Will AI Replace Us? - Shelly Fan ***

Although a review should largely be about the content of a book - and we'll get onto that - sometimes, the format it is presented in can have such an impact that this comes to the fore. Will AI Replace Us? falls into this category as part of Thames & Hudson's The Big Idea series which makes use of 'Quick recognition text hierarchy'. Throughout the book there are four different font sizes used. The idea is that you can read the biggest two sizes to get an overview in about half an hour, add in the next size down for an hour's quick read, or go the whole hog for a two-hour in-depth read.

It's an interesting concept, but the execution is horrible. The designer seems to have no idea how book pages should be laid out. The text is chunky sans serif - hard on the eyes on paper - and has far too little white space, getting uncomfortably close to the edge of the page and looking like it has been thrown in, rather than carefully set. To make matters worse, there are two other text styles - a tiny one for image captions and another for hypertext-like definitions of phrases (which are heavily highlighted in the text), causing more visual confusion. That attempt at imitating hypertext reflects the reality that this approach would work far better on a web page, where you can have proper hypertext and collapse lower level sections without resorting to eye-crunching text size differences. It just doesn't work well as a book format. It's novel and creative, which is excellent - but a key lesson of creativity is that you will sometimes fail. And this is one of those times.

As for the content on artificial intelligence, it's a mix of so-so history, good on the science, over-enthusiasm about achievements and interesting lengthy consideration of ethics and impacts - the last of which is probably the best.

A couple of examples of iffy history: in the caption for the Science Museum's Difference Engine (rather disappointing it isn't mentioned where it is) we are told it was 'based on Babbage's original drawings of the Analytical Engine' - but, of course the Difference Engine and the Analytical Engine were totally different things. The text is also fast and loose with Ada Lovelace's role. We're told 'Charles Babbage and Ada Lovelace originated the concept of... the Analytical Engine' - where Lovelace's first involvement was a commentary on someone else's paper on Babbage's idea - there is no basis for the suggestion she was involved in the origination of the design.

Similarly, the early work on expert systems is made to sound like a commercial triumph, talking of the 'dominant success of the first AI boom' and 'AI seemed unstoppable.' It really didn't - it seemed niche and rapidly collapsed. Later we get a lot of material on self-driving cars, which spends some time addressing the societal issues of killing people, but doesn't take on the essential point that the statistical argument of hypothetical lives saved is of little value to the families of those killed. We're told that people's concerns are about a lack of understanding of AI mechanisms, but that's really not what this is about. Nevertheless, along with the descriptions of the technology, the parts of the book focusing on the ethics of AI are the best, engaging the reader in thought about the implications of the technology.

Oddly, the 'future' section seemed one of the most dull, in part because it focussed on medical applications - which tend to be worthy but not something we think of in everyday life terms. Even so, it covered the issues well.

A distinct curate's egg, then. The format is wonderfully creative but really gets in the way of the message. And there are some useful spoonfuls of information in that egg, if you can get past the problems.

Paperback:  

Review by Brian Clegg