As the pace of technological change seems to become ever more speedy, it’s easy to just shrug and gloss over the terminology and accept what is said at face value – or at least it is for me. So self-driving cars are a couple of years away, facial recognition software works fine, and machines will be churning out Bach fugues before we know it. Also the words have become increasingly interchangeable to me – machine learning, algorithms, artificial intelligence and so on – without any real understanding of what they mean. Or whether they will be ultimately positive or negative in our lives.
So thank heavens for Hannah Fry who crucially has not only the knowledge and expertise to understand the technology, but also the examples to bring it to life and the ability to articulate and draw out the key aspects for readers like me. That is the central achievement of Hello World, which managed to make me feel about 10 times more informed than I was before starting to turn the pages, but not feel like I’d gone through a reading experience 10 times less enjoyable than a novel (which is an all too common occurrence of supposedly ‘accessible’ science books).
There is some genuinely fascinating stuff in here – and also a fair amount that is fairly terrifying. The use of algorithms in the criminal justice system in American states is a section that particularly sticks in the memory. The racial biases and more general failings of facial recognition software also stayed with me for some time. Fry is excellent at identifying the strengths and weaknesses of various technological developments, and where the difficulties really lie.
She is similarly grounded when it comes to transport: her description of the reality of self-driving cars, bolstered by the examples from air travel (where automation and its effect on pilots has so recently and tragically been in the news), is eye-opening – as well as the way in which she brings to life the sheer difficulty and technical challenge of doing so in any place which isn’t straight roads with unchanging environments. It was clear to me at the end of that section that we are way off a world that the car companies have been trying to sell us in recent years; and that that’s probably a good thing.
The book has two main conclusions. The first is that we should stop seeing humans and machines as either/or choices, and actually they could be better when working together – cars being a good example, where some automation (for example, braking systems on a Volvo) has been proven to supplement human reactions. The second conclusion is that openness of algorithms and approaches is crucial; because the only more scary thing than an algorithm making a wrong decision is if you can’t get inside the algorithm to see why and how. This is also connected to the fact that we tend to instinctively ‘trust’ algorithms because we have been conditioned to do so – making them more open would help to address that, and to see the human hand in design and implementation, and promote it in ownership and regulation.
All this, and much more – read it, it’s excellent.