Liu cixin: artificial intelligence won, human fears? -Artificial intelligence, AI-IT information
"Alfa Laval dog" and Li Shishi go 4:1 to end the war. With man-machine war, in addition to issues relating to human dignity, is the fear of the rapid development of artificial intelligence. This fear is not totally unreasonable, but might be wrong.
Like science fiction or movies described as artificial intelligence will "rise" and eventually rule the human, it is the fear shared by many people who care about this issue. The computer came shortly, calculation speed is much faster than a man, and now artificial intelligence in the game of go against human, only quantitative rather than qualitative changes, because the original calculation, retrieve its domain. In other areas, artificial intelligence is very slow. Human is easy to do a lot of things, is basically a computer can′t do, the most obvious is to understand natural language, computer only that language out and match the database for the corresponding interpretation, understanding of complex language, computers can do. In addition, human intuition, imagination and ability to rely on incomplete information-judge, on today′s computers is very difficult. At present, including "Alfa Laval dog" weak artificial intelligence, and intelligence totally beyond human reach. As for "SkyNet" wake up, just need a few science fiction writers on the line.
"SkyNet"-like artificial intelligence such as does emerge, humans can still respond to. Artificial intelligence cannot be strong without weakness. Human also impossible not to predict a level of strong artificial intelligence is about to occur, so will appear before taking a series of measures, such as don′t let it on hardware connection or with the outside world cut off from all power supply and its no way to live. 10,000 steps back again, even strong artificial intelligence last self to survive and avoid being human eradication, humans may also find a way to live with it, may not necessarily be the slavery, such as the man-machine combination is a way to coexist.
Human to reach strong artificial intelligence stage, there are at least three obstacles have been encountered, they can eventually be overcome, is still unknown. First, our own consciousness, intelligence and brain science is still at an early stage. Number of neurons in the brain are like stars in the Milky Way Galaxy qianyige on, each neuron structure is extremely complex. Currently, scientists believe that human intelligence is probably occurred at the quantum level, and the laws of Physics of the quantum world is much more complex than the macroscopic world. And second, the existing computer performance increases very fast, but the use is 40-von Neumann architecture of the last century, this parallel with our brain structures are completely different. But progress has been slow on the structure of the new computer. Third, more traditional hurdle is computer calculates capacity at this stage are not up to the capacity of the human brain. Computer-based Moore had predicted computers would soon reach the calculated capacity of the human brain. But at present developing trend of IC industry there are slow signs, Moore′s law might be about to fail.
In fact, we should be concerned that artificial intelligence will take over the work of mankind, and this process has already begun. Now, choose goods and payment network sales system operations rely on artificial intelligence. In the foreseeable future, more services, mechanical work will be replaced by artificial intelligence, it may make great changes in our social structure.
Now, therefore, also to the need to consider the relationship between human and artificial intelligence at the moment. Now, artificial intelligence and human relationship both partners, friends and competitors. Nuclear energy initially as weapons of terror could destroy the world, but finally beat crazy human reason, nuclear energy is now locked into a cage, became a human partner. Artificial intelligence, too, developing artificial intelligence must be at risk, but any technology development at risk, human beings do not develop as a result. Artificial intelligence is a hope for the development of human civilization, but if the development process is in danger, that we need to face together, and try to eliminate it.
刘慈欣：人工智能赢了，人类该恐惧吗？ - 人工智能,AI - IT资讯
“阿法狗”和李世石的围棋大战最终以4：1落幕。伴随人机大战而起的，除了有关人类尊严的话题，就是对人工智能快速发展的恐惧。这种恐惧并非完全没道理，但可能方向错了。
像科幻小说或电影描述得那样，人工智能将“崛起”并最终统治人类，这是许多关心这一问题的人共有的恐惧。计算机刚出现不久，计算速度就远比人快，现在人工智能在围棋比赛中战胜人类，也只是量变而非质变，因为原本计算、检索等就是其擅长领域。而在其他领域，人工智能进展十分缓慢。很多事人类做起来易如反掌，计算机却基本上做不了，最明显的是理解自然语言，目前计算机只能做到把语言抽出来和数据库相匹配进行对应式理解，对于复杂语言的理解，计算机也无能为力。此外，人类的直觉、想象力和依靠极不完整的信息做出判断等能力，对目前的计算机来说非常困难。目前，包括“阿法狗”在内的弱人工智能，要想在智力上全面超越人类还遥不可及。至于说“天网”觉醒，只需要科幻作家考虑就行了。
“天网”般的强人工智能如真能出现，人类依然可以应对。人工智能不可能强大到没有弱点。人类也不可能预测不到某种级别的强人工智能即将出现，所以必然会在其出现之前采取一系列措施，譬如不让它与外部世界有硬件上的连接或是断绝所有能量供应，让其没办法生存等。再退一万步，即便强人工智能最后自我生存下来，避免被人类消灭，人类也可能找到与之共存的方法，并不是一定会被其奴役，比如人机结合就是一个共存的途径。
人类要想抵达强人工智能阶段，至少还面临三道技术障碍，最终它们能不能被克服，目前仍不得而知。第一，我们对自身意识的产生、智力以及脑科学研究还处于初级阶段。大脑神经元数量就像银河系的星星一样有上千亿个，每一个神经元结构都极其复杂。目前，有科学家认为人的智力过程可能是在量子层面上发生，而量子世界的物理规律比宏观世界的要复杂得多。第二，目前计算机的性能提升速度很快，但使用的还是上世纪四十年代冯·诺依曼结构，这与我们并行的大脑结构完全不同。而对新结构的计算机研究进展却很缓慢。第三，更为传统的障碍是现阶段计算机总体计算容量还没达到人脑的容量。基于计算机摩尔定律，人们曾预测计算机不久就会达到人脑的计算容量。但目前集成电路工业的发展趋势已有放慢迹象，摩尔定律可能行将失效。
其实，我们更应担忧的是人工智能将抢走人类的工作，而这种进程已经开始。目前，电子商务中的选货、付款等网络销售系统操作过程都需依赖人工智能。在可见的未来，更多的服务性行业、机械性的工作都将由人工智能替代，它可能让我们的社会结构产生很大改变。
因此，现在也到了需要考虑人类与人工智能关系的时刻。现在看来，人工智能和人类兼具伙伴、朋友和竞争对手的关系。核能最初作为恐怖武器曾可能毁灭世界，但最后还是人类的理性战胜了疯狂，现在核能基本被关到笼子中，成为了人类伙伴。人工智能也是这样，发展人工智能肯定面临风险，但任何技术发展都面临风险，人类不会因此而不去发展。人工智能是人类文明发展的希望，但发展过程中如果有危险，那我们必须共同去面对，并努力消除它。