As it happens, my cousin's cat has had kittens and we now have daily squabbles at home over which of the four kittens is the cutest. That question is quite relevant, because it determines which of the kittens will become our cat. The pictures I posted on Twitter - 'Which one is the sweetest: the black, the striped, or the grey?'- resulted in so many contradictory responses that we have as yet been unable to reach a decision.
Humans (some weird exceptions excluded) love cats. Even my gloomy friend, who is a data scientist, owes his comfortable income indirectly to the popularity of our miaowing friends. Research into artificial intelligence was given a strong impetus after 2012, when Google announced that a network of a thousand computers had taught itself to recognise cats. The system had been fed ten million random images taken from YouTube. So many cats were on these images that, by looking at them, the computer learned to identify the main characteristics of cats. In the same way, the system learned to identify humans as well, even better than cats. Yet, what stuck was the idea of a computer that recognises cats.
By now, artificial intelligence is able to beat a human world champion at a game of Go. Within a few decades, a professor of law and computer science told me recently, computers will be able to make a more accurate judgement than lawyers. However, even assuming that computers do indeed become just as intelligent as my gloomy friend and the likes of Elon Musk predict, why would these computers treat humans as pets? What neural network would autonomously reach the conclusion that petting humans calms and cheers up a robot? And would this be desirable, or not?
These questions lingered after the gloomy friend had rung off and I put my phone into my pocket. Fow a few minutes the clouds looked even darker, the rain felt even wetter and my hands were colder. Until I received an app from my cousin with a photo of the kittens. I think we will go for the grey one.
