Might an AI system one day decide to turn us all into paperclips?

This unlikely scenario was dreamed up by the philosopher Nick Bostrom. Suppose an AI has been given the only goal of producing as many paperclips as possible. Without restraints, it might come to the conclusion that, first, humans could choose to switch it off. It would want to prevent this.

Second, humans contain atoms that could be put to better use as paperclips. So we would all have to go.

Bostrom meant this as a thought experiment but, as Eleanor Drage points out in this stimulating book, many thoughts about the future of AI tend towards the apocalyptic. Alternatively, it’s presented as the answer to all our problems. AI will find the cure for cancer, end poverty, and create utopia. As Drage notes, ‘it’s hard to know what to think, and where we should direct not only our fear but our optimism’.

Is our future with AI going to be annihilation or salvation? The nuanced truth is that it’s unlikely to be either. In Drage’s words, ‘AI has become a buzzword.’ Startups use it to attract funding, it helps companies sell products, and it provides material for Hollywood film-makers to create dystopian movies.

The problem lies with Big Tech. Powerful companies such as Google, Meta and OpenAI control the information and the news that we receive. They are, Drage claims, ‘selling us a future which they also claim is the only one worth desiring’. And we shouldn’t be too concerned about all this – our real worry should be that AI will destroy us all. As Drage says, it’s an ‘odd paradox’ that ‘the people most worried about the apocalypse are also those who are building the machines which they claim will cause it’.

Drage, a leading researcher into AI ethics, argues that ‘the unfettered pursuit of profit’ is what makes AI dangerous, both now and in the future. Certainly not rogue AI eager to transform us into paperclips. The climate crisis, for example, is more likely to be exacerbated than solved by AI. Big Tech AI eats up resources. Google’s data centre in Dalles, Oregon uses more than 350 million gallons of water annually, a third of the city’s total water use.

By 2027, the AI sector will consume up to an estimated 134 terrawatt-hours of energy each year: the same as the Netherlands.

Her answer to the problems is for us to turn away from our dependence on Big Tech. The AI industry doesn’t just need input from computer geeks and technological wizards. It needs people who are ‘sensible and have life experience’. And she asks the question, ‘What if AI wasn’t built by Big Tech but by local communities?’ Drage argues that getting AI right shouldn’t just be left to Big Tech and those claiming to be experts. We all have a stake in this. It’s hard to disagree.