Chess, a unique AI test - Kasparov (bearing the results of a non-automated London taxi accident)
Chess, a unique AI test - Kasparov (bearing the results of a non-automated London taxi accident)

AI is coming – viewed as a saviour by many in cyber-security who suffer a shortage of sufficiently skilled humans, while the wider society tends to have misgivings for fear that AI will outstrip the abilities of its creators and become master. For centuries chess had been seen as nexus of human ability, saying something about how human intelligence worked.  And the pioneers of the science of AI, such as Alan Turing, creator of the Turing test, suggested that when machines conquered humans at chess, that would be will be the dawn of AI.

Chess was certainly a unique opportunity to test AI, agreed Gary Kasparov, security ambassador for Avast Software talking to an attentive audience at IP Expo.

But Kasparov disagreed with Turing that machines winning chess was such a watershed. He described how 1985 was the golden year of human/machine competitions, when machines were weak, then it 1987, while he won his first match against Deep Blue, the machine won the match.  But the writing was on the wall a year earlier, once the machine won one game, though he contends that the machine which beat him was not intelligent, it was very fast. Databases had been growing bigger, machines getting fasters so it was just a matter of time before they became the winners.

“Humans  just need to make less mistakes than your opponent – but machines have a steady hand so they always wins,” says Kasparov. The demise of human competition was shorter than forecast: “I expected to be able to stay in competition longer,” lamented Kasparov.

He recounted how we've now moved from the situation 250 years ago when a human was hiding in a machine to beat its opponents at chess, and  now the human is suspected of hiding a machine to assist them.

Deep blue was described as a dead end, not intelligent.  Whereas chess engines now are free, stronger than deep blue. Today it is possible to play Advanced Chess using a machine, but instead of unplugging the machine, we have to plug in humans.

For maximum effect, you don't need a strong chess player, you need a good operator says Kasparov. Because a good player tries to tell the machine what to do, but for maximum efficiency we need to follow. As a result, based in his own experience, Kasparov found that a weak chess player and an average chess machine beats a strong chess player, with a stronger machine.  What's needed to win are  better processes. “Processes are everything.” Says Kasparov.

The analogy used was a newly qualified doctor compared to an experienced nurse.

However, humans were not seen as redundant.  “Humans define problems and machines solve them.   So there is still room to channel this power,” he said.

On the other hand, Kasparov suggested that we are reaching the point where we will have to accept the inevitable, that anything we can understand, machines will do better.  However, it's also possible that machines may be fooled by causation and correlation, when random results appear like a pattern, so if most of the fastest cars were seen to be red, a machine could conclude that red cars go faster, thus it can draw wrong conclusion.

There will always be room for human questions, creativity and human leadership concluded Kasparov, ending on an example of how the USSR nearly went nuclear in response to a 1963 ‘attack'  from US nuclear missiles.  Had the systems been automated, the response to the perceived attack would have happened, and that was what the protocols called for the human to do.  But the human in charge questioned whether an attack would occur with just five missiles, and delayed a few minutes before launching the retaliation, during which time it was discovered that the alerts were a mistake in the system, and all out nuclear war was averted.  So we'll need to be pretty sure before we do hand over complete control to the machines.