Jürgen Schmidhuber (1963, Munich) started to be interested in the artificial intelligence when I was a teenager. I dreamed to build a robot that is more intelligent than himself. Some 40 years later presides over the company NNAISENSE —designing robots that learn like children do— and pursuing the same dream. The artificial intelligence advances by leaps and bounds and, with it, the possibility of attaining his desire: “In two decades there will be robots smarter than humans”.
MORE INFORMATION The 10 branches of technology with more job offers Networks 5G and artificial intelligence: China wants to be the superpower
“My intention is to build an automaton that will learn to be smarter than me and resolve the issues that I cannot resolve. When I get I will be retiring,” he says with a smile Schmidhuber, who is also co-director of the Institute Dalle Molle of Research for Artificial Intelligence in Manno (Switzerland). In 1997, this German was the architect of an algorithm called Long Short Term Memory (LSTM), which has revolutionized the processing of natural language by machines and is used by companies such as Google, Facebook or Amazon.
Now your computer in NNAISENSE about making robots that are autonomous and learn as children do: “most of The time do not imitate their parents, but they learn by themselves. See processes, interpreted, and reproduced”. The smaller, ” explains Schmidhuber, invent “your own experiments” to learn how the world works. “For example, to drop their toys look like these fall to the ground faster each time and learn about gravity,” he says. Children can also make predictions about what actions will cause pain or satisfaction: “If a baby is given with a table, you are going to feel pain, and the next time will try to avoid the stroke surrounding the object, or Matadorbet without touching it”.
The robots manufactured by this expert ‘ learn as children to do experiments that will teach them how the world works”. At the beginning, are provided with a series of orders but after that, through systems of machine learning, they themselves learn new skills on their own, according to this German participant in the SHOW by 2018 —the congress of robotics, the world’s largest—.
neural Networks
Schmidhuber gives the example of a robot that learns to run “without any teacher to teach you” in the same way as a child: “A baby may need a year to control their muscles, learn to stand up, walk and run. Neural networks of this robot would also need weeks to do it.” While this humanoid at the beginning it fell to the ground continuously, the expert in artificial intelligence, ensures that “there came a point when I was able to move one foot behind the other, and even to run.”
Jürgen Schmidhuber. KIKE
But this ability to learn as children is not only important for humanoid robots. In another project, Schmidhuber and his team worked with small cars of Audi, which reached 120 kilometers per hour to learn on your own to park: “they Were equipped with cameras and sensors. It is collided with other vehicles and had to try and try again, little by little, they avoided hitting and was finally able to park in difficult situations”.
This expert in artificial intelligence, explains that the key is to equip robots “the curiosity of a baby that tries to do experiments and understand how the world works and the consequences of their actions”. This form of learning, he argues, does not differ much from that of the adults: “a child is A little scientist who does experiments with their toys. Can that 20 years later that child is for example a scientist to do experiments at the european laboratory for particle physics CERN”.
Despite the fact that the robots of the future will be able to learn in a similar way to the people, “will be very different to the humanoid robots that are manufactured today.” “The robots of the future, for example those who go to the Moon or Mercury, do not appear physically to us, will be composed by systems and sensors that humans don’t have, and will be much more intelligent,” concludes Schmidhuber.