Suleyman starts by saying that containment – that is, the restriction of a technology – never works. If something is possible, it will end up being done regardless of whether or not it is permitted. He supports this with a handful of cherry-picked examples, before making the surprising claim that nuclear weapons are the only technology that has ever been successfully restricted.
Suleyman then proceeds to regurgitate a few years’ worth of Silicon Valley press releases about how powerful the technology is about to become. He says that we don’t have to believe that all this is possible in the future for his argument to be true, but as someone who receives a lot of tech industry press releases, this reviewer is particularly sceptical about claims of technologies ‘just around the corner’.
He never quite sets out exactly what he feels the future problem will be, but most of the times that he tries to, his concerns seem to rest somewhere around the fact that large language models (LLMs), like ChatGPT, will soon become more intelligent than people. This, at least, seems to be the thrust of the solution he proposes in the final chapter.
Finally, Suleyman concludes with the solution: containment. The same containment that he dismissed at the start, only this time we’ll do it with our fingers crossed. Maybe it’ll work this time if we really mean it. However, it is worth pointing out that he seems to view the word containment as meaning something subtly different at the end than how he used it at the start. In the first chapter, when he talked about how it never worked, he used the word to mean the complete removal of a technology. Here, he uses it to mean something much closer to regulation. He doesn’t acknowledge this change in use of the word.
When used this way, it’s obvious that containment has been successful many times. Guns, cars, chemical weapons, electrical installations – even medieval technology such as swords and crossbows – are regulated, and while they do still have their dangers, they cause far fewer calamities than they otherwise would. The lesson of history, then, isn’t that regulation isn’t successful, it’s that it’s imperfect, frustratingly slow, but often ultimately effective.
We are perhaps being a little glib here. There’s no doubt that Suleyman is a highly intelligent person and is very well-placed to both see and understand the potential risks of AI. There is also no doubt that AI has progressed significantly in the previous couple of years, and it will probably continue to do so. While a civilisation-ending catastrophe is a bold claim, there are undoubtedly big changes coming.
AI is putting people out of jobs, though how far this trend will continue remains an open question. Suleyman very briefly touches on the potential for human suffering caused by increased unemployment, though this may end up being a more thorny question. Whether or not this trips over into Ludd-style rebellion, there is the possibility (perhaps even likelihood) of AI creating real suffering in communities. For example, regions where call centres account for a large proportion of the employment. Suleyman’s containment of advanced AI will not help them.
The End is AI? The rate of technological change feels like it’s increasing, with whole new types of technologies springing up all the time. Any one of these technologies could have catastrophic consequences for the human race. It’s surely only a matter of time until one of them does finally finish us off.
That previous paragraph feels a little scary, but it’s perhaps comforting to realise that this sentiment would feel familiar to many people at any point in the past 500 or so years. Many technologies have come and shaped our world and, so far, we’ve always been able to adapt to them, or them to us. Maybe we’ll get unlucky this time and AI will be the one that finally does us in, but there’s nothing in this book that convinces this reviewer that it’s more dangerous than many others.
It will, of course, create winners and losers – new technology always does. There will be a champagne reception in some areas and unemployment in others. It will be unfair, and it’s unlikely that those who lose out will be adequately supported. This, however, is not the civilisation-ending disaster Suleyman is concerned with in this book.
We’re not sure if this makes us optimists or pessimists about AI, but after reading The Coming Wave, our money’s still on it being climate change that finishes us off.
Verdict
7/10
Passionate and articulate, and ultimately unconvincing.
Buy if
You are more pessimistic than us
Avoid if
You’re more concerned about the implication of current technology than the as-yet undeveloped future products
