[Abstract] With the development of artificial intelligence technology in recent years, the issue of " goodness / evil " in human-computer fusion intelligence has attracted more and more attention from the society. This article expounds several thoughts on the human mind problem in the era of human-computer integration and intelligence. First, it briefly introduces human-computer and its fusion intelligence, and expounds the relevant research background of human-computer fusion intelligence; and then discusses the current era of human-computer intelligence integration. Faced with various difficulties and bottlenecks; finally analyzed the good and evil problems of human heart in the era of human-computer intelligence integration and its impact.
[Keywords] artificial intelligence, human-machine integration, intelligent human mind
The origin and future development direction of human-machine fusion intelligence
Intelligence, including artificial intelligence, are complex systems. Many of these things cannot be explained clearly by logical thinking. There are also a large number of non-linear and non-logical components, interpretability, lifelong learning, dynamic representation, strong and weak reasoning. Analogy is needed, but the mechanism of analogy is far from being solved by simply using science and technology, especially when it comes to emotions, situations, and virtual bodies. Trying to use mathematics alone, especially modern incomplete mathematics, to solve the core problems of intelligence or artificial intelligence is tantamount to seeking fish from trees, drawing cakes to satisfy hunger, and fishing for the moon in the water, just like holding a wooden stick, a rock to build an airplane and a rocket, and the reason is simple: Qualitative real things are still being explored in the foreseeable future, but quantitative ones can only be automation!
The interaction between humans, machines, and environmental systems produces intelligence, which is not only a scientific problem, but also includes non-scientific research (such as humanities, art, philosophy and religion). Among them, humans are complex systems, machines are relatively simple systems, and the environment fluctuates greatly. Therefore, the human-machine environmental system we study has both "determinism" and "randomness", and is a "complex giant system". ". Qian Xuesen believes that human beings have not yet found a general principle and method to solve the "complex giant system", and the theory of human-computer fusion intelligent system may be a useful attempt. However, the inconsistency of time, space and logic mechanism between human and machine is the key to the difficulty of intelligent fusion. The bottleneck of human-computer fusion intelligence is how to achieve rational, advantageous, and regular rhythm and rhythm. For example, in the process of human-machine integration, some problems will become more and more prominent: how to effectively allocate human-machine functions? When, where, and how to allocate humans and machines? When the human and machine speeds do not match, is it better to use the human speed as the basis or the machine speed? How to integrate learning between man and machine? How to integrate understanding between man and machine? How does human and machine integrate decision-making? How does human and machine integrate reasoning? How does human and machine fuse perception? How do humans and machines integrate intentions? And how do data, information, knowledge, intelligence, and wisdom interact and transform? "I Ching" analyzes the movement, "The Tao Te Ching" discusses right and wrong, "Sun Tzu Art of War" talks about fictitiousness and reality, Wittgenstein proves whether or not, Boolean algebra says 0, 1, Turing test is similar, von Neumann coexistence calculation, these What is the relationship between concepts? In digital logic, does the relationship between AND or NOT exist? In non-digital logic, how to define the calculation of the degree of analogy? Is the people-oriented thinking right? If not, what is the basis? If it is right, will it separate the coexistence between people, things (machines) and the environment? How to characterize both yes and no? Should and should not? Can and can't? Want it or not? Both affirmative and negative? What about... and not...? How to characterize the superposition and entanglement of the macroscopic and mesoscopic states? How to produce and cultivate tolerance, concession, compromise and other mechanisms in the integration of man and machine?
Contemporary artificial intelligence has developed from the initial fully artificially compiled machine automation to artificially pre-compiled machine learning. The next development may be to realize machine cognition through the method of human-machine fusion intelligence, and finally realize machine awakening.
Difficulties faced by human-computer integration and intelligence today
One of the bottlenecks of human-machine fusion intelligence today is that there is no physical theorem or law that defines what human-machine fusion intelligence is. To study this issue clearly, we must first explore the nature of intelligence. Fundamentally speaking, machine intelligence and artificial intelligence are reflections of the conceptualization, systematization, and programming of human intelligence. Fragmented knowledge + fragmented logic constitute a variety of complex human intelligence, fragmented knowledge + fragmentation The logic + implicit/explicit ethical and moral laws and regulations constitute human wisdom. Human real intelligence requires the combination, mixing, and fusion of logic from different fields and different perspectives. Therefore, the real intelligence is the stitched and connected Baina clothing, not Beautiful and beautiful finished products.
After several ups and downs of artificial intelligence, the authority of rationality and the optimism behind formal methods have attracted a lot of questions. Symbols, connections, and behavioral technologies have not only brought progress, but also brought more confusion to the field of intelligence. At the same time, transformative technologies such as artificial intelligence and genetic engineering have made it seem that human beings are more likely than any other era in history. I bring psychological, ethical and thinking challenges. It is no longer important to use reason to find the truth, the meaning of existence is the core issue. Enlightenment replaced God with rationality, but the result was the collapse of all values, and existence became more and more nihilistic. Heidegger used "Being and Time" to try to answer the question of the essence of existence. Husserl turned from the early "Logic Research" to the later At the same time, Wittgenstein, the leader of analytic philosophy, also shifted from the early "On Philosophy of Philosophy" to the later "Philosophy Research" thinking. The two changes invariably removed the "logic" and moved to the "concept" respectively. And "philosophy", this may not be a coincidence or accident.
"Logic" mainly involves judgment and reasoning. It belongs to higher-level conscious activities. To clarify the theory of judgment, it is also necessary to conduct research on lower-level conscious activities such as sensation and perception. In particular, it requires physical data, Psychological information and knowledge, etc., have been studied in depth. The "name" in "names can be very famous" mainly refers to this dynamic representation, naming, definition, and categorization. The "Tao" in "Tao can be very Tao" reflects the mixture of facts and values, things and relationships, including both being and subjective, involving the universal validity of logic and mathematics. It also covers the contingency of psychological laws, which is a collection of logic and illogical. Emotion may be a complex of own logic and others' illogical.
Lao Tzu's Tao is extremely natural. For example, he said "wisdom comes out, there is great hypocrisy", which means that the more people pursue wisdom, the more man-made things, and the greater the element of self-righteousness. This requires turning the usual view of the world upside down and letting things see us like the compound eyes of an insect. Cezanne said: A good painter sees the world not from the outside but from the inside. People with wisdom often see the world in the opposite way to their daily lives, such as Sai Weng who lost Mali and the sparks on Jinggang Mountain. We should not only use scientific methods to see the world. Wisdom and its philosophy should make people look at the world. Truly open to nature, so that the human-machine environmental system in nature speaks to us in its own form and demonstrates the ethical issues of intelligence. The question of whether artificial intelligence or digitization can "be good" is on the surface a question of human-computer interaction or integration. The cooperation of humans and machines to show the world is essentially a question of human nature. If science is regarded as the reality of things and the being of objective existence, then art is the kindness of human beings, the should of subjective imagination, the fusion of the two is beauty, and finiteness is used to represent infinity. Goodness is the original intention. The original intention of a person is also the meaning and value of a person’s life. Of course, human nature also has weaknesses. Individuals will linger between goodness and hypocrisy from time to time, but for the entire human race, goodness is the mainstream and value of the should. Direction, otherwise, the consequences will be self-evident!
From the source point of view, intelligence and ethics are born from the same root, that is, "the mind of right and wrong, wisdom is also". From the perspective of development, being able to distinguish right from wrong is also the main continuous dependence of individuals and groups. There is only the truth of science and technology without the goodness of art. People are often nervous and anxious on the edges of various fields, such as how to deal with the eggs (bombs) laid by physical atoms ? How should physiological gene editing (taboo) be treated? How to grasp the control (governance) of mathematical intelligence control? How to make up for the lack of management? ……All of these will involve the question of how natural persons interact with the materials they have created and the civilized world. Finally, they have to answer questions such as will the love and affection between people be replaced? Does beauty technology reflect the weaknesses and strengths of human nature?
人机融合的关键是人的机化+机的人化之平衡，其本质是人、机、环境系统各种资源的优化配置与适宜调度，涉及人力资源、物理装备资源、时空管理资源、各种环境资源等的一多分有。其中的一个难点和瓶颈是——感觉不等于知觉，生理测量值（心电、脑电、皮肤电等）不一定反映心理情绪值（记忆、理解、认知等），更不要说思维/意识/智能值等，如人脸识别与人心理解的非线性。人机融合的态、势、感、知之间既不是简单的时序关系，也不是复杂的无序关系，而是一种事实与价值的主客观融合关系。
也许，从道德角度分析，宗教不仅是一种迷信，而且是扬善去恶一种途径和关系，关系本身就有人赋予的意义部分，科技是中性的，只有功能而没有能力，这是因为能力是有人性的，功能是没有人性的，如何实现这两者的结合呢？这就需要考虑科技所不能单独解决的部分——善的问题，例如，针对技术的不确定，人在设计算法时，要提前考虑数字世界之外存在着的现实世界，适时地融入相应的基本前提和规范，以使人们面对不可能完全掌控的未来时，不过分地担忧焦虑，重启内在的认知原力，进而在更大程度上把握可以掌控的部分。也许每个人身体上都有一种原力，有些可以发现世界（事实），有些可以发明世界（价值）。每件事物都有一种待开发的原力，有些可以被发现，有些可以被发明，但这需要在一定的情境下。
准确地说，智能不仅包含自然科学和工程技术，还涉及许多社会科学的领域，如人文、哲学、宗教乃至艺术等等，这从世界上最早的兵书之一《孙子兵法》的英文名字可见一斑：The Art of War。这说明：好的智能有时候不仅是技术还是艺术。美军2016年发布的《自主性》研究报告中指出，AI可用于对部队和指挥官进行告警及提供行动方案的建议，但还远远没有达到能够代替人类制定决策的程度。这个观点是比较客观、务实、有效的。看待人机智能化发展的这个难题，不同视角会得出不同的结论。从新技术落地发展的角度看，当前发展人机融合智能面临着三大瓶颈问题：缺大小实在样本数据、缺算法人因验证手段、缺复合专业融合。三大瓶颈问题，说到底是缺乏研究的认识论及方法论——人都说不清楚，指望机器说清楚，在短期内是很难做到的，所以人机融合的研究确实必要：人解决“做正确的事”，机解决“正确地做事”。人机融合智能的本质就是把事实与价值统一起来：人负责价值，而机处理事实。人机融合智能也许可以破解“休谟之问”：Being与Should、自然与自觉的一致性。曾有人说，在计算开始的地方，理解便终结了。而人机结合在一起的深度态势感知就可以实现可理解的计算+算计。
以人为本的思想，在初级的人机融合智能中是可以理解的，机器的主要角色是辅助性的，但随着机器各种功能的不断提高，尤其是隐约出现类人能力的迹象时，“以人为本”的观念可能会被发生变化，我们从《道德经》中的启示可见一斑：一、要把道的存在本身和人为构造区分开来；二、为了克服以人为中心就需要避免人为努力。这是由于要克服人为努力，又需要另外一种人为。老子主要是要克服人道主义，这种人道主义以儒家为代表，突出人的地位和作用，天在人道主义里是被遮蔽的。他认为：人并不重要，自然中存在的道才是中心。人只是自然中的一分子，只是物的一种而已，无限广阔的宇宙里存在着无穷无尽的变化，只有与之随动的人才能更好地发挥主观、客观能动性吧！例如，从文艺复兴到工业革命再到后来，每隔一段时间，世界情境就会更新一次。在西方，每一个阶段都会产生新的对世界的看法。
人文艺术之所以比科学技术容易产生颠覆原创思想，主要是追求主观价值和意义，而不是单纯的客观事实存在。人文艺术哲学宗教给人提供了更广阔的想象空间，正可谓人们看见什么并不重要，重要的是人们如何诠释看见的事物。
情感的本质就是价值的判断。价值的量化非常困难，这需要把价值的本质和计算的本质都搞清楚，才可能做价值计算。有人认为“绝对价值不好搞，能计算相对价值也行”，其实，相对价值计算更难，各种因素都在变，连坐标系都在变。
价值、意义本质上应该是随主体、客体与环境相互作用的变化而改变的，比如，有可能一条信息上一秒有价值，下一秒就完全没有价值了，也可能下下一秒又有更大的价值了，事物之间关系改变很快，价值因此在过程中的变化也很大。价值的计算怎么通过算法实现，理论都还没搞清楚。
人机认知不一致性问题。好的人机融合智能主要解决机器对人的助智（辅助分析、决策）和学伴（个性化、弹性成长），可以通过知几、趣时、变通实现不同数据、信息、知识的往返跳跃、匹配对接、循证查询，即不断使得多种数据/信息/知识进行动态演化组合以达到任务要求。其中的抽象提炼，举一反三就是类比、归纳、演绎混合机制使然， 其中类比起着非常重要的作用，相比归纳、演绎两种推理而言，类比具有鲜明的穿透效应，它可以穿透物质与意识、主观与客观、表象与本质、真实与虚拟、感性与理性、伦理与道德、线性与非线性、确定与不确定等之间的隔阂，让休谟之问不再成为问题，让莱布尼兹的理性演算更为感性，让维特根斯坦的逻辑与非逻辑融合的越发完美，让人、物（机）、环境 之间的协调绵绵不断。
人机智能难于融合的主要原因就在于时空和认知的不一致性，人处理的信息与知识能够变异，其表征的事物/事实千变万化，一直具有相对性，机器处理的数据标识缺乏这种相对变化性。
在西方伦理学界一般认为伦理学的基本问题有两个基准：一个是我们应该如何行动？另一个是我们应该成为什么样的人？前者以行为为中心，属于规范伦理学研究范畴，也是休谟之问的应该（should）问题；后者以行为者为中心，属于美德伦理学研究范畴，也是休谟之问的being问题。这与智能生成的基本问题：“事实与价值能否相符”是一致的。智能的生成将涉及到主观目的与行为动机，并与情境中的客观事实变化密切相关。产生智能不仅需要形式化的计算，更需要意识性的类比。掌握事实性与价值性的因果关系，深研人机融合智能，开展深度态势感知，将是智能研究的重大突破口之一。
人机之间的适时干预要求：干预的既不能早也不能晚，既不超前也不滞后，干预早了，往往会造成双方的措手不及，干预晚了，时过“时迁”，为时已晚。所以如何实现人机之间结合任务要求的适时、适当干预将是未来人机一致性研究的难点和关键。
区块链就是把人机群体智能之间的事实与价值关系统一起来的一种工具，优点是可以通过信用/智能合约手段扬善抑恶、安全/高效整合各种资源；缺点是这是一个乌托邦，除了明链，许多暗链会并行不悖，如（一开始就）预置隐患，还有真实的虚假问题（某人有某种学位，但是别人给他写的答辩论文，而答辩委员们又不太了解他的研究内容，造成事实上的学位，价值上的假学位），所以单纯数字上的区块链很可能只有有/ 无，没有阴/阳和虚/实。
人机融合智能的难点——人心研究
近来，许多学者和普通人都提出未来人工智能的问题。这是因为现在的人工智能还远远未达到大家的期望，现在大家看到的AI在某种意义上 都是自动化或者是高级自动化，那智能化和自动化有什么区别呢？自动化是这样的，固定的输入及可期望的输出，如很多生产线都是自动化生产线，而智能化不是这样，输入可以固定也可以不固定，但是输出一定是非预期性的，绝大部分是非预期性，出乎意料的东西，这才是智能。什么叫做智能？有两个说法：第一个说法是孟子写过的“智者，是非之心也”，是非之心就是“智”，你可以有意识，但不一定有智能，意识是无关乎是非的，而智能是要知道是非的，明白伦理的。根据我们的研究，伦理和智能应该是一回事。什么是伦理？在古希腊看来，伦是分类，分类的道理就是伦理；智能的本质也是分类——是非之心。第二个说法是，西方著名学者米塞斯曾认为：“所谓知识就是区别A与非A，智能也与之相似”。从这两个说法我们可以看出，东西方对智能的认识是类同的。中国传统文化倡导“仁义”。什么叫仁？孔子的仁就是一撇一捺——人，通假字通到那个人身上。什么叫义？义就是应该，孟子特别讲义和仗义。东方和西方智能的共同交界处可能就是这个义——应该（should）。
智能研究领域里面这几位东西方的先驱值得关注：第一是休谟，是智能哲学起源的根，休谟之问，即从事实里面能不能推导出价值观，这是休谟很重要的观点，笔者认为这可能是未来强人工智能的突破点，也是人机融合智能的关键之处。第二个是智能科学之源莱布尼兹，他第一次提出“普遍语言”和“理性演算”，在这两个词的基础上弗雷格提出了分析哲学里面的涵项 一词。后来出现了布尔，也是从莱布尼兹的思想里面演化出布尔代数的，再后来是图灵、冯诺依曼都是从这延伸出来的……真正的技术起源是莱布尼兹。若还有点牵强的话，那么大家都知道图灵，实际上图灵的老师和朋友是维特根斯坦，这是一个很厉害的人，他是图灵的老师和朋友，他给了图灵很多好的智能和哲学思想，但是很多人很少知道他，这是非常遗憾的，也是人工智能界的遗憾，不提他是不行的，他人生里面两部书：第一部书是翻译成中文只有几十页的《逻辑哲学论》，这部书里面讲人的语言是撬开人和人、人和世界关系的切入点，规范化的社会语言是非常重要的，即按照规则语法一句话一句话地说可以理解世界；当他40多岁之后，他又回到剑桥，他写了一些手稿，后来他的一个女学生安斯康姆给他整理出版了第二本书叫做《哲学研究》，这本书也不厚，翻译成汉语一两百页，他否定了第一本书的思想，认为真正了解人类智能最重要的切入点是自然化和生活化的语言，如集贸市场人与人之间省略了主谓宾补的无语法的对话等。这两部书就是弱人工智能和强人工智能的哲学基础，第一部书有关逻辑哲学论，第二部书则涉及哲学研究，没有逻辑。真正强人工智能里面肯定不全是逻辑，仅有逻辑，那都是自动化，那都是规则化的东西，非理性的东西才是揪人心的东西，才是人类智慧的东西。总之，智能的哲学思辨起源于休谟、笛卡尔、弗雷格、维特根斯坦，从事实到价值，从生理到心理，从形式到意向，从逻辑到非逻辑；而智能的科学技术起源于莱布尼兹、布尔、图灵、冯·诺依曼，从文字到演算，从表征到推理，从指令到编码，从存储到计算……
东方思想这方面知识从《易经》开始，《易经》非常棒。但西方是不承认东方有哲学的，认为东方思想里面少逻辑性的东西，只有结果性的东西。最早是伏羲氏，他看到四季变化就开始写了《易经》；《易经》第二个作者是周文王—— “ 文王拘而演周易”，文王把社会管理放在《易经》里面，除了自然以外放入了社会化管理；《易经》第三个作者是孔子，他把人与人之间的伦理放入了《易经》里面，《易经》就是这三个人接力集成所著。《易经》里面有三个词六个字是人类智慧的核心：第一个词叫“知几”，就是要看到事物发展的苗头、兆头；第二个词是“趣时”，即要及时抓住时机；第三个词是“变通”，即随机应变、因时而变。现在智能产品里面有这些东西吗？什么是苗头，什么抓住是时机，什么是变通，你看看大部分都是自动化。第二本书是老子《道德经》，第一句话可能是智能里面最重要的一句话，“道可道非常道，名可名非常名”，老子的“名”里面涉及到智能里面的第一阶段——输入表征阶段，英语叫representation，就是表达、表示，一个事物有万种表征，一花一世界，一树一菩提，白马非马，人能把一个事物表征为很多方面，但在知识图谱里面就非常糟糕，知识图谱的对象、属性、关系都是死的，那是一个标签的世界，什么时候出现活的知识图谱现在还看不到边界，因为现有的表征手段解决不了这个问题。老子《道德经》里面的“道”，包含了算法，算法不是单纯的数学计算方法，也包含了人非逻辑性的算计之法在里面，这是人特有的直觉性的东西。
“人智”之所以很难转成“机智”，是因为人脑不是电脑，“人智”主要是明辨是非，“机智”侧重于模式识别。只有在把一物与它物区分开来，才会对该物有了认知；只有把一个人的知识或信仰状态与他人的区分开来，才对一个人有了解。哲学上最难，也是最重要的任务之一，是明确世界的两类特征，即那些独立于任何观察者而存在的内在特征和那些相对于观察者或使用者而存在的外在特征。例如，一个物体有质量（无论对谁而言）与这个物体是浴缸（也可是水缸、饰缸、粮缸）。所以对智能态势感知系统研究的下一步工作，就是将其具体应用到某一或某些情境中，检验其有效性和可靠性。科学和数学中的陈述乃是描述性的being，因为它们描述了客观事实；而伦理学和美学中的陈述则是评价性的should，因为它们被用来表达感觉和态度、指导行为，而不是陈述事实。而人机融合智能中的深度态势感知就是试图让人、机、环境三者像石榴籽一样紧紧地抱在一起。
一个学科领域常常是以一个错误为基础发展起来的，但这并不必然是致命的，事实上许多领域都是以错误为基础的。比如，化学就是以炼金术为基础的，认知科学就是假设“脑是一个数字化的计算机”这个错误为基础的，同样，人工智能则是以逻辑推理这个错误为研究基础开始的。
人总是从无知开始的：知道的越多，知道的越少。人之智能中可程序性的一部分被转化成了人工智能，而人智中许多可陈述但不可程序的部分和更多不可陈述的部分远远没有被转化，尤其涉及到不可解释、不可学习的非逻辑部分（如情感的深层次、意识的随机变、思维的模糊化）。人，有一种更抽象的能力仍未被发现，这就是从非逻辑开始的智能系统。这种智能也可以从错误开始，但总能找到适合的正确；这种智能可以从混乱开始，但总能找到依稀的有序；这种智能可以从事实开始，但总能找到意向的价值……同时，这种智能也可以从正确开始，但总能找到各种的错误；这种智能也可以从有序开始，但总能找到各样的混乱；这种智能也可以从主观价值开始，但总能找到相关的客观事实。正可谓，数据价值化产生了信息，信息系统化形成了知识， 知识逻辑化构成了智能， 智能非逻辑化演化成了智慧，智慧落实到行为就表现为有意义……
结束语
有专家仔细研究机器学习后发现，机器学习就像一个理发的故事：老和尚让小和尚用南瓜练习理发，小和尚理完后常常下意识地把剪刀插在南瓜上，所以，对于可以实现人们期待的强智能而言，现有的数学、物理等形式化工具都需要改变，只有诞生出新的数学、物理、生理、心理、伦理、法理、管理……也许才可能衍生出新的人机融合智能形式。
世界是复杂的，复杂即不确定性，而人的认知往往是从一个确定的部分开始探索不确定性的，其中充满了试错和反思，进而完成从事实到价值的一跳。义，也就是应该、should机制，功不可没，也许人机融合智能里的“人心”就是“应该”吧，即人和其他智能体的内在驱动力，不断驱使数据、信息、知识、概念、意识、物质在不同时间空间中进行流转、变异，并在物理、心理、数理等业务时/空间中得到拓扑应用。