Skip to main content

Artificial Intelligence (Ladybird Experts) - Michael Wooldridge ****

As a starting point in assessing this book it's essential to know the cultural background of Ladybird books in the UK. These were a series of cheap, highly illustrated, very thin hardbacks for children, ranging from storybooks to educational non-fiction. They had become very old-fashioned, until new owners Penguin brought back the format with a series of ironic humorous books for adults, inspired by the idea created by the artist Miriam Elia. Now, the 'Ladybird Expert' series are taking on serious non-fiction topics for an adult audience.

Michael Wooldridge takes us on an effective little tour of artificial intelligence. Given the very compact form, he fits a lot in, taking us through some of the historical development including the 'golden age' (when everything seemed possible and very little was done), through the rise and fall of expert system, robotics, and the modern split between machine learning and 'good old-fashioned AI'. He emphasises how much in the past expectations have far outreached reality (then does something rather similar himself at the end).

It's just such a shame the format wastes half the space available with pointless and rather childish illustrations that don't add anything at all to the content. This book is supposed to be aimed at adults, but I read this on a train (the format's ideal for a short journey), and felt embarrassed to be seen looking at what appears to be a children's picture book.

The format isn't helped by the problem of using an academic to put something across to the general reader - there was sometimes a lack of appreciation of the sort of questions that people would want answering. For example, there is a page dedicated to the program SHRDLU. Immediately the reader thinks ‘Why SHRDLU?’ And Wooldridge leaves us hanging with the unhelpful ‘curiously named’. (As I couldn’t be so cruel, it’s the second block of letters on a Linotype printing machine, where the first block, in approximate frequency of use order was ETAOIN.)

Occasionally, the tightness of space of the format led to an oversimplification that was confusing. So, for example, when talking about the travelling salesman problem, we are told: ‘The best we seem able to do with NP-complete problems [never properly defined] is to exhaustively consider all possible solutions.’ But it depends what you mean by best. That's the only way to be sure of finding the optimal solution, but there are methods that will get within a small percentage of optimal in practical times (or satnavs wouldn’t work), which are surely better than a non-feasible approach? Similarly, we are told ‘The type of logic used in mathematics can’t cope with this seemingly trivial scenario [moving away from the idea Tweetie, who is a bird, can fly when you discover it’s a penguin], because it wasn’t designed for retracting conclusions.’ But this is exactly what happens when using Bayesian methods... which bizarrely are covered on the next page.

A final, and important oversimplification is over the negatives of AI. Some parts ignore this. The section on driverless cars is upbeat about all the lives that could potentially be saved. But this ignores the psychological issue that we aren't good at weighing up virtual lives saved against the actual people who will definitely be killed by driverless cars (the first example occurred just before this review was written). Though Wooldridge does mention problems from job losses, loss of privacy and algorithmic bias, he also misses the negatives arising from a point he makes earlier that machine learning can’t explain its decisions. The inability to explain why, say, someone is refused a mortgage runs counter to increasing move towards corporate transparency and could prove a real problem.

Overall, Wooldridge does a surprisingly good job, though, given the limitations of the format.

Hardback:  

Kindle:  


Review by Brian Clegg

Comments

Popular posts from this blog

The AI Delusion - Gary Smith *****

This is a very important little book ('little' isn't derogatory - it's just quite short and in a small format) - it gets to the heart of the problem with applying artificial intelligence techniques to large amounts of data and thinking that somehow this will result in wisdom.

Gary Smith as an economics professor who teaches statistics, understands numbers and, despite being a self-confessed computer addict, is well aware of the limitations of computer algorithms and big data. What he makes clear here is that we forget at our peril that computers do not understand the data that they process, and as a result are very susceptible to GIGO - garbage in, garbage out. Yet we are increasingly dependent on computer-made decisions coming out of black box algorithms which mine vast quantities of data to find correlations and use these to make predictions. What's wrong with this? We don't know how the algorithms are making their predictions - and the algorithms don't kn…

Infinity in the Palm of your Hand - Marcus Chown *****

A new Marcus Chown book is always a treat - and this is like a box of chocolates: a collection of bite-sized delights as Chown presents us with 50 science facts that are strange and wonderful.

The title is a quote from William Blake's Auguries of Innocence: 'To see a World in a Grain of Sand, / And a Heaven in a Wild Flower, / Hold Infinity in the palm of your hand, / And Eternity in an hour.' It would seem particularly appropriate if this book were read on a mobile phone (so it would be literally in the palm), which could well be true for ebook users, as the short essays make excellent reading for a commute, or at bedtime. I found them distinctly moreish - making it difficult to put the book down as I read just one more. And perhaps another. Oh, and that next one looks really interesting...

Each of the 50 pieces has a title and a short introductory heading, which mostly give a feel for the topic. The very first of these, however, briefly baffled me: 'You are a third mus…

How to Invent Everything - Ryan North ****

Occasionally you read a book and think 'I wish I'd thought of that.' This was my immediate reaction to Ryan North's How to Invent Everything. The central conceit manages to be both funny and inspiring as a framework for writing an 'everything you ever wanted to know about everything (and particularly science)' book.

What How to Invent Everything claims to be is a manual for users of a time machine (from some point in the future). Specifically it's a manual for dealing with the situation of the time machine going wrong and stranding the user in the past. At first it appears that it's going to tell you how to fix the broken time machine - but then admits this is impossible. Since you're stuck in the past, you might as well make the best of your surroundings, so the aim of the rest of the book is to give you the knowledge you need to build your own civilisation from scratch.

We start with a fun flow chart for working out just how far back in time you are…