American researchers from Carnegie Mellon University to feed a computercluster with millions of images in the hope that the system through the laying of the bandages a form of artificial intelligence can develop.
The used cluster has the name NEIL gotten, or Never Ending Image Learning. NEIL, that consists of 200 processors has, for the past months, millions of images of the web before been, writes AP. Through improved beeldherkenningssoftware, the system is able to not only shapes and colors to distinguish, but also to make connections.
With the research project, which is partly funded by Google and the American army, attempt researchers a computer system the ‘common sense’ with the possibility of gaining knowledge without the system specific instructions for getting. These associative qualities are, according to ai-researchers it is necessary to artificial intelligence as possible.
The system, after analyzing millions of photos, more than 1,500 objects and 1,200 scenes can recognize. Know NEIL 2500 associations, but a significant portion is incorrect or irrelevant. Nevertheless, researchers NEIL as an interesting alternative for providing knowledge where normally often descriptions in the form of language to be used. In a next step, NEIL fed is going to be with YouTube videos.