By Laura Shin
Posting in Energy
A neural network of 16,000 computers was let loose on YouTube images for three days. What did it learn by the end? How to recognize cats.
Q. How many computers does it take to start thinking like a human?
Q. And what will a "brain" of that many computers do when turned loose on the Internet?
A. Learn to recognize cats.
Well, yes. At least that's what happened when Google created a neural network.
After seeing 10 million images from YouTube videos within three days, the 16,000-computer network, which had one billion connections, began to recognize cats, even though it had never been taught what a cat looked like.
How the neural network learned to recognize cats
The neural network was built by the secretive Google X laboratory, which has also developed self-driving cars and augmented reality glasses.
The researchers, led by the Stanford University computer scientist Andrew Y. Ng and the Google fellow Jeff Dean, then fed it randomly selected images from 10 million YouTube videos. They did not, however, give it any information about any specific objects, cats or otherwise.
“We never told it during the training, ‘This is a cat,’ ” Dr. Dean told The New York Times. “It basically invented the concept of a cat.”
The researchers then tested the network to see whether it responded to certain images more than others. Dr. Ng told NPR that since images of cats are commonly found on the Internet:
We probed around to see if any of the little simulated neurons will respond strongly only to a picture of a cat but not to pictures of other things. And maybe a little bit to our surprise, we actually found one such neuron, and that's when we thought that, gee, this neuron consistently responds to pictures of cats even though no one had told the algorithm in advance to be learning to look for cats.
The neural network appears to work in the same way that scientists theorize the human mind works: Individual neurons in the brain are dedicated to detecting specific objects.
The network taught itself to detect cats. So what?
It didn't only learn to recognize cats. It also learned to recognize human faces and the shape of the human body. But even that wasn't the point. As Dr. Ng told NPR:
The point wasn't to find a cat. The point was to have a software, maybe a little simulated baby brain, if you will, wake up, not knowing anything, watch YouTube video for several days and figure out what it learns. And I'm sure it's learned tons of other things other than, you know, cats. And it's just that cats was one thing we happen to look for and found.
This experiment showed what computers can do when they are clustered together. Such neural networks are helping advance artificial intelligence in areas such as machine vision and perception, speech recognition and language translation.
While this breakthrough shows the potential to improve artificial intelligence with the "deep learning" technique of training computer processors, I can't help but finish this post with ... LOL cats.
Related on SmartPlanet:
- How your cat could be making you crazy
- Glow-in-the-dark cats could shed light on AIDS research
- By year's end, Google glasses to stream info to eyeballs
Jun 26, 2012