Google's Artificial Neural Network of 16,000 CPUs Learns to Detect Cats

Posted on June 26, 2012

Google says its Research at Google team has been exploring an alternative approach to machine learning that focuses on unlabeled data, such as random images fetched off the Internet. Current machine learning technology is difficult to adapt to new uses because it relies on labeled data to train the system.

Google says it has built artificial neural networks, which loosely simulate the brain's learning processes. One of Google's artificial neural networks is spread across 16,000 CPU cores. Google showed the network 10 million stills from YouTube videos to see what it might learn. What the artificial neural network discovered after observing the Youtube stills was cats.

Google says its neural network learned to respond strongly to pictures of cats. Google says, "This network had never been told what a cat was, nor was it given a single image labeled as a cat. Instead, it 'discovered' what a cat looked like by itself from only unlabeled YouTube stills."

Google has a more in-depth report on its unlabeled data machine learning research here. The researchers say, "Our work shows that it is possible to train neurons to be selective for high-level concepts using entirely unlabeled data. In our experiments, we obtained neurons that function as detectors for faces, human bodies, and cat faces by training on random frames of YouTube videos. These neurons naturally capture complex invariances such as out-of-plane and scale invariances."

More from Science Space & Robots