One seemingly small step for mankind, one giant leap for artificial intelligence – Computer beats the World Champion in the ancient game of Go
Although on March 13the newspapers took little or no notice of it, preferring to focus on the economic problems of Swiss hydropower operators, the crude idea of extreme right wing Christoph Blocher to subsidize nuclear power plants, or the doping confession of a Russian tennis glamor girls, what happened the day before truly made history. The extraordinary event might have not made it into the next day’snewspapers, but 25 years later will surely be found in the history books (sadly, those who have made it the purpose of their professional lives to explain the world to us have little to say when it comes to one of the most important mechanisms and causalities of our modern living conditions: progress in the natural sciences).
It was in the end just a victory in a board game, but this win possessesa great symbolic power and groundbreaking importance for a technological development that could dramatically shape our future in the coming years: The computer “AlphaGo” defeated the human world champion in the traditional Chinese board game „Go“. Had the Go professionals before the game been predominantly confident that the 33-year-old exceptional player Lee Sedol would beat the machine, to the surprise of most commentators the best human player appeared to have no chance against the artificial intelligence. Now we might wonder what is so special about a computer defeating the world’s best human player in a board game. This happened in chess almost 20 years ago when the computer „Deep Blue“ won against the world champion Gary Kasparov. And have computers not been able to crunch numbers faster than humans since the early days of the pocket calculator? However, this victory of the machine means much more than an invincible chess computer, as until recently Go had been considered too complex for any artificial intelligence. It comes with a number of possible positions and hence potential game developments that is magnitudes higher than for chess, so a “brute force” calculation of every possible move, as chess computers are mostly programmed to do, represents no useful strategy. Many top Go players play more intuitively which makes their moves very difficult to predict for a computer. In other words, while chess computer can simply put immense computing power on the table and thus predict the course of a game in a way that they are destined to be superior to man, the new Go software imitates human behavior and his creativity explicitly. In yet other words, like the best human players it plays according to inspiration and intuition.
For this very reason, computer chess is nowadays not a big research topic anymore, as the scientists do not expect from it much further insights into the human thought processes and its creativity. Unlike for Go: Since March 12, 2016 the Go-experts talk about „ingenious moves“ on the part of artificial intelligence, which no human ever expected, much less would have made. These are moves that only 50 moves later show their decisive impact, and thuscan by no means be the result of pure computing power. This is already reflected within the structure of the software: Unlike chess computers AlphaGo is based on learning – one is tempted to say: living – neural networks, which follow the architecture of the human brain and are trained imitating the learning and thinking processes of its human counterpart, which improves their abilities constantly over time (referring to the complex structures of neural networks the scientists speak of „Deep Learning“). What indeed shocked the Go-professionals is how much Alpha Go has improved its strength further since last October when the software already beat the European Go Champion. By training, initially by being fed with past human matches and then by continuously playing against itself, it has in only five months’ time been able to achieve a much higher level of mastery.
This success of the machine puts into the spotlight a scientific development, which is likely to rise into new technological spheres in coming years: artificial intelligence („AI“). To provide some background:Computer scientists generally distinguish two types of AI and assign to those the attributes „strong“ and „weak“. The „strong AI“ is a form of intelligence which similar to human beings performs acts of creative thinking and is thus able to solve problems accordingly. The „weak AI“ in contrast is about solving concrete problems by the means of mathematics and computer science, such as recognizing patterns, process language, play chess, detect web pages on the Internet, learn and automate simple mechanical processes, or as of latest search specific information from large disordered data sets. While despite years of research the strong AI has to date failed to deliver on its basic promises, the weak AI in contrast has made significant progressin recent years and decades. Is with the new architectures of complex neural networks and improved versions of „deep learning“ algorithmsperhaps finally a new technological development towards strong AI taking shape? The Go experts are definitely impressed by the unexpected creativity and „genius“ of the new software.
Here lies the potential of a groundbreaking technological development. Thus last year leading AI researchers reached out to the public with a (link) „We are ready now“ appeal to pay attention to the latest results in their field. In the past two decades machine learning has made „dramatic progress”, which is why AI is one of those technologies that will drastically change our lives. Do most of us not already leave sizable data tracks behind us, and are we not constantly encouraged to divulge more data? Are sensors not already able to measure in real-time almost any aspects and expressions of our lives? With the help of AI algorithms these data can be made visible to the creators and owners of these algorithms (the Googles and Facebooks of this world), which will give the munprecedented power. They will be able to recognize or potentially manipulate people’s behaviors and opinions, to the extent that they could sooner or later even identifylarger political or social events before these take place. And at last, the future military infrastructure and weaponry will be heavily based on AI. For example, the NSA has already categorized and scored tens of millions of people in Pakistan on how likely it is that they are planning a terrorist attack (on the basis of these data 2,500 to 4,000 people have been killed, including as it seems Osama bin Laden).
Dramatic changes in our living conditions are developing due to progresses in numerous other technologies, as the author of these lines has hardly grown tired of emphasizing. That this is illustrated to us by something as entertaining as a board game is fabulous, if it helps us to grow aware of the dramatics of the underlying developments. As in any case exciting and at the same time scary connections are to be recognized between our all future lives and the scientific and technological progress, which is taking shape in front of our eyes.