2022.08.04 了解人工智能的五本最佳书籍

shiyi18

Artificial intelligence
The five best books to understand AI
Specialists outside the field do better at explaining the implications
Side face of AI robot by network form.
Aug 4th 2022

Save

Share

Give
In recent years artificial intelligence (AI) has undergone a revolution. After decades of modest progress that never quite lived up to its promise, a different approach—relying on big data and stats, not clever algorithms—made huge strides in solving real-world problems like voice- and image-recognition and self-driving cars. Also in the past ten years, a lot of books have been published that aim to explain what AI is, where it’s going and why it matters. Books by AI academics do well at explaining the technology but are less robust at foreseeing the implications. For that, bright lights from outside the AI world do better. Five books in particular stand out.

Novacene: The Coming Age of Hyperintelligence. By James Lovelock with Bryan Appleyard. The MIT Press; 160 pages; $22.95. Allen Lane; £14.99


In a slim volume, James Lovelock—a British scientist who originated the Gaia hypothesis that the Earth acts as a self-regulating, living organism—predicts how “cyborgs” may help people in the future, notably by keeping the planet temperate so both man and machine can survive. The AI-infused bots will be vastly more intelligent and faster-thinking than humans (and may keep us as pets, he suggests). But we should be willing to cede our place on the pedestal to a superior intelligence, he says, even if their answers are so intricately reasoned that they are ineffable to the human mind. Mr Lovelock passed away last month on his 103rd birthday, so the book represents some of the final thoughts of a very special thinker (read our review).

The Age of AI. By Henry Kissinger, Eric Schmidt and Daniel Huttenlocher. Little, Brown and Company; 272 pages; $30. John Murray; £20

The crux of this book is a big idea: AI marks “a new epoch” since it ends the Enlightenment’s ethos that placed humans at the centre of all that is knowable (displacing God), putting in their stead machines with superior intelligence. The book does a great job of explaining how AI systems work, though a chapter on business is lacklustre. Yet the analysis on international security is exceptional. Henry Kissinger, America’s pre-eminent foreign-policy thinker, is clearly spooked by AI weapons and calls on the West to develop them (so as to not cede them to a foe) while urging immediate AI arms-control talks. Since AI responses may be unlike human ones, the environment is more unpredictable and dangerous, the book argues (read our review).

AI Superpowers: China, Silicon Valley and the New World Order. By Kai-Fu Lee. Houghton Mifflin Harcourt; 272 pages; $28


Born in Taiwan, the author studied AI in America, became an executive at Apple, Microsoft and Google and now runs a venture-capital fund in China—so is well placed to evaluate the two countries’ rivalry. American AI dominance will give way to China, he believes, because the Chinese work harder and have more data (in part because of looser standards on privacy). The liberal belief that openness is required for innovation is being proved false, he claims. Yet a neocolonial situation is forming whereby all countries will need to align with an AI ecosystem from China or America for everything from health-care systems to corporate IT and military equipment. Mr Lee’s prognostications are fascinating, regardless of whether they are borne out (read our review).

Prediction Machines. By Ajay Agrawal, Joshua Gans and Avi Goldfarb. Harvard Business Review; 272 pages; $30

Three economists at the University of Toronto wisely frame AI as lowering the cost of making predictions, in the same way as computers lowered the cost of doing maths. When doing so is exponentially less expensive, society finds ways to transform problems into ones the technology can handle. Hence books, photos, video and music became domains for computers to handle, displacing physical formats. Likewise, AI will replace humans, from self-driving cars to office automation. But all is not lost: as the cost of predictions go down, the value of human judgement will increase, the authors argue. (Read our economics column on their recent work).

Possible Minds: Twenty-Five Ways of Looking at AI. Edited by John Brockman. Penguin Press; 320 pages; $28

This compilation of essays from an all-star cast of experts from across scientific fields draws together smart thinking about the technology. Judea Pearl, who brought causal reasoning into statistics, explains AI’s blind spot from number-crunching alone. Stuart Russell, a computer scientist, synthesises his work on developing “provably” beneficial AI to ensure the machines don’t go rogue. Other essays by thinkers such as Steven Pinker and Alison Gopnik look at human cognition compared to AI’s answers, and how humans can stay relevant. The essays are a tasting menu of intriguing ideas and a good introduction to the topic.

These books provide an optimistic look at data and AI, yet there is a budding genre of negative takes. They, too, are good—if read with a critical eye. The best is Privacy is Power by Carissa Véliz on the need to take control of personal information (read our blurb). Weapons of Math Destruction by Cathy O’Neil is an unfair hit-job: the data are innocent—it’s people who are to blame! (Listen to our podcast with the author.) The Atlas of AI by Kate Crawford is a litany of AI’s woes, from its environmental toll to unemployment.

To pick just one book, it would be Mr Kissinger’s, et al. Artificial intelligence humbles humanity and their analysis shrewdly explains why. ■

_______

Our deputy executive editor is the co-author of a book on AI and human decision-making:

Framers: Make Better Decisions in the Age of Big Data. By Kenneth Cukier, Viktor Mayer-Schönberger and Francis de Véricourt. Dutton; 272 pages; $28. WH Allen; £20

An exploration of what people can do that AI can’t: use mental models to see the world in a new way. “Different, and better, than the usual recipes for smart thinking”, said the Financial Times.
_______________

More from The Economist reads:
We recommend the five best books for understanding Silicon Valley’s history
Our defence editor picks five books that help make sense of modern warfare
Our senior economics writer suggests reading to understand the history of Western capitalism
Our language columnist, Johnson, picks five of his favourite recent reads



人工智能
了解人工智能的五本最佳书籍
该领域之外的专家更善于解释其含义
按网络形式划分的人工智能机器人的侧脸。
2022年8月4日



近年来，人工智能（AI）已经经历了一场革命。在经历了数十年来从未完全实现其承诺的适度进展之后，一种不同的方法--依靠大数据和统计数据，而不是聪明的算法--在解决语音和图像识别以及自动驾驶汽车等现实世界问题方面取得了巨大进展。在过去的十年里，也有很多书籍出版，旨在解释什么是人工智能，它将走向何方，以及它为什么重要。人工智能学者的书在解释技术方面做得很好，但在预见影响方面却不那么有力。对于这一点，来自人工智能世界之外的明灯做得更好。有五本书特别引人注目。

Novacene。即将到来的超智能时代。詹姆斯-洛夫洛克与布莱恩-阿普利亚德合著。麻省理工学院出版社；160页；22.95美元。Allen Lane；14.99英镑


在这本薄薄的书中，詹姆斯-洛夫洛克--一位英国科学家，他提出了盖亚假说，认为地球是一个自我调节的活的有机体--预测了 "机械人 "在未来会如何帮助人们，特别是通过保持地球的温度，使人和机器都能生存。充满人工智能的机器人将比人类更聪明，思维更敏捷（他表示，可能会把我们当成宠物）。他说，但我们应该愿意把我们的地位让给一个卓越的智能体，即使它们的答案是如此复杂的推理，以至于人类的思维无法表达。洛夫洛克先生上个月在他103岁生日时去世了，所以这本书代表了一位非常特别的思想家的一些最后想法（阅读我们的评论）。

人工智能的时代。作者：亨利-基辛格、埃里克-施密特和丹尼尔-胡滕罗彻。Little, Brown and Company；272页；30美元。约翰-默里；20英镑

这本书的核心是一个大概念。人工智能标志着 "一个新的时代"，因为它结束了启蒙运动将人类置于所有可知事物的中心（取代了上帝）的精神状态，取而代之的是具有卓越智能的机器。这本书很好地解释了人工智能系统是如何工作的，尽管关于商业的一章很乏味。然而，对国际安全的分析却很特别。美国杰出的外交政策思想家亨利-基辛格（Henry Kissinger）显然被人工智能武器吓坏了，他呼吁西方国家开发这些武器（以便不把它们让给敌人），同时敦促立即进行人工智能军备控制谈判。该书认为，由于人工智能的反应可能与人类的反应不同，因此环境更加不可预测和危险（阅读我们的评论）。

AI超级大国。中国、硅谷和新的世界秩序。作者：李开复。霍顿-米夫林-哈考特出版社；272页；28美元


作者出生于台湾，在美国学习人工智能，成为苹果、微软和谷歌的高管，现在在中国经营一家风险投资基金--所以他有能力评估这两个国家的竞争关系。他认为，美国的人工智能主导地位将让位于中国，因为中国人工作更努力，拥有更多数据（部分原因是对隐私的标准更宽松）。他称，自由主义者认为创新需要开放的信念正在被证明是错误的。然而，一个新殖民主义的局面正在形成，所有国家都将需要与来自中国或美国的人工智能生态系统保持一致，包括从医疗保健系统到企业IT和军事装备。李先生的预言很吸引人，不管它们是否被证实（阅读我们的评论）。

预测机器。作者：Ajay Agrawal、Joshua Gans和Avi Goldfarb。哈佛商业评论》；272页；30美元

多伦多大学的三位经济学家明智地将人工智能归结为降低预测的成本，就像计算机降低了做数学题的成本一样。当这样做的成本成倍降低时，社会就会找到方法将问题转化为技术可以处理的问题。因此，书籍、照片、视频和音乐成为计算机处理的领域，取代了物理格式。同样地，人工智能将取代人类，从自动驾驶汽车到办公室自动化。但一切都没有失去：作者认为，随着预测的成本下降，人类判断的价值将增加。(阅读我们关于他们最近工作的经济学专栏）。

可能的头脑。看待人工智能的二十五种方式。由约翰-布罗克曼编辑。企鹅出版社；320页；28美元

这本由来自各科学领域的全明星专家组成的论文汇编汇集了对该技术的明智思考。将因果推理引入统计学的朱迪亚-珀尔（Judea Pearl）从数字计算的角度解释了人工智能的盲点。计算机科学家斯图尔特-拉塞尔（Stuart Russell）综合了他在开发 "可证明的 "有益人工智能方面的工作，以确保机器不会失控。史蒂芬-平克（Steven Pinker）和艾莉森-戈普尼克（Alison Gopnik）等思想家的其他文章探讨了人类的认知与人工智能的答案相比，以及人类如何才能保持相关性。这些文章是引人入胜的想法的品尝菜单，也是对该主题的良好介绍。

这些书对数据和人工智能提供了一个乐观的看法，然而也有一个正在萌芽的负面看法的流派。它们也很好，如果用批判的眼光来阅读的话。最好的是Carissa Véliz的《隐私就是力量》，讲述了控制个人信息的必要性（阅读我们的简介）。凯西-奥尼尔（Cathy O'Neil）的《数学毁灭武器》（Weapons of Math Destruction）是一个不公平的打击：数据是无辜的，应该责备的是人！（听听我们与播客的对话）。凯特-克劳福德(Kate Crawford)的《人工智能地图》(The Atlas of AI)是一连串关于人工智能的困境，从环境损失到失业。

如果只选一本书，那就是基辛格先生等人的书。 人工智能让人类感到谦卑，他们的分析精明地解释了原因。■

_______

我们的副执行编辑是一本关于人工智能和人类决策的书的共同作者。

Framers: 在大数据时代做出更好的决定。作者是Kenneth Cukier、Viktor Mayer-Schönberger和Francis de Véricourt。Dutton；272页；28美元。 WH Allen；20英镑

探讨人能够做什么，而人工智能不能做什么：使用心理模型，以新的方式看世界。金融时报》说："与通常的智能思维食谱不同，而且更好"。
_______________

更多来自《经济学人》的阅读。
我们推荐了解硅谷历史的五本最佳书籍
我们的国防编辑挑选了五本有助于理解现代战争的书籍
我们的高级经济学作家建议通过阅读了解西方资本主义的历史
我们的语言专栏作家约翰逊挑选了五本他最近最喜欢的读物