 1, 2, 4. Neither of these alternatives is more mathematically correct than the others. What it comes down to is that most people have learned the 1-2-3 pattern.' The group is therefore using a psychological model of human patterns in their software.Making machines smarter by making them think more like us seems like a sound approach. The more like us they are, the more potentially sympathetic they will be with our goals and desires. In fact, this research suggests a potential safe model for getting to greater-than-human intelligence. If we continue to increase the complexity of problems that non-sentient systems can solve, eventually we can hand the following problem over to them:
They have integrated a mathematical model that models human-like [thought processes.] The group has improved the programme that specialises in number sequences to the point where its score implies an IQ of at least 150.
'Our programmes are beating the conventional math programmes because we are combining mathematics and psychology.'
How do we construct a greater-than-human intelligence that will not be a threat to us and that will be aligned with our goals and desires?
Of course, that wouldn't really be a single problem -- it would be reflected in myriad individual problems solved sequentially or in tandem. Bud it each step we take towards that answer produces a machine that is closer to us in its way of looking at the world, that can only help.
Ultimately we partner with our machines to produce a machine that will be better at solving our problems than we could ever hope to be. And that is a mighty big payoff.