Google's artificial brain learns to recognise cats

By Raj Saxena Published Date
09 - Jul - 2012
| Last Updated
09 - Jul - 2012
Google's artificial brain learns to recognise cats

What do you get when you give a million people access to YouTube? They watch cat videos. And what do you get when you give a cybernetic brain made up of 16K processors access to YouTube? It learns to recognise cats! Notice any similarity? Cats have not only taken over the interwebs but now even artificial intelligence has fallen prey to their feline charms.

This “deep learning” experiment being conducted by Google and Stanford University researchers at their top secret lab in Mountain View, California is a result of Google’s Blue Sky research. The cybernetic brain was shown ten million random images from YouTube and the sixteen-thousand-processor-rich computer learned to recognise cats all by itself.

The researchers made it very clear that at no point during the training was the computer instructed that “this is a cat”. Scientists say that this is a major breakthrough as the computer basically invented the concept of a cat.

One of the neurons in the artificial neural network, trained from still frames from unlabeled YouTube videos, learned to detect cats.

The ground breaking research also has been believed to have resulted into activities similar to those happening in the cerebral cortex of the human brain. Just for the record, the artificial brain has not only learnt to recognise cats, it also can now recognise human faces, body parts and probably other images as well.

Andrew Y. Ng, a computer scientist at Stanford University who lead the research with Google Fellow Jeff Dean said that the results were similar to the ones expected if a toddler was shown YouTube videos for a week. The baby wouldn't be able to tell you that it saw a cat but the idea of a cat would be imprinted in its brain. This new find can be used by Google in its most basic form for Search Engine Optimization.

Researchers point out that they worked with 16K cores to work out a model with more than one billion connections. The above figure may be impressive but its nothing when compared to an adult human brain’s hundred trillion connections.

The next step in this feat will be developing a more powerful unit with higher number of processors to enhance the capabilities of this find. Who knows, one day we might finally have robots making discoveries for us!

Can you think of some application of this new find? Share your comments below or join us on our Facebook or Twitter page.