Google's AI learns about cats

26/06/2012

Google dedicates 16,000 computers with more than one billion connections to mimic aspects of the human brain's activity. Using the project Google Brain, the company trained its artificial intelligence (AI) by feeding it 10 million thumbnail images taken from YouTube videos.

The project is led by Andrew Ng, Associate Professor of Computer Science at Stanford University. He led a research team at Google X division in creating the neural network, and in three days learning, the AI learns about cat and its characteristics.

Google cat

The Google scientists and programmers at the team was surprised, not that the internet is full of cat videos, but the simulation performed much better than any previous efforts. The research also concluded that the software-based neural network appeared to closely mirror theories developed by biologist that suggest individual neurons inside a human brain can be trained to detect significant objects.

The process for Google's "cat" machine learning had no human intervention. The AI was just set loose.

"The idea is that instead of having teams of researchers trying to find out how to find edges, you instead throw a ton of data at the algorithm and you let the data speak and have the software automatically learn from the data," Ng said.