The news about Google AlphaGo vs. Lee Sedol surprised many people worldwide. Google AlphaGo is a computer program developed by Google DeepMind lab to play the board game Go. Lee Sedol is a South Korean professional Go player and he is ranked Top 3 in the world. Before the contest, many professionals expressed their point of Google AlphaGo could not beat this professional player. However, a 4: 1 final score was quite dramatic.
Go is one of the oldest and the most complex board game play to
day. It has more than 2*10^170 legal way to put a “stone” (which is the white or black pieces in the above figure). The big bang (or the age of the universe) was 15 billion years old or 5*10^17 seconds! That means, no matter how fast a computer is, this computer cannot predict the result of the game and beat a professional human player for a limit time. However, AlphaGo does. AlphaGo has different strategy decision functions, and those are the computational functions to decide the next step of the game. The core function is the deep learning. Deep learning means AlphaGo can study the game like a normal human being. AlphaGo can obtain the experience from previous games. Hence, it looks like AlphaGo will be able to think and make a decision as a human.
So, what does this news mean?
Artificial Intelligence (AI) is “smarter” and it starts to evolve. The threat of AI becomes a stronger voice, and these people who believe the threat is true also think the movies I, Robot, or The Terminator will finally happen.
But, I don’t believe this threat theory.
An article, Where Computers Defeat Humans, and Where They Can’t, New York Time, has addressed the self-learning power of AlphaGo. AlphaGo’s victories illustrate the power of a new approach, which instead of trying to program smart strategies into a computer, it builds systems that can learn winning strategies almost entirely on their own by seeing examples of successes and failures. Such ability may be the reason for people to believe the threat of Artificial Intelligence and worry about if AI will be out of control with a self-conscious mind and has an against ideology – against the human being, the founder and the master of AI. With such high development and assumption, even Steven Hawking expressed his concern about Artificial Intelligence during the interview of BBS on Dec. 2nd, 2014.
However, I do not believe this threat theory. In an article entitled with To Beat Go Champion, Google’s Program Needed a Human Army, Melanie Mitchell, a computer scientist at Portland State University and the Santa Fe Institute indicated the biggest difference between human being and a computer program – the pattern recognition. Computer programs have their limitations: the goal and method. Since programs performance computing with binary, 0 and 1 only, and following the certain rules, they will only do what human wants to. For example, they cannot use the same rules and program to play different board games.
Yes, the true reason of Artificial Intelligence cannot defeat humans is the different ways to realize and solve a problem. The developed AI cannot solve creative problems! But we can. We can solve the problem A by studying and using the experience from problem B even if A and B are 100% different, because humans can always make some connections between two things no matter how tiny this is. But AI doesn’t. Artificial Intelligence cannot solve the problem of Go from the result of chess. Also, the way of the human brain working is far more different from the way of the computers are. In spite of mimicking how the human brain works, AI is still an algorithm designed for machines. Human brains store or pass information through countless neurons with complicated chemical changes, which is of great importance for signals, whereas AI is much simpler than that.
Artificial Intelligence is not to create a new brain. It seems like that machines are “thinking” and then making decisions, but all they do is just following established rules and performing calculations. Artificial Intelligence is still computing, not “thinking”, in high-speed with some rules. Machines with high-level Artificial Intelligence can only proceed under the pre-set rules. No matter how “smart” AI is, it has no ability to create rules of its own. But, the human can create the rules, break the rules, and change the rules.
Regardless, the victory of Google AlphaGo represents a new level of Artificial Intelligence. But, the advantages of it should be more noticed by people, instead of worrying it become a threat to humans in the future. Through the victory of Google AlphaGo, researchers can now confirm that computers possess the ability to make the optimal decision with highly developed algorithms and performing huge computations. In fact, I think the victory of Google AlphaGo relies on decent algorithms and high-quality hardware. People can take advantages of the benefits. The essence of Artificial Intelligence is to study how to make computers perform tasks that used to be done only by people. We control AI. Artificial Intelligence is not a threat! Instead of performing with instinct like humans do, AI is not “smart” enough! So, don’t worry about your coffee maker will try to kill you in the future.