London: You might expect an artificial brain made up of 16 000 computers to have rather more sophisticated interests.
But when Google let its latest cyberbrain loose on the internet, it did what many humans do when looking for light entertainment – it homed in on cats.
And while it may seem a frivolous activity, the scientists say it is a major breakthrough.
In the past, humans have supervised the process by which such computers identify objects by labelling certain features to give the machine a reference point. But in the latest experiment, at Google’s secretive X laboratory in Mountain View, California, the computer brain was given no help at all.
After being fed 10 million random YouTube images, it taught itself to recognise a cat, assembling its own digital picture from those it had been “shown”.
Significantly, this suggests that if a computer is big enough, and programmed correctly, it can make sense of random, unlabelled data without human help.
Dr Jeff Dean, who headed the study, said: “We never told it during the training: ‘This is a cat.’ It basically invented the concept of a cat.”
The researchers believe they have created a computer that replicates what goes on in the human brain’s visual cortex.
“You learn to identify a friend through repetition,” explained Gary Bradski, a neuroscientist at Industrial Perception in California.
The researchers will present the results of the experiment to a conference this weekend in Edinburgh. – Daily Mail