We're seeing the beginnings of Skynet, with Intel working on launching technology that mimics the human brain and "learns" of its user, but the Google X Lab has built a one-billion-connection "neural network" that can identify... cats... on YouTube. Yes, cats.
The project did have an aim, where it looked to simulate object recognition by humans, and was able to more than double the accuracy of item identification from a list of 20,000. Using 16,000 cores in 1000 connected machines (is that all?), the system was able to identify objects, all without human supervision.
The technology represents a big departure from current vision-learning methods. The system was fed 10 million images from YouTube thumbnails, at the low resolution of 200x200. After 72 hours of learning, the system was capable of recognizing not only the human face and body, but also cats, a subject seen frequently on video clips.
The improved 15.8-percent accuracy rate is said to be a jump of over 70-percent from the previous state-of-the-art system.