Subscribe
E-Paper
 |
| / Graphic by Cho Sang-won |
The machine appears to be smirking behind the computer screen after scoring two straight victories in the five-game series against the go grandmaster.
Google's DeepMind team, the creator of the artificial intelligence (AI) program, said the machine played some "creative" moves in the second game.
Regardless of the outcome of the remaining three games — of which Lee said he wants to win at least once — AlphaGo has already set a milestone in the development of AI, underscoring the potential of AI to affect the future of humanity immensely.
It demonstrated the ability to apply what it has learned in a creative way.
"Unlike previous go programs, AlphaGo makes the right moves based on deep learning," said Ahn Sung-won, manager of the new technology diffusion research team at the Software Policy and Research Institute.
As the complex board game has a near-infinite number of potential positions, it is impossible to calculate them all, even for computers. Professional go players use their intuition — the realm once believed unreachable by AIs.
However, AlphaGo showcased the possibility of AIs evolving into something far smarter — just like or superior to humans.
It is nearing human intuition through deep learning, or a kind of machine learning method in which machines autonomously analyze massive data.
"Even if a human studies 1,000 games a year, he or she wouldn't be able to master 160,000 games in their lifetime," Ahn said. "For AlphaGo, it took only five weeks."
Such an advanced AI is based on abundant resources for calculation — AlphaGo, which also beat Europe's go champion Fan Hui, used 1,202 CPUs and 176 GPUs.
Experts also note the amazing pace of development.
"An AI program beat the world's greatest chess masters 19 years ago. As go is subtler and more complex than chess, it was expected it would take much longer for an AI to beat a human being, but it's happened now," said Son Young-sung, a principal researcher at the Electronics and Telecommunications Research Institute.
He said that AlphaGo is significant as it shows that technological development is accelerating.
"The time is nearing when it will have a major impact on society," Son said. "AI is surpassing boundaries previously deemed impossible."
Experts agree that AI will affect the future of humanity significantly, though it's too early to fathom how.
Jeong Doo-seok, a senior researcher at the Korea Institute of Science and Technology (KIST), said AlphaGo proved that machine learning is rapidly developing and it will be applied to diverse fields that will affect people's lives in good ways.
"What's significant is that deep learning enables wider applications that reach other areas of technology," Jeong said. "AlphaGo apparently made a big stride in the AI industry. It seems unavoidable that AI will be able to carry out tasks and behave like human beings in the future."
AlphaGo has the ability to learn, practice and study.
Lee Seong-whan, a professor at Korea University, said AlphaGo showed that AI could be smarter than humans in certain areas through deep learning.
"It surpassed specialists in the specific area of go, which people believed to be reserved for humans," he said. "What we currently see, such as deep learning and recognition, is ‘weak' AI, while ‘strong' AI includes creativity, understanding of abstract concepts and judgment."
With the current pace of AI development, the advent of cognitive and humanlike machines that will outperform humans in many tasks is not long in coming.
The AI revolution is in the making and the application of AIs will help resolve many real-world problems. But the emergence of more intelligent machines could cause dire consequences such as the elimination of jobs for human workers.
As depicted in many sci-fi films featuring AIs, whether these intelligent machines will be a blessing or curse for humans remains a big question.
