Book review: ‘Machines Behaving Badly’

Books about the societal impacts of artificial intelligence and the ethical questions it raises are a dime a dozen, so it is always worth asking: does this book do anything differently? In the case of ‘Machines Behaving Badly: The Morality of AI’ (The History Press, £20, ISBN 9780750999366), the answer is yes, thanks to the author’s decades entrenched in the community from which AI has sprung.

The first third or so of ‘Machines Behaving Badly’ examines the people and companies developing AI technologies. This is packed with nuggets of little-known information. It is not much known, for instance, just how small the group of people building AI is (“There may never have been a planet-wide revolution before which was driven by such a small pool of people”).

Walsh paints this community with more than broad strokes and statistics. He talks in revealing detail about the uncomfortable sexism and philosophies entrenched in the community: objectivism, techno-libertarianism, and transhumanism. In the most thought-provoking part of ‘Machines Behaving Badly’, he discusses the influence of Ayn Rand’s philosophy on Silicon Valley and how this informs how AI is built: “A robot has direct contact with the reality of the world via its perception of that world. its sole purpose is to maximise its reward function. And it does so by reasoning rationally about those precepts.”

He then vivisects the AI corporations: their cheap money, their intent to disrupt, how the voting rights associated with Big Tech shares tend to liberate executives from accountability. Importantly, Walsh sets out alternatives to this prevailing model.

The rest of the book covers more familiar ground. There are the same case studies (social media microtargeting for political ends, racial bias in facial recognition software) and debates found in dozens of other books. However, Walsh employs accessible language and explanations suitable for readers unfamiliar with AI and its unfortunate impacts. He does a particularly thorough job of dissecting the issue of lethal autonomous weapons, having been a key organiser in the campaign to halt the development of these technologies.

In a distinctive touch, an entire chapter is dedicated to AI and the climate; it is encouraging to see that subject getting the same prominence as better trodden AI issues.

Walsh leaves the reader with some important lessons, particularly about the limitations of AI. If only more people recognised how far-removed AI is from human intelligence, he writes, we may see fewer road fatalities caused by drivers placing too much trust in self-driving car algorithms.

The book’s aim is, according to Walsh, to “open your eyes to this strange intruder, to get you to think about the unintended consequences of AI”. Machines Behaving Badly is a good introduction to the unintended consequences of AI, and contains some genuinely thought-provoking insights into the quirks of the people behind them.