AI in the Wild - Peter Dauvergne ***

Sometimes a science book can highlight a totally new connection between two disciplines, and that was certainly the case here - linking environmental science and sustainability with artificial intelligence. Peter Dauvergne shows how (as is also in the case in many other fields) AI can both be a positive and a negative influence on the environment.

On the plus side, we see how AI is being used for everything from sending semi-intelligent drones out to look after the Great Barrier Reef to detecting illegal activities in protected areas by monitoring sounds and identifying those identified with, say, illegal logging in a forest. Perhaps the biggest impact comes from the use of AI in smart resources to reduce climate impact of everything from domestic houses to data centres.

This is all great stuff, but Dauvergne also shows the dangers that AI can present to the environment. This can come from misuse of the technology, but also from the resources needed to make the technology work. Often this results in a balancing act. So, for example, self-driving electric cars are good for the environment when used, but have a negative impact when the raw materials for the batteries and electronics are mined. What we don't really get is a feel for how to quantify this balance. This is a notoriously difficult activity - see, for example, the (failed) attempts to assess which is more environmentally friendly of reusable and disposable nappies.

The subject, then, is important, and Dauvergne uncovers some real positives and issues than many readers will not find familiar. However, the way he goes about it is not great. I was struck by one of the blurb comments on the back where a professor says 'this book is as fast-paced and thrilling as any sci-fi storyline.' All I can say is, this professor must read really boring novels, as the writing style here is classic dull academic: the book is packed with fact statements and is almost entirely lacking in any narrative flow.

To make matters worse, a lot of these statements that are thrown at us are not backed up with evidence or balance. As an example, there are dramatic statements of the way AI is sweeping the world, for example in the deployment of self-driving cars - but very little about all the issues self-driving car manufacturers face in going from localised trials to proper implementation, for example, the limitations of computer image recognition which can be fooled by small constructed patterns and the lack of consideration of resistance to the very idea of self-driving cars as more people are killed by them. 

The author's politics also come through extremely strongly, which leads to the extension of the argument well beyond the environment, diluting the thrust of the book. So, for example, we are told (without evidence) that ‘Global sustainability is going to require a fairer, more just distribution of wealth and resources’. I am not saying this is necessarily untrue, but it’s not an obvious conclusion and it's not really about the environment. Dauvergne emphasises the existence of inequality but doesn’t mention the vast improvements in the circumstances of many at the bottom end of this scale - it's almost as if Hans Rosling's Factfulness didn't exist.

There's a core of interesting and useful information here, but it's a shame it's not presented better.

Paperback:

Bookshop.org

  

Kindle:  

Using these links earns us commission at no cost to you

Review by Brian Clegg