16,000 Computers Needed to Spot a LOLCat: DNews Nugget
Google Brain Loves Cats: In the fascinating world of neural networks and “learning” algorithms, it always seems that we’re a heartbeat away from Skynet becoming “self-aware” and Terminators turning against the robots’ creators. But in reality, replicating even the simplest of the brain’s functions is hard and requires a huge amount of processing power.
But for the guys of Google X labs, they may have taken a small, tentative step toward Skynet. They’ve taught a cluster of 16,000 processors to recognize… a cat. But it’s not as frivolous as it sounds. The algorithm running on the cluster taught itself what a cat was and then delved into images pulled from 10 million random YouTube videos to identify which ones featured cats.
Although impressive, the “Google Brain” isn’t a patch on good old fashioned Human Brain: “It is worth noting that our network is still tiny compared to the human visual cortex, which is a million times larger in terms of the number of neurons and synapses,” say the researchers. via The New York Times