剑桥五重奏
出版日期:1999
ISBN:9789578306462
作者:John L. Casti
页数:252页
作者简介
本書獲1998年亞馬遜網站
年度科學類推荐第一名
《劍橋五重奏》是一本談「人工智能之哲學」的書, 談得深入淺出, 值得一讀。
作者以一個假設的「社會」：「聚社會飲」的場合做開場, 讓這場「會飲」 (symposium)發生的地點由古雅典的阿加頓(Agathon)家轉移到今日的英國劍橋基督學院; 會飲的談話主題由「美之為物」轉為「機之思維」。主角不是亞里斯多芬 (Aristophanes)和蘇格拉底, 而是亞蘭．痄F(Alan Tuning)和路德維希．維根斯坦(Ludwig Wittgenstein)。
痄F是現代計算機思想的奠基者, 而維根斯坦則是兩套重要哲學理論的建立者。至為關鍵的是痄F的計算機思想和維根斯坦的第一套哲學有交集之處, 而維根斯坦以第二套哲學來反對第一套哲學的論證, 此時就派上用場, 用來駁斥痄F的觀點。本書的主要內容就以狺顝M維氏的思想為軸心來發展, 並進行兩者的交鋒, 間或有艾文．薛丁格(Erwin Schrodinger)替痄F助陣, 賀爾丹(J.B.S.Haldane)則較支持維根斯坦, 也扮演理論提供的角色, 會飯談話主持人史諾(C.P.Snow)則溝通兩個文化：科學與人文, 穿針引線串場。僕人西蒙則是跑龍套, 藉進奉食物以串場。
痄F的天才在於將數學的運算轉換為機械的計算, 這整個模式稱為「痄F機器」 (Tuning machine)。痄F機器比歷史上其他計算機佔優勢的地方在於數學思想的嚴密以及技術的提昇; 前者是和「判定問題」(Decision Problem)的研究有關, 後者則與電子學的發展有關。簡言之, 痄F將數學上的「可行方法」(effective method)轉化成電子計算機的「機械方法」(mechanical method)。所以只要一個問題可以有限步驟內判定其真假──屬於該集合或不屬於, 那麼電子計算機透過有限的機械步驟, 一定可以得出正確答案出來, 本書的第二幕「大腦與機械的對話」便是在說明這其中的原理。這是我看過談計算機原理的書籍中, 最能以淺顯的文字說明其背後哲學思想與數學理論的書。它能將庫特．哥德爾(Kurt Godel)的不完備性定理(theorem of incompleterness)或不可判定定理(theorem of undecidability) 與痄F機器的關連加以說明, 這也是一大特色。
另外書中也討論了計算機本身的邏輯語言和神經元的類比, 認為大腦神經元接受並發射訊息的方式, 和邏輯命題所蘊含的方式一樣, 都可以轉化成電路訊號的接通方式。因此這就是建立起邏輯／數學、神經生理學、以及電子學三者之間的平衡關係, 這是「人工智能」理論和技術上的重要基礎, 這個「三位一體」的概念也正是維根斯坦反駁的重點。…………
发布书评
精彩书评 (总计1条)
-
Yesterday, I read a book called Cambridge Quintet: machines that can think For all the years, at least from the time I was in college. I am fascinated with concept of AI, which stands for Artificial Intelligence. I watched all the movies I knew and all the books I found that related to it. Not long ago, just about less one year, I started to really delve into the real research of this field. From what I have learnt, the field of AI is really broken into pieces. In my spare time, which is very few, I learnt some superfacial knowledge about natural language processing, how to solve problem by searching, which is to some extent about logic, and finally machine learning. I tried to bring those different, broken pieces into one whole integrated one. That is, I am, trying to get a big picture of the research field of AI. However, there seems a long way to go. Also, not long ago, not more than a month, I started to read the most famous textbook in this field -- Artificial Intelligence, A Modern Approach, which in the past I thought it is too thick to read thoughtly, thus I found lots of other books in this field, which is thinner. I want to list them for whatever reasons: * Artificial Intelligence - A Guide to Intelligent Systems * Artificial Intelligence - A Modern Approach (3e) * Artificial Intelligence A New Synthesis * Artificial.Intelligence.Illuminated * Artificial Intelligence, Introduction (Springer 2011) I read some parts of those books from time to time. And with some help got from some specific books exclusively about problem solving, natural language processing, and courses about machine learning, and especially the book writting by a Chinese researcher -- The Beauty of Mathematics, at the time I finally started to read the very thick book, I found, though I did not know all the academical detail, which is a lot, I seemed to know what the author want to talk about in different parts. To be honest, I indeed feels some kinds of accomplishment. Thus, I quickly read the introduction part and the summary part. In the introduction part, I got a concept of agent, which I think is already in my mind, but not that clear. In the summary part, I got somewhat similar big picture that I was trying to construct -- mainly it is hsitory events, what progress have scientists reached, where the constraints are. However, though I got some kind of big picture, I still have no idea what should I start with. Right now, I am in a lab where the main focus is using machine learning technique to recognize human face, which I have almost done nothing to get a insight into it. Maybe this could be a good start point? Who knows.Now, leave the question above. Finally return to the book I really want to talk about. Artificial Intelligence field is really a very hard field, in my opinion, there are linguistists, biologists, phylosiphier, mathematicians, and finally computer scientists in it -- it's really a inter-disciplinary subject. After all the tries to get the big picture, I asked myself days ago: what is needed to construct a intellectual being? And without clear aid(It is sure that the thought in the history has influenced me by osmosis), I seems to come to similar question, if not answers, to the point made in the Cambridge Quintet: machines that can think. This is my attempt to summarize the book with my understanding: In the book, the discussion begins with the question, can consciousness be made? To begin with, the discussion goes into the knowledge human knows about brain, which is mainly known as neural networks. This knowledge gives scientists the inspiration to develop a subfield of machine learning call neural networks, which simulates the mechanism of human neural networks. However, the mysteries of human brain still seems infinite. In the book Aritifical Intelligence - A Modern Approach, the author states that the understanding in Neuroscience could give hint to profound progress in AI. Based on the neural networks of human brain, Turing believes that the thoughts of human being is nothing more than the electrons exchange between neural cells, which builds the overall enormorous system of human mind. However, how could such hypotheis explains the meaning in the word of human life. For example, how could the word "sole", which is the conbination of s-o-l-e, have meaning? How the meaning is represented? In the book, an abstract answer may like this: if every part of a object who has meaning is treated individually, there exists no meaning. But if they are viewed as a whole, the meaning appears. A more concrete example is China Room Argument.Up to now, the discussion only focus on individual being, saying one human. However, humans live as a whole society. And intelligence could not be realised if there is only one. When society develops, language, gestures and other things which only have meaning under certain context, have been invented to convey thought with others. Some, which discussed in the precious part of the book, only makes sense under society context. Good examples have been made in the book. In my opinion, intelligence may consist of inborn capability and experience, which to some extent stands the ability to learn. We humans have the basic mechanism to survive and the learning mechanism to acquire more knowledge. To build intelligence, the two must be understood thoroughly.After taking into account the influence of the phycical basis of human, the essence of meaning and the role of context, the book comes to the uniqueness of humans. The only thing I could make sense is this part tries to answer whether physical exstence determines the uniqueness of one person? Finally, a section is exclusively made for the role of society played in intelligence.To sum up, in the book, the author may try to discuss the following points, and appended with my belief, not proof: consciousness is physical; in the evolution process, humans have been hard wired the ability to survive, and the capability to learn, which I would like to call self programming. And the latter one significantly relates to interaction, which means society, culture, language and so on. One last word: after writting down all those words, it seems my understanding of artificial intelligence becomes clearer.
精彩短评 (总计2条)
-
Must Read -
在这个年代读这部大作，很有启发意义