Microsoft has taken the wraps off a new artificial intelligence project that claims to perform tasks like identifying a dog or car in a picture up to 50 times faster than existing systems. Project Adam, as it's called, is modeled after the human brain in that it's a large, unstructured neural network that has to be trained on massive amounts of input data. At first it's a blank slate, but when given a few basic rules and a set of millions of images, Project Adam gets progressively better at identifying objects.
After looking at thousands of images of dogs, it might know how to tell a dog from a cat; after looking at thousands more, it can tell breeds; a few more and it can tell them from more angles. It's not a new idea but Microsoft's system learns far faster and performs better than the rest. Adam could eventually analyze food, people's faces, landscapes — if it can be seen, Adam should be able to figure it out. A more technical explanation can be found at the Microsoft Research blog.
Sign up for top Technology news delivered direct to your inbox
IN-DEPTH
New Technique Identifies Human Actions in Videos
A Robot Worthy of a Standing Ovation? There's an X Prize for That
Elon Musk: A 'potentially dangerous outcome' with AI
SOCIAL
— Devin Coldewey, NBC News