The project works by creating a database of millions of images with no context. A 200 processor super computer is given free reign to move through those images, without any associated data to help it interpret what it is seeing. Those on the project hope that the computer will begin to find patterns and develop its own interpretation of those images, given enough time.
But it is more than just finding associations. The researchers want to see the computer develop “common sense”. From there, they hope that it will be able to learn from that common sense, and essentially begin to think on its own.
Any intelligent being needs to have common sense to make decisions,” robotics Professor Abhinav Gupta explained. “When we started the project, we were not sure it would work. This is just the start.
In fact, it is working. NEIL has made 2,500 associations since it began four months ago. These associations were developed through the identification of 1,500 objects, and 1,200 scenes. Not all of these associations are correct, but many of them are. It shows that the process is moving forward, and that these could be the first tentative steps in creating A.I.
What is probably most interesting about this story isn’t the research itself, but who is funding it. Google is at the head of funding on the corporate side, showing their interest in such technology that could have major implications for future products.
On the military side, funding has been coming in from Naval Research. It isn’t hard to see how use of artificial intelligence could be beneficial in a battle setting. Especially with the military having always been interested in technology that limits the number of soldiers put directly into the line of combat. With a computer that can base its decisions around elements on its environment, it would reduce the need for human interference, and so risk.
However, neither Naval Research or Google are talking about what their plans are for the project if NEIL is successful. We can only continue to speculate as we keep an eye on what is one of the most fascinating research efforts in modern robotics.