Machines Like Me

by Ian McEwan

‘Autonomous’ was never the right word, for the new cars were as dependent as newborn babies on mighty networks of computers linked to satellites and on-board radar. If artificial intelligence was to guide these vehicles safely home, what set of values or priorities should be assumed in the software? Fortunately, in moral philosophy there already existed a well-explored set of dilemmas known in the business as ‘the trolley problem’. Adapted easily to cars, the sort of problem manufacturers and their software engineers now posed was this: you, or rather, your car is driving at the maximum legal speed along a narrow suburban road. The traffic is flowing nicely. On the pavement on your side of the road is a group of children. Suddenly, one of them, a child of eight, runs out across the road, right into your path. There’s a fraction of a second to make a decision – either mow the child down, swerve onto the crowded pavement, or into the oncoming traffic and collide head on with a truck at a closing speed of eighty miles an hour. You’re alone, so that’s fine, sacrifice or save yourself. What if your spouse and your two children are in the car? Too easy? What if it’s your only daughter, or your grandparents, or your pregnant daughter and your son-in-law, both in their mid-twenties? Now, take into account the occupants of the truck. A fraction of a second is more than enough time for a computer to give thorough consideration to all the issues. The decision will depend on the prior...

Asimov’s laws of robotics