The master algorithm

Pedro Domingos
Penguin
2017 | 352pp | £9.99
ISBN 9780241004548

Buy this book from Amazon.co.uk

Machines are taking over the world. They’ve been doing the hard work for quite some time. But now they want to do the interesting things too, like caring for the sick, waging wars, going on dates for us and basically giving us whatever we wish for. Doesn’t it sound wonderful?

Pedro Domingos thinks so. His book The master algorithm: how the quest for the ultimate learning machine will remake our world tells, in three parts, what machine learning is doing for us, how it works, and what it could do in the future. It also tells us of Domingos’ belief in the existence of, and his personal quest for, the ultimate learning algorithm. In his words, the machine that uses it will be able to derive all knowledge – past, present, and future.

The first few chapters were those I enjoyed the most. Domingos reveals just how much machine learning is doing for us already: sequencing DNA, driving cars, managing stocks, filtering spam and recognising our voices, to list just a few. He proceeds to discuss some fascinating subjects: what exactly are learning and knowledge, a partial history of computing, and why traditional methods of knowledge acquisition, such as ‘doing science’, are so limiting. The author’s explanations are clear and he is honest in discussing the obstacles that obstruct the path to intelligent machines.

The middle section of the book describes how machine learning actually works at present. Domingos informs us there are five approaches: symbolism, connectionism, evolutionism, Bayesianism, and analogism. He sets aside a chapter for each, but for me, this is where the book gets the balance wrong. The descriptions go into too much detail. They require a great deal of concentration, the kind that would be better devoted to poring over a text book rather than what is appropriate for a popular science book. Without a technical background or some familiarity with the subject I fear readers will struggle with this section. I certainly did.

Those who persist, however, will be rewarded in the final section, where Domingos envisions a world with the master algorithm. He is very optimistic, predicting less poverty, greater happiness, better relationships, more humane war and humans still in charge. As Domingos wittily says, he for one welcomes our robot underlings. It’s all exciting speculation. You may not share his optimism, but you will be persuaded that we are living through the infancy of a revolution that could change the very nature of humanity.