Once upon a time, we might have believed that on the Internet, nobody knows you’re a cat. Unfortunately for all the felines masquerading as humans, that’s no longer true.
Several years ago, a team of researchers at the Google X laboratory began developing a neural network that would allow computers to learn.
The group connected 16,000 computer processors together and let the network go wherever it wanted on the web. The idea was to simulate a human brain’s learning process.
Computer scientist Andrew Ng of Stanford University and Google fellow Jeff Dean led the team, which fed the neural network 10 million randomly selected digital images from YouTube videos, from which it taught itself to recognize a cat.
“We never told it during the training, ‘This is a cat,'” said Dean. “It basically invented the concept of a cat.”
A computer network teaching itself how to identify a cat probably sounds frivolous. But when you look at the big picture, the findings are pretty important.
The current vision technology used by computers is created by having humans supervise the learning process by labeling parts as needed. In Google’s research, they just threw tons of data at a huge array of processors and let the software learn by itself.
Speech recognition, image search, and machine-language translation could be greatly improved by this new process.
But the Google X team, being composed of pragmatic and logical individuals, isn’t ready to pop the champagne corks just yet. Although the research has caused a huge forward leap in artificial intelligence and machine learning, “my gut feeling is that we still don’t quite have the right algorithm yet,” according to Ng.
All right then. Play him out, Keyboard Cat!
(In a reader? Watch the video here.)
Source: New York Times