AI The good, the bad, and the downright terrifying

Everybody knows the story of Superman. How a baby from another world is found and adopted by a humble farming couple who discover the boy possesses incredible powers

- WILLIAM LANGLEY

Mr and Mrs Kent bring their son, Clark, up with strong moral values and teach him that his duty is to help and protect humanity.

"But what about the opposite scenario?" asks Mo Gawdat, a former top executive at Google's artificial intelligence division. "Suppose the Kents had done things differently, and Superman had learned greed and self-interest. What would have stopped him from destroying the world? "This is where we are with AI."

In a sense, all of us - the scientists who have created a new breed of hyper-intelligent machines, the governments and corporations that will deploy them, and everyone who will come to rely on them - are the new Kents. If we can safely steer artificial intelligence towards good purposes, it has the potential to vastly improve our lives. If not, it is terrifyingly capable of destroying us.

In May, a group of AI-pioneering researchers and executives issued a blunt warning that reducing the dangers of the new technology should become a global priority. A few weeks earlier Professor Geoffrey Hinton, a British scientist known as the 'Godfather of AI', said that he was retiring from the field and now regretted some of his achievements.

"It's hard to see," he said, "how you can prevent the bad actors from using it for bad things."

Especially when the bad actors may turn out be the machines themselves. By doing what you ask, AI won't necessarily do what you mean. Let's imagine the governments of the world giving an advanced AI network the task of solving global warming. The machine puts its mind to work and instantly sees that the primary cause of the crisis is human activity. The solution is obvious: No humans, no problem. It covertly commissions a bio-research program, which produces what becomes a super-deadly virus to kill us all.