Design robots that can learn from their mistakes
German scientists have developed robots that can learn from their mistakes, thanks to complex algorithms included in its programming. In the twentieth century, prominent writers, exponents of the genre of science fiction, addressed the issue of the fine line between human and artificial enough to remember the replicants of Philip K Dick, the questioning of Brian Aldiss automatons or robots rebel Isaac Asimov. Well, apparently, in addition to exhibiting these authors a great narrative talent, showed a unique ability to advance. And, recently, some of her thematic approaches seem to have become reality.
This is an example of this: engineering from the Technical University of Munich, using sophisticated algorithms and advanced technologies have managed to build robots capable of learning from their mistakes.
A breakthrough:
This development in the field of robotics engineering, it is still so evocative and disturbing, as a revolutionary. For a common robot being designed today, the average guy, solve a complex chess move or a specious mathematical problem is a simple activity compared with the action of a sausage in a frying pan. However, this paradoxical situation seems to be close to positively changed, thanks to software created by scientists in Germany.
Errors sobering:
Experts from the Technical University of Munich have developed some algorithms of great complexity, capable of making a robot can learn from their mistakes, and so gradually be able to perform activities that people perform without problems. The purpose of these algorithms is to encourage faculties to perceive robots, handling and reason. So, little by little, the robots can learn from their mistakes and thus, there is a step in the still far-attempt-to achieve self-conscious automata. So that the horizon of robotics engineering is evident the ideal goal: to provide robots precisely that which distinguishes human beings.
Action learning:
Michael Beetz, professor of computer science and one of those involved in the project, told Reuters that these robots able to learn, they can anticipate the effects of their actions before proceeding to perform them. Beetz figuratively exemplifies the objective of this project, see these robots, as fit to raise a hot-cake a pan without damaging the hot-cake. One of these robots is TUM-Rosie, which, thanks to control systems and models of computation present in your programming, you can understand the function of a spoon. Another of these robots, TUM-James, using similar mechanisms and schedules, you can cut pieces of perfectly bread.
