The impact of deep learning
By tAtSuYA KiMurA and YOMiuri SHiMBun
THE AI ( artificial intelligence) technology called deep learning has the potential to bring about revolutionary changes to society.
The self- learning technology is already being used not only for games but also for medical diagnoses, self- driving cars and controlling robots.
Until recently, it was difficult for conventional computers to recognise, say, a cat in a picture, something humans can easily do. In the past, to enable a computer to instantly recognise a cat, it was necessary for programmers to input detailed features of a cat.
Now, deep learning allows computers to overcome this weakness. By scanning large quantities of data, computers can rapidly learn by themselves to recognise features of cats – such as their whiskers and ears – with an accuracy that surpasses that of humans.
By analysing 160,000 games played by professional Go players and recording the 30 million moves in games played by human experts, the computer program AlphaGo learned moves that enabled it to defeat a professional player.
When a human player places a Go stone, AlphaGo recognises where the stone was placed and then determines its next move based on what will give it the highest probability for victory.
Not only does AlphaGo monitor the entire board and predict the patterns of how a human player will respond to its moves, but it can also make effective countering moves even in the event of unexpected developments in a match.
Though computers cannot immediately transfer the experience they obtain by playing Go to other fields in the way human beings can, there is no doubt that deep learning will be applied even more in such fields as image and voice recognition in the future.
Another application of the technology is expected to enable self- driving cars to avoid hitting pedestrians and other cars, with Toyota Motor Corp undertaking joint research with a domestic start- up company.
Deep learning technology also can be used for medical diagnoses to spot cancer, among other medical problems, in images. Also, security software using deep learning can help detect suspicious persons in street images.
It will play a big part in the future as companies and governments have stepped up efforts in AI research. – Japan News/ Asia News Network
— The Yomiuri Shimbun