Onlookers will watch a way: go watch man-machine war-Li Shishi, go, AI-IT information

Beijing time on March 9, a by far the most important game of "man-machine war" in Korea Seoul staged game is Chess World Championship, Korea star Li Shishi section, the other party to Google artificial intelligence program developed a LPHA g o (a g). 5 the parties will contest, the winner will win a $ 1 million bonus. What is Li Shishi of the last 10 years has the most world championship titles won, there is a 1200 CPU g rule the roost? Suspense was too large.

Prediction of Li Shishi WINS, of course, is long or long; think winning a g, you are short or short.

In animal farm, George Orwell said, the real struggle is between animals and humans. However, since the summer of 1956, John McCarthy and Edward Albert Feigenbaum and other men in the United States presented at the Dartmouth artificial intelligence Conference, board games and through a program to enable computers to complete other tasks, but also of human intelligence can be a variety of features and learning in all aspects of the machine can be achieved. This is artificial intelligence. Since then, real fight of the world have been transformed into artificial intelligence and human struggle.

From dark blue to a g today is artificial intelligence game and struggling with the people. A g is a deep learning capabilities of a computer program, by being bought by Google Inc for 4 billion yuan of United Kingdom d eepM IND company research and development. A g is the most terrible place, "algorithm for multilayer neural networks" and "Monte-Carlo search tree" combined game. This combination not only increase its strength, and can greatly accelerate the pace of a g chess, winning against the people in the speed force.

G the first neural net is a policy network will make viable the hand and sprawling Monte-Carlo search tree will narrow your search. Second layer neural net game where black and white---value net will prompt party advantage, what advantages and tips, and limit the scope of a road. Now, therefore, g a next search when in the game of chess can be shrunk to only 100,000 about 200 million change change makes the game faster.

Although g has such powerful features and a subtlety, but also resist most choose to do more. There are two main reasons. A computer go program like Goldbach′s conjecture in the fields of artificial intelligence, AI now also reached a level not to crack the Goldbach′s conjecture. Second, any artificial intelligence is merely imitation and simulation of human thinking, a-g, too, it moves it had collected only in the light of what has happened and to calculate, and decide how to deal with people a game. If Li Shishi throw a different play or tactics, results may be completely different.

These explanations from shorting seem as if some shortness of breath, according to a g great strength and have over human achievement, a g beat Li Shishi is not impossible. October 2015 5-9th, a and g was the career of the chess player Fan Hui, China (second paragraph, now France go team head coach) conducted five competitions, a g 5:0 big score victory over Fan Hui, this was the first artificial intelligence go program to overcome the human go professional chess players. G has demonstrated a strength.

Whether it is long or short, are available from the 1997 IB m deep blue against chess master Garry Kasparov thought one of the most important factors and prediction on artificial intelligence and the attitudes toward artificial intelligence and the way, that′s artificial intelligence can be thought.

Deep blue defeated Garry Kasparov is not the program Super careful, impeccable and algorithms, but both appeared accidental. Accident is the deep blue program appeared in a b u g, cause it was out in the game without the weird logic. This terrified Kasparov, distracted, intuitive feeling that the dark blue that is certainly not to go, dug a hole behind. As a result, thinking the resulting confusion, time out, finally abandoned the towel.

Although Kasparov thought out, but he can think. Deep blue can′t think only according to the established procedure or even a virus changed the program to play chess, so deep blue beat Garry Kasparov was an accident. Now, a g also run from deep blue pattern, they are obviously not thinking. However, researchers in artificial intelligence, neurons in the brain′s storage and pattern very similar to the corresponding change to your computer, the relationship between brain and mind is like the relationship between hardware and software. Once the computer simulated the neurons and the connections between them all, computer simulation of intelligence and intelligence of the brain there is no longer any distinction, and they can think.

However, to date, any artificial intelligence has not yet shown its ability to think like a human. Even a g can beat Li Shishi, may not prove it will think, alone are not worth worrying about. However, today the Li Shishi and a-g games long and short attitude are not fully comparable with the financial markets. The long and short of the latter is not entirely out of spite and good will, but by different means in order to profit. However, the consequences to the shorting of the artificial intelligence in it.

If one day, AI represented by a g like a man thinking, and thinking capacity in people, you have to ask a question, whether malicious elements of artificial intelligence, AI will replace humans dominate human society. If short humans no harm, AI designs need to left hand, both drivers for them, as well as configure them brakes. Otherwise, the artificial intelligence is likely to end in the future humans, humans really want to become its own gravedigger.

围观要看门道：围棋人机大战看点在此 - 李世石,围棋,人工智能 - IT资讯

北京时间3月9日，一场迄今为止最重要的围棋“人机大战”将在韩国首尔上演，比赛一方是围棋世界冠军、韩国名将李世石九段，另一方为谷歌公司研制的人工智能程序A lpha G o（A G ）。双方将进行5盘较量，胜者将赢得100万美元的奖金。究竟是最近10年拥有最多世界冠军头衔的李世石取胜，还是拥有1200个CPU的A G称雄？悬念太大。

预测李世石胜，当然是做多或多头；认为A G取胜，则是做空或空头。

乔治·奥威尔在《动物庄园》里说，真正的斗争是在牲口和人之间。但是，自从1956年的夏天约翰·麦卡锡与爱德华·费根鲍姆等人在美国达特茅斯举行的人工智能研讨会上提出，通过程序使计算机能进行棋类游戏并完成其他任务，而且还可使人类智能的各种特点和学习的各个方面都能够在机器中得以实现。这就是人工智能。从那时起，世界上真正的斗争就已经转变为人工智能与人的斗争。

从深蓝到今天的A G就是人工智能与人进行博弈和斗争。A G是具有深度学习能力的电脑程序，由被谷歌公司以40亿元人民币收购的英国D eepM ind公司研发。A G最厉害的地方是，能把“多层神经网络算法”与“蒙特卡罗搜索树技术”结合起来进行博弈。这种结合不仅使其棋力大增，还能大大加快A G的下棋速度，在与人的较量中体现速度的制胜力量。

A G的第一层神经网是策略网，会提出可行的下一手，然后无限蔓延的蒙特卡罗搜索树会缩小搜索范围。此后第二层神经网———价值网会提示博弈的黑白哪一方占优势，占多少优势，并提示和限制算路的范围。因此，现在的A G可以把在国际象棋博弈时每下一步搜索约2亿个变化缩减到只考虑10万个变化，使得博弈速度更快。

尽管A G有如此强大的特点和精妙之处，但是也抵挡不住多数人选择做多。主要原因有两点。一是电脑围棋程序如同人工智能领域里的哥德巴赫猜想，现在的人工智能还达不到破解哥德巴赫猜想的水平。二是任何人工智能都只是在模仿和模拟人的思维，A G同样如此，它只能根据已经发生的和搜集到的下法来计算，并决定如何与人博弈。如果李世石抛出一个与以前不同的下法或怪招，结果就可能会完全不一样。

这些解释从做空的人看来好像有些气短，按照A G强大的实力和已经打败人类的成绩，A G战胜李世石并非不可能。2015年10月5-9日，A G与曾是中国棋院的职业棋手樊麾（二段，现为法国围棋队总教练）进行了五局正式比赛，A G以5：0的大比分完胜樊麾，这是人工智能的围棋程序首次战胜人类围棋的职业棋手。已足以说明A G的实力。

无论是做多还是做空，都可以从1997年IB M的深蓝战胜国际象棋大师卡斯帕罗夫思考一个最重要的因素，并预测人工智能的前景和提出对待人工智能的态度和方式，这就是人工智能是否可以思考。

深蓝打败卡斯帕罗夫并不是程序缜密、无懈可击和算法超强，而是由于两者都出现了意外。深蓝的意外是其程序中出现了一个B U G，导致它在比赛中走出了一步毫无逻辑的怪招。这一招吓坏了卡斯帕罗夫，让其心烦意乱，直觉认为深蓝这一招肯定不是乱走，背后一定挖了个坑。如此一来，思维就乱套了，时间也没有了，最后弃子认输。

尽管卡斯帕罗夫思维乱套了，但说明他能思考。深蓝却不能思考，只能按既有的程序甚至是病毒更改后的程序下棋，因此，深蓝战胜卡斯帕罗夫是意外。现在，A G也脱离不了深蓝的运行模式，它们显然还不会思考。但是，在人工智能研究人员看来，大脑神经元的存储行为和模式改变跟计算机的相应行为非常相似，人脑与思维的关系则如同硬件与软件的关系。一旦计算机模拟了所有的神经元和它们之间的连接点，计算机的模拟智能就与大脑的智能不再有任何区别，并且它们能够思考。

不过，迄今为止，任何人工智能还没有显示出其能像人类一样思考的能力。即便A G能战胜李世石，也未必证明它会思考，仅此一点就不足为虑。不过，今天人们对李世石和A G博弈的做多与做空态度并不与金融市场完全相似。后者的做多与做空完全不是出于恶意和善意，不过是为了赢利而采取的不同手段。但是，对人工智能的做空却有恶果在里面。

如果有一天，以A G为代表的人工智能像人一样会思考，并且其思考能力更强于人，就得问一问，人工智能的设计是否有恶意的成分在内，因为，人工智能将替代人类主宰人类社会。如果说做空人类没有恶意，对人工智能的设计就需要留有后手，既要为它们安装驱动，还要为它们配置刹车。否则，未来人工智能很有可能终结人类，人类就真要成为自身的掘墓人。