A computer named AlphaGo has beaten Grandmaster Lee Sedol at the game of Go. As the machine wowed spectators around the globe, it also raised some questions. Like how advanced are AI technologies already? And when will they start to have commercial applications?
The AlphaGo computer is now officially the Number three Go player in the world. But while the achievement is remarkable, many experts are pointing out that there really isn't much new in AlphaGo and that it is based on classical AI technology dating back to the 1980s. Professor Yuan Feng Zheng, chairman of the Department of Electrical and Computer Engineering at Ohio State University, agrees. He says Alpha Go may be great at the game, but that there is a long way to go to marketize the technology.
"AlphaGo is developed to play the go game but again, this Alpha Go only is good for playing go not for anything else so today the success of Artificial intelligence is more specially for a specific task, for a specified intelligence of human beings, not generically to replace a person," said Prof. Yuan Fang Zheng from Ohio State University.
So you mean 20 years ago when IBM developed Deep Blue, and then we have Alpha Go. So those are two different programs. Even though both are artificial intelligence but they are in nature are different?
"Yeah. Alpha Go is definitely a significant improve of the Deep Blue because Deep blue to my knowledge relies more on so called exclusive search. Alpha Go combines search plus more human like algorism," said Prof. Yuan Fang Zheng.
What is a human-like algorithm? It's simple - trying to make computers think and behave like human beings, the basic goal of artificial intelligence. This might have made sense to only a few back in 1997, when IBM's Deep Blue beat Russian Chess Grandmaster Garry Kasparov. In today's society, however, different forms of AI are all around us. For example, my phone can talk to me.
Like a smart phone with many different add-on features, some cannot be classified as artificial intelligence, but others can. Facial and voice recognition and hand-writing on digital devices are just a few examples. Those are so called artifical narrow intelligence, however, because they can only do one particular task. But what about creating an AI with the ability to learn, and learn multiple tasks, as humans can?
"In addition to new technologies, low costs of the computing machines and one important thing is so-called open source. Every intelligence needs substantial amount of time and resources to develop so it is not realistic for one organization to take on the task," said Prof. Yuan Fang Zheng.
Prof. Zheng says that collective efforts should be able to make a machine capable of many types of artificial intelligence. Oxford Professor Nick Bostrom's surveyed hundreds of AI experts three years ago, and found half of them believe the day for this will come in about 20 years. Today in Seoul, however, Mr Lee may have seen the future.
