Machines and learn to write as human – Clarín.com

For the first time, they developed an algorithm that can make a machine learn and write as a human being. It was done by scientists from the United States and Canada as a way to get the machines to be more creative.
The achievement was in charge of Brenden Lake, New York University, Ruslan Salakhutdinov, University of Toronto, and Joshua Tenenbaum, the Massachusetts Institute of Technology (MIT), and is published today in the journal Science.
So far, it had managed machines, after much training, they could do some homework. What made the team now is to develop an algorithm that enables computers to recognize and draw visual concepts that are indistinguishable from those created by humans.
“It is a breakthrough from the scientific point of view. You may not have an application available today for people. But what did the researchers from the United States and Canada is important as it came to getting a machine learn to write like a child does. Just seeing the letters, and you can copy and write, “he told Clarin Ricardo Rodriguez, Department of Computing Faculty of Natural Sciences of the UBA. When humans are exposed to a new concept, such as a letter in a stranger alphabet just they need some examples to understand its configuration. However, the machines need to be given hundreds or thousands of examples to operate accurately. With the algorithm it will allow greatly shorten the time it takes for computers to “learn” new concepts. This advantage will expand its application to more creative tasks. “These machines could help care for elderly or persons with disabilities,” said Monica Paves, director of Robot Group, Argentina the only company that makes robots for education.
“As early as the fifties, it is looking to learn how the human machine. There are artificial systems that approximate their learning skills: IBM’s Watson computer that feeds on many data and today provides services for the care of cancer patients, “noted Riccillo Marcela, doctor of sciences Computation. In the future, artificial systems are tools that will give more suggestions based on their analysis, but human judgment will always be valuable and should predominate in making decisions. “