Scientists from Google X Lab, who are responsible for Glass Project, and from Stanford University have caused yet another breakthrough in building a Neural Network by connecting 16,000 computer processors to do something unexpected.
The neural network has more than a billion interconnections with each simulating human brain. To put their innovation to a test, the team of scientist handpicked, but randomly, 10 million videos from YouTube and held back to see how this network performs.
It outdone the expectations of everyone by doubling its accuracy (15.8 percent) as compared to the previous networks of such kind. Out of 20,000 different images it recognized the face of cat with zero training from its makers and that was a nail biting moment for all.
We never told it during the training, ‘This is a cat,’ it basically invented the concept of a cat. We probably have other ones that are side views of cats. said Dr. Dean from Google.
The researchers are to present their findings this week at a conference in Edinburgh, Scotland.