Researchers are trying to create a computer that can think. In their artificial intelligence experiment, they've let a computer do the searching millions pictures to learn, trying to explain on his own what each of them means.
The system exists in Carnegie Mellon University and is called NEIL, from its initials Never Ending Image Learning. In mid-July the computer began searching for Internet photos 24 / 7 and slowly begins to decide how these images relate to each other. The goal of the Project is to create a machine with what we call common sense or the ability to learn new things without someone teaching them.
The project is funded by Google and the Department of Defense (Marine Research Office).
"Every intelligent person has the common sense to make decisions," said Abhinav Gupta, a professor at the Carnegie Mellon Institute of Robotics.
The NEIL system uses computational vision to analyze and detect the shapes and colors of the photographs, as well as to discover the links between the objects on their own without anyone planning it. For example, he has calculated that zebras tend to be in savannas and that the tigers look somewhat like zebras.
In about four months, the network of 200 processors detected 1.500 objects and 1.200 scenes and connected the dots to make 2.500 associations.
Some of the compounds that NEIL created are wrong, such as "the rhino may be a species of antelope," or something more bizarre, "the actor can be found in a prison cell".
But as Gupta pointed out, a computer that makes its own connections is a totally different kind of challenge from a planned supercomputer that can do something specific very well, or quickly. For example, 1985, researchers from Carnegie Mellon programmed a computer to play chess. 12 years later, a computer defeated world chess champion Garry Kasparov.
Catherine Havasi, an Artificial Intelligence expert from the Massachusetts Institute of Technology, said people are constantly making decisions using silent assumptions, while computers do not do that.
But Gupta seems to be quite pleased with the initial progress. In the future, NEIL will analyze huge numbers of YouTube videos searching for links between objects.
"When we started the project, I wasn't sure it would work," he said. "That's just the beginning."
The worrying thing about this story is its website Naval Research stating that "the combat environment today is far more complex than it has been in the past" and that "the rate at which data reaches the system λήψηof decisions is increasing, while the number of people available to turn data into decisions is decreasing.”
In other words, computers will be able to make some of the decisions in the wars of the future. The Navy website notes: “In many operational scenarios, the human presence does not constitute choice. "