Book review: ‘Confessions of an AI Brain’

There are two major problems confronting any author wishing to write a manual on artificial intelligence. The first is that by the time you've written it, the glacial pace of the book publishing world has effectively rendered your work obsolete before the ink has dried. Second, there are simply so many books produced on the subject that it's virtually impossible to find a differentiating niche.

'Confessions of an AI Brain' (Springer, £25.00, ISBN 9783031259340) overcomes the first issue paradoxically by not attempting to say anything new, while the second is countered by the clever literary construction of telling the story from the first-person viewpoint of the AI 'brain' itself.

Miranda – as authors Elena Fersman, Paul Pettersson and Athanasios Karapantelakis have called their AI brain – is our patient guide throughout this thoroughly entertaining primer that might just as well have been called 'Everything you always wanted to know about AI but were afraid to ask'. The advantage in having the AI tell its own story is that the reader can be taken back to basics – reminding us of the difference between data, information and knowledge, for example – without being made to feel inferior.

This is more important than it sounds because while we all know that, given enough time, machine learning will work out the difference between cats and dogs, we may have forgotten how this is achieved. And we should remember this, says Miranda, because how else will be able to apply Maslow's hierarchy of needs to the AI pyramid that has data at the bottom and insight at the top?

Miranda is also helpful in providing end-of-chapter 'what have we learned?' bullet points. These are the 'confessions' of the title - and it's not much of a reviewer's confession to state that you could, if you chose to, read the book in ten minutes just by skipping to these bite-sized chunks of AI wisdom that will leave you better informed than when you started. All those embarrassing gaps in your knowledge about the internet of things and digital twins, bias and diversity, model drift and sustainability, get neatly plugged with Miranda's no-nonsense summaries, presented in (mostly) jargon-free vocabulary.

However, don't skip the main chapters, because for all its concision and humour – at one point Miranda suggests the optimal strategy for saving Earth is to get rid of the humans – 'Confessions of an AI Brain' is one of the best introductions to AI available. This is because it brings common sense to the general reader who has been bombarded by mainstream media sensationalism surrounding the subject.

My only quibble concerns the metaphysical rabbit hole of presenting Miranda as author, which means we have human authors pretending to be artificial intelligence in order to explain artificial intelligence to humans. On reflection, it all makes perfect sense. Wonderful stuff.