Is it a key step for artificial intelligence?
Like the victory of computer IBM's Deep Blue against chess champion Garry Kasparov in 1997, the unexpected triumph of AlphaGo designed as circuits human neurons, is a new success for artificial intelligence, obtained much earlier than expected . This demonstrates its ability to self-learning and even random exploration of possibilities, a technique that, according to some researchers, approaches the imagination.
"It's a feat. We had hoped that one day the machine defeated man, but not so fast. It is a milestone in the evolution of the machines, as the defeat of Kasparov against a computer," said Jean-Gabriel Ganascia, specialist artificial intelligence at the Pierre and Marie Curie University in Paris.
"This achievement gives us an indication of the development of deep learning (deep learning or ability of a machine to learn alone) and the computing power that makes this possible," estimated meanwhile Nick Bostrom, the Institute for the Future of Site of the University of Oxford.
AlphaGo, who has played millions of games against itself, "seems to have managed to break the game, which would be a crucial step, but unfortunately we do not know exactly what it does," said Stuart Russel, another expert on artificial intelligence at Berkeley.
Is it a defeat for man? Is it justified to be afraid?
Torn between enthusiasm and caution, the researchers considered necessary to develop research to ensure that artificial intelligence will remain under human control. In late 2014, the British physicist Stephen Hawking expressed fear that longer-term "development of a complete artificial intelligence can wipe out the human race."
"This is not the beginning of the end of humanity, as long as we decide we want a safe and beneficial artificial intelligence and not only efficient," he said Anders Sandberg, University of Oxford.
"I think, luckily, Deep Mind (artificial intelligence subsidiary of Google) shares this view, but still have to carry out many more research so we can rely on our artificial intelligence," he added.
"The methods of artificial intelligence progressing much faster than expected, raising more urgently the issue of long-term goals. Will be necessary to develop a research discipline completely new to ensure that systems increasingly powerful remain under human control, "confirmed Stuart Russel.
What scientific perspectives open now?
Artificial intelligence and has applications of all kinds, GPS, voice recognition or image processing, but compared to Deep Blue, "AlphaGo is more interesting because it uses potentially more universal algorithms and their approach could be applied to numerous games" , said Nick Bostrom.
In addition, the media impact of what happened "encourages manufacturers to be tested methods of artificial intelligence, for the sake of technology, which is good for everyone," Olivier Teytaud, researcher at the French Research Institute enthused computer.