Machines are good at complex tasks, but not at activities that humans
find simple. The answer is for people and computers to work together
There are those who argue that up to 60% of the work we currently do will, within two decades, be entirely replaced by machines. This process is already happening. The hollowing out of what we understand as work has already seen many medium-skilled jobs disappear as executives’ smartphones replace their secretaries and assistants are made redundant by automatic filing systems and Wikipedia.
The first issue with the problem of future decision-making is well-known to scientists specialising in artificial intelligence. It is best described in Moravec’s paradox: why is it that computers are very good at undertaking tasks that require speed and precision, and which humans find difficult (such as solving mathematical equations, playing chess, or even driving cars), and yet bad at tasks we humans find simple, such as clearing coffee cups from a table?
The psychologist Alison Gopnik studies babies. She states that, today, machines are not as smart as a two-year-old child. Though they may not be able to play chess or drive, two-year-olds can do lots of things that a computer cannot: create theories, make hypotheses, figure out how things work, imitate and – most importantly – learn and be creative. Moravec’s ultimate argument is that computers are great at reasoning, but the often unconscious, sensorimotor knowledge that has evolved in the human brain over billions of years is impossible to imitate.
The second issue is the question of values and morality. We humans make decisions on the basis of our values. Can computers have values? The artificial intelligence expert Stuart Russell argues that it is already possible to programme computers on the basis of utility, in terms of gaining the highest-value outcome. However, programming is still carried out on the basis of the values of the programmer.
