AI 3.0 cover

AI 3.0

Rating 4.5
Word count
197K words
Chapters
137
ISBN
9787572700378

Tags

Summary

Artificial intelligence is now profoundly affecting our production and life, and even bears on the future destiny of humanity. But what exactly is artificial intelligence? What are the principles behind it? What historical changes has AI undergone from its birth to its present evolution? Where are the boundaries of AI's capabilities today? What is the difference between artificial intelligence and human intelligence? What challenges and opportunities will AI face in the future? To these questions, AI 3.0 will reveal the answers one by one. AI 3.0 is a new work by Melanie Mitchell, author of the mega-bestseller Complexity and a frontier scientist of complex systems, who has thought for ten years to clarify artificial intelligence and human intelligence. The book stems from Mitchell's years of recording the real state of development in the AI field; in five parts she reveals "what AI can do now and what we can expect from it in the coming decades." After describing the history of AI, the author clarifies the relationship between artificial intelligence and human intelligence by exploring the current state and limitations of four major hot application areas: visual recognition, games and reasoning, natural language processing, and common-sense judgment. The case analyses of face recognition, autonomous driving, machine translation, and other topics are full of great insights. And these are precisely the difficulties facing AI development today and what must be rethought if AI is to achieve breakthrough progress.

Editor's review

The strength of this book is that it answers "what AI can and cannot do now" with a relatively sober framework. The author starts from the history of development and, around visual recognition, games and reasoning, natural language processing, and common-sense judgment, analyzes the boundaries and limitations of current capabilities rather than merely listing technical terms or inflating future imaginings. For ordinary readers, it helps distinguish "demo effects" from "real capabilities" and understand the differences between AI and human intelligence; for practitioners, it offers a perspective for reflecting on technical routes and managing expectations. While reading, do not try to remember every case; focus on the capability boundaries and reasons for failure of each technical route. It is suitable when you feel confused by AI news, and also as general reading to help build steadier judgment about artificial intelligence.

Who it's for

A good fit for

  • General readers who want a systematic understanding of AI's capability boundaries
  • Professionals who follow AI news but lack a technical background
  • Product people or managers who need to explain AI's current state and limitations to others
  • Readers interested in cognitive science, complex systems, and intelligence

Not a good fit for

  • Readers hoping to directly learn programming or model training practice
  • People who only want quick scans of frontier papers or reviews of the latest products
  • Popular-science readers expecting light entertainment and strong storytelling

Reading tips

  1. Read the overview first to build a framework of "what it can and cannot do," then choose the four application chapters according to interest.
  2. After each application area, summarize one capability boundary in your own words to avoid remembering only cases.
  3. Read alongside daily AI news; when you encounter exaggerated claims, check the corresponding chapter to train independent judgment.
  4. Do not force yourself to finish in one sitting; read by theme such as visual recognition or natural language processing week by week and take brief notes.

About the publisher

Sichuan Science and Technology Press China · Founded in 1982

Sichuan Science and Technology Press features science and technology, popular science, and science fiction publishing, while also covering agricultural technology and medical and health books. In science fiction, the press has continued to release works by Liu Cixin, including The Wandering Earth: Selected Short Stories of Liu Cixin, boxed sets of Arthur C. Clarke's science fiction classics, and masterworks such as Ringworld, Snow Crash, and Minus Zero, organized into series including the Chinese Science Fiction Cornerstone Series and the World Science Fiction Masters Series. In popular science, classics such as George Gamow's One Two Three... Infinity reflect its orientation toward public science communication. Its author resources in the library include Liu Cixin, Yao Haijun, and Cheng Xi, among other science fiction and popular science editors and translators; the library contains 190 titles, classified mainly as popular science, science fiction, quality fiction, and foreign literature. Public information is limited, and the above is summarized only from the library's titles and existing profile.

Contents

  1. 版权信息
  2. 中文版序 今天的机器距离真正像人一样理解世界还有多远
  3. 序 等那一口仙气儿
  4. 译者序 未来智能:人有人用,机有机用
  5. 引言 创造具有人类智能的机器,是一场重大的智力冒险
  6. 第一部分 若想对未来下注,先要厘清人工智能为何仍然无法超越人类智能
  7. 人工智能流派大变迁
  8. 01 从起源到遭遇寒冬,心智是人工智能一直无法攻克的堡垒
  9. 达特茅斯的两个月和十个人
  10. 定义,然后必须继续下去
  11. 任何方法都有可能让我们取得进展
  12. 符号人工智能,力图用数学逻辑解决通用问题
  13. 感知机,依托DNN的亚符号人工智能
  14. 感知学习算法,无法重现人脑的涌现机制
  15. 学习感知机的权重和阈值
  16. 感知机是一条死胡同
  17. 泡沫破碎,进入人工智能的寒冬
  18. 看似容易的事情其实很难
  19. 本章要点
  20. 02 从神经网络到机器学习,谁都不是最后的解药
  21. 多层神经网络,识别编码中的简单特征
  22. 无论有多少输入与输出,反向传播学习都行得通
  23. 联结主义:智能的关键在于构建一个合适的计算结构
  24. 亚符号系统的本质:不擅长逻辑,擅长接飞盘
  25. 机器学习,下一个智能大变革的舞台已经就绪
  26. 本章要点
  27. 03 从图灵测试到奇点之争,我们无法预测智能将带领我们去往何处
  28. “猫识别机”掀起的春日狂潮
  29. 人工智能:狭义和通用,弱和强
  30. 人工智能是在模拟思考,还是真的在思考
  31. 图灵测试:如果一台计算机足够像人
  32. 奇点2045,非生物智能将比今天所有人类智能强大10亿倍
  33. 一个“指数级”寓言
  34. 摩尔定律:计算机领域的指数增长
  35. 神经工程,对大脑进行逆向工程
  36. 奇点的怀疑论者和拥趸者
  37. 对图灵测试下注
  38. 本章要点
  39. 第二部分 视觉识别:始终是“看”起来容易“做”起来难
  40. 视觉识别的两次飞跃
  41. 04 何人,何物,何时,何地,为何
  42. 看与做
  43. 深度学习革命:不是复杂性,而是层深
  44. 模拟大脑,从神经认知机到ConvNets
  45. ConvNets如何不将狗识别为猫
  46. 激活对象特征,通过分类模块进行预测
  47. 不断从训练样本中学习,而非预先内置正确答案
  48. 本章要点
  49. 05 ConvNets和ImageNet,现代人工智能的基石
  50. 构建ImageNet,解决目标识别任务的时间困境
  51. 土耳其机器人,一个需要人类智慧的工作市场
  52. 赢得ImageNet竞赛,神经网络的极大成功
  53. ConvNets淘金热,以一套技术解决一个又一个问题
  54. 在目标识别方面,ConvNets超越人类了吗
  55. 我们离真正的视觉智能还非常遥远
  56. 本章要点
  57. 06 人类与机器学习的关键差距
  58. 人工智能仍然无法学会自主学习
  59. 深度学习仍然离不开“你”的大数据
  60. 长尾效应常常会让机器犯错
  61. 机器“观察”到的东西有时与我们截然不同
  62. 有偏见的人工智能
  63. 人工智能内心的黑暗秘密以及我们如何愚弄它
  64. 本章要点
  65. 07 确保价值观一致,构建值得信赖、有道德的人工智能
  66. 有益的人工智能,不断改善人类的生活
  67. 人工智能大权衡:我们是该拥抱,还是谨慎
  68. 人脸识别的伦理困境
  69. 人工智能如何监管以及自我监管
  70. 创建有道德的机器
  71. 本章要点
  72. 第三部分 游戏与推理:开发具有更接近人类水平的学习和推理能力的机器
  73. 始于游戏,不止于游戏
  74. 08 强化学习,最重要的是学会给机器人奖励
  75. 训练你的机器狗
  76. 现实世界中的两大绊脚石
  77. 本章要点
  78. 09 学会玩游戏,智能究竟从何而来
  79. 深度Q学习,从更好的猜测中学习猜测
  80. 价值6.5亿美元的智能体
  81. 西洋跳棋和国际象棋
  82. 不智能的“智能赢家”深蓝
  83. 围棋,规则简单却能产生无穷的复杂性
  84. AlphaGo对战李世石:精妙,精妙,精妙
  85. 从随机选择到倾向选择,AlphaGo这样工作
  86. 本章要点
  87. 10 游戏只是手段,通用人工智能才是目标
  88. 理解为什么错误至关重要
  89. 无须人类的任何指导
  90. 对人工智能而言,人类的很多游戏都很具挑战性
  91. 它并不真正理解什么是一条隧道,什么是墙
  92. 除去思考“围棋”,AlphaGo没有“思考”
  93. 从游戏到真实世界,从规则到没有规则
  94. 本章要点
  95. 第四部分 自然语言:让计算机理解它所“阅读”的内容
  96. 自然语言的本质
  97. 11 词语,以及与它一同出现的词
  98. 语言的微妙之处
  99. 语音识别和最后的10%
  100. 分类情感
  101. 递归神经网络
  102. “我欣赏其中的幽默”
  103. “憎恶”总与“讨厌”相关,“笑”也从来伴随着“幽默”
  104. word2vec神经网络:口渴之于喝水,就像疲倦之于喝醉
  105. 本章要点
  106. 12 机器翻译,仍然不能从人类理解的角度来理解图像与文字
  107. 编码器遇见解码器
  108. 机器翻译,正在弥补人机翻译之间的差距
  109. 迷失在翻译之中
  110. 把图像翻译成句子
  111. 本章要点
  112. 13 虚拟助理——随便问我任何事情
  113. 沃森的故事
  114. 如何判定一台计算机是否会做阅读理解
  115. “它”是指什么?
  116. 自然语言处理系统中的对抗式攻击
  117. 本章要点
  118. 第五部分 常识——人工智能打破意义障碍的关键
  119. 无法逾越的人类智能之火
  120. 14 正在学会“理解”的人工智能
  121. 理解的基石
  122. 预测可能的未来
  123. 理解即模拟
  124. 我们赖以生存的隐喻
  125. 抽象与类比,构建和使用我们的心智模型
  126. 本章要点
  127. 15 知识、抽象和类比,赋予人工智能核心常识
  128. 让计算机具备核心直觉知识
  129. 形成抽象,理想化的愿景
  130. 活跃的符号和做类比
  131. 字符串世界中的元认知
  132. 识别整个情境比识别单个物体要困难得多
  133. “我们真的,真的相距甚远”
  134. 本章要点
  135. 结语 思考6个关键问题,激发人工智能的终极潜力
  136. 致谢
  137. 注释

FAQ

Does this book require programming or math foundations?
No. The author discusses AI's capability boundaries, development history, and failure cases in a popular-science way, with technical principles placed in understandable narrative. Basic scientific literacy is enough; programming and math are not prerequisites.
How is it different from ordinary AI popular-science books?
It is not content with listing technical achievements but focuses on analyzing the capability limitations and failure causes in visual recognition, games and reasoning, natural language processing, and common-sense judgment, helping readers understand the differences between AI and human intelligence and make steadier judgments.
Will the content become outdated quickly?
Specific cases will change over time, but the issues discussed in the book—capability boundaries, the difficulty of common-sense judgment, and human-machine differences—remain relevant for a long time. Readers can use it to build an analytical framework and then update their judgments with the latest progress.

Comments

Find this book

This site is an index: you start “Search all sources” in your browser, and the entries it lists all point to third-party sites; please judge their availability yourself.

About the author

Melanie Mitchell is a popular-science author whose works are carried by this Chinese book site, with five titles currently listed, commonly categorized under popular science, artificial intelligence, and management. Because public information is limited, her nationality, dates of birth and death, and full career details remain unclear, so it is more worthwhile to get to know her through her works. Her representative titles include AI 3.0 and Complexity (new edition), as well as several works in the First Push Book Series comprehensive collection, plus the eight-volume Future Whizzing By and the collection Understanding the Logic and Progress of the Intelligent Age. These works focus on artificial intelligence, complex systems, and the logic of the intelligent age, combining scientific explanation with discussion of key issues for general readers who want to understand technological progress and its social impact.

王飞跃 Author

Wang Feiyue is an author about whom publicly available information is currently limited, and the existing details are not sufficient to give a complete picture of his personal life or academic background. Judging from the works held in the collection, his writing mainly concerns two fields: artificial intelligence and management. His representative works include AI 3.0 and the eight-volume series The Future Is Coming Fast. These books address general readers and managers interested in technological change, discussing the stages of AI development, technology trends, and their possible effects on organizations and society, with a clear emphasis on trend observation and forward-looking discussion. In the two works currently available, his writing tends to use conceptual frameworks to organize complex issues and to connect technological evolution with management practice, helping readers understand the changes brought by intelligent technologies. Because information such as nationality, dates of birth and death, education, and professional positions has not been confirmed, no inferences are made here; readers may learn more about his areas of focus and argumentative style through his works.

李玉珂 Author

Li Yuke is an author with limited publicly available information; the only verified record in our database shows a single work, AI 3.0, generally classified under artificial intelligence. As the author's dates of birth and death, nationality, educational background, and professional experience remain unconfirmed, no speculation is offered here. Judging from the existing work, Li Yuke's writing centers on the frontier field of artificial intelligence, with topics close to the hot spots of technological evolution and industrial change. The "3.0" in the title suggests an intention to trace the development of AI in chronological stages. Writing of this kind is typically aimed at general readers who wish to understand the trajectory and future direction of AI technology, seeking a balance between professional depth and readability. Constrained by the available materials, the author's specific writing style, academic stance, and social influence cannot yet be fully assessed; readers may form their own judgment based on the content of AI 3.0.

王晓 Author

Wang Xiao is a Chinese author whose works span a notably wide range of subjects. Public information about the author is limited, and details such as nationality, dates of birth and death, and educational background are not clearly documented. Judging from the nine works held in the collection, Wang Xiao's interests extend across artificial intelligence, medicine, popular science, mystery and detective fiction, finance, and language and linguistics. The representative work AI 3.0 addresses artificial intelligence, while Ecological Agriculture of Chinese Materia Medica focuses on the intersection of medicine and agriculture. The four-volume I Am a Physics King series is aimed at popular science, and the Mysterious Island: Detective Kunk series builds mystery and detective stories around clues such as a gold medal, a diamond necklace, and a champion bracelet, suiting young readers. In addition, A Small Window on China and the World and Theories and Methods of Pragmatic Research show the author's engagement with financial observation and linguistic research. Overall, the works combine informative content with readability and appeal to a broad readership.

张慧 Author

Zhang Hui, about whom limited public information is available, emerges as a writer through works on a wide range of subjects. Thirteen of her titles are held in the collection, spanning artificial intelligence, medicine, art, inspirational growth, finance and economics, and programming and design. Representative works include AI 3.0, Case Analyses of Tuberculosis Prevention and Control in Schools, Talking About Poetry and Chatting About Painting: Subtle Words on Art, The Complete Works of Carnegie (8 volumes), The Joint Operation Institutions of the Four Northern Banks, A Step Ahead: The Way to Improve System Performance, Tuberculosis Prevention and Control Training Materials: Prevention and Control Volume, and The Power of Habit. These works suggest that her writing combines knowledge popularization with professional application, addressing both technological frontiers and programming practice as well as public health, art appreciation and personal growth. Her subjects are broad and her approach practical, with an emphasis on readability and usefulness for readers of different backgrounds.

More by the same author

Top rated in the same category