Google has created an ‘artificial brain’ from 16,000 computer processors, with more than a billion connections. The team led by Google’s Jeff Dean then fed it random images culled from 10 million YouTube videos and let it ‘learn’ by itself. Surprisingly, the machine focused in on cats. “We never told it during the training ‘this is a cat’,” said Dean. “It basically invented the concept of a cat.”
Contrary to the widely-held intuition, experimental results reveal that it is possible to train a face detector without having to label images as containing a face or not. They also find that the same network is sensitive to other high-level concepts such as cat faces and human bodies. Starting with these learned features, they trained their network to obtain 15.8 per cent accuracy in recognising 20,000 object categories from ImageNet, a leap of 70 per cent relative improvement over the previous state-of-the-art.
The ‘brain’ was a creation of the company’s ‘blue sky ideas’ lab, Google X, reportedly located in Google’s Mountain View, California, headquarters — known as ‘the Googleplex.’