We like to think that we would stick up for ourselves after being wronged. No one wants to be a coward. Often, though, faced with the realities of power, wealth, and superior resources, we shrink from the good fight. More worryingly, humans can misdiagnose or externalise an issue, rationalising it away. We take a problem grounded in interpersonal relationships, politics, or some other social arrangement, and convince ourselves it is an objective, natural state of being. After all, as distinguished artificial intelligence researcher and author Toby Walsh, author of Machines Behaving Badly: The morality of AI, says: ‘We are, for example, frequently very poor at explaining ourselves. All of us make biased and unfair decisions.’

The last in a trilogy exploring the near and far future of AI, Machines Behaving Badly should be commended for its focus on the relationships being constructed with new intelligent machines. Walsh is clearly passionate about AI. For him, the topic of machine intelligence is one requiring compromise, self-improvement, and clear communication. Building fruitful, equitable AI requires the patient nurturing of a healthy relationship, like friendship or romance. Unlike some breathless accounts of AI, for Walsh, there are no easy fixes.

Morality is a question of politics and a vision of the good life. Politics and the good life are, of course, open-ended questions. Walsh is aware of this: ‘There is no universal set of ethical values with which we need to align our AI systems.’  The reader gains a general understanding of Walsh’s values in Machines Behaving Badly. He is concerned with AI’s impact on equality, racial and gender bias, and climate change. Intervening effectively in the development of AI requires making the case, based on an ethical philosophy, that others should care too.