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

EARLY LAST year, researchers at the Massachusetts Institute of Technology (MIT) used a machine-learning algorithm to look for new antibiotics. After training the system on molecules with antimicrobial properties, they let it loose on huge databases of compounds and found one that worked. Because it operated in a different way, even bacteria that had developed a resistance to traditional antibiotics could not evade the new drug.

Behind the success was a deeper truth: the algorithm was able to spot aspects of reality that humans had not contemplated, might not be able to detect and may never comprehend. The implications of this general development for science, business and warfare—and indeed, for what it means to be human—are the subject of these ruminations by Henry Kissinger, America’s pre-eminent living statesman, Eric Schmidt, the former boss of Google, and Daniel Huttenlocher, an expert on artificial intelligence (AI) at MIT.

In their telling, the most important way that AI will change society is by redefining the basis of knowledge. “Whether we consider it a tool, a partner, or a rival, [AI] will alter our experience as reasoning beings and permanently change our relationship with reality,” the authors write. “The result will be a new epoch.” If uttered by a Soylent-quaffing coder, that sentiment might be dismissed as hyperbole. Coming from authors of this pedigree, it ought to be taken seriously.

In an essay in the Atlantic magazine in 2018, Mr Kissinger argued that AI represents the end of the Enlightenment. This book substantiates that thesis with an engaging romp through the history of reason and decision-making. Unsurprisingly, the crunchiest discussion is of international security. The authors show that AI is radically changing the way states challenge one another, and why fighting wars with autonomous weapons—which could devise strategies, identify targets and kill opponents—invites calamity. Interactions between rivals will become harder to predict, and conflicts more difficult to limit.

Hitherto in the era of nuclear weapons, the central objective of national-security policy has been deterrence. That rests on the premise that a rival state’s capabilities are visible, its doctrine known and its actions foreseeable. The rise of AI systems undermines those consoling assumptions, since responses may diverge from human expectations, and to disclose capabilities may be to forfeit them. A winning manoeuvre may be horrific yet inscrutable, like the algorithm that crushes human players at chess and the board game Go.

The recommendation for policymakers is realism. Since the technology cannot be un-invented, the book calls on America to develop and shape the military applications of AI, rather than surrendering the field to countries that do not share its values. Some of the new capabilities will involve non-violent tasks, such as managing logistics and helping wounded servicemen. At the same time, the authors write, major powers should pursue arms control in AI, as they have for nuclear weapons. The fact that both Mr Kissinger and Mr Schmidt have advised America’s armed forces gives their counsel special weight.

Living with this technology will be tough. Based, as they are, on correlations and elaborate statistics, rather than on a sense of causality, AI’s decisions may seem otherworldly; when the stakes are high, they must be diligently validated. And as AI becomes more widely used, and makes findings that surpass human understanding—whether concerning the laws of science, medicine, managing businesses or navigating roads—society may seem at once to be hurtling towards knowledge and retreating from it. If an AI co-pilot or surgical robot experiences an emergency, who should seize the controls, the human operator or the algorithm? The book calls for a “partnership” between people and machines, but is silent on how to achieve it.

Master and man

There are other shortcomings. A chapter on “global network platforms” (ie, big tech companies) is alternately banal, over-abstract and mealy-mouthed—as if wary of offending any particular business or government. In light of the many ways in which AI bolsters state power, human rights ought to be an essential consideration. Yet there is no discussion of the camps in which, abetted by technology, China’s rulers have imprisoned many Uyghurs. The disappointing final chapter is merely a recapitulation of the first.

Despite these faults, “The Age of AI” is a salutary warning to handle this technology with care and build institutions to control it. Human values and peace must not be taken for granted, the book urges. “While the advancement of AI may be inevitable,” it concludes, “its ultimate destination is not.” With his co-authors Mr Kissinger has, at the age of 98, used his vast experience and versatile mind to make a muscular contribution to one of the 21st century’s most pressing debates. ■

This article appeared in the Books & arts section of the print edition under the headline "Mechanical minds"