Replacing humans in jobs, violating privacy, creating autonomous weapons, becoming intelligent enough to act beyond human control: the apocalyptic ways in which artificial intelligence (AI) could go wrong are all too familiar. Something that is harder to envisage is a less gloomy alternative: a future in which AI harmoniously enhances human life.
AI seems to have cemented itself into modern life, both personal and professional. The latest findings from Gov.UK indicate that 73 per cent of the public have used AI at least once in the past month. Systems like ChatGPT, Claude and Gemini are streamlining tedious admin, planning holidays and becoming unexpected (not to mention unqualified) therapists. But what about the environmental impact of systems that run on huge data centres? Or the creeping sense that our growing reliance on AI is gradually corroding our brain cells?
These are some of the questions Dr Eleanor Drage broaches in her new book, What If We Got AI Right: How to stop catastrophising and build an ethical future. Drage, a senior research fellow at the University of Cambridge’s Leverhulme Centre for the Future of Intelligence, is considered one of the UK’s leading “AI ethicists”, but is sceptical about that specific job title.
“I don’t believe that there can be an AI ethics expert”
“I don’t believe that there can be an AI ethics expert,” she says. “Part of what I’m trying to show in the book is that expertise in AI is always distributed. These systems have effects on the people who use them that the people who build them don’t even understand. They’ve got environmental effects, and if we don’t treat the environment as an expert, as a stakeholder, then we miss that too. So the idea of expertise lying in a single person is probably the product of consultancies who are pitching us one individual to solve all our AI problems. I can’t back that.”
The Silicon Valley paradox
Drage’s central argument is that the biggest risks of AI are not the dystopian doomsday scenarios that consume us, but the systems, incentives and power structures currently shaping it. She argues that it’s the people behind these who are paradoxically pushing the apocalyptic narrative.
“I wanted to call the book The Apocalypse Fetish, but my publisher wouldn’t let me,” she says. “I was interested in why people get so excited about death and dying, and the excitement you see in Silicon Valley, for example. There’s a paradox that you see a lot with people like Elon Musk, who is like ‘we have to build AI’ but then signs open letters about how AI might also kill us and how we need to have a six-month pause on it. It has nothing to do with living a good life alongside AI. It’s just to do with how exciting it is to sit in a room and think about the possibility of a global apocalypse.”
Dr Eleanor Drage
But how do we actually go about getting AI “right”? We could start by decreasing its scale and making it more tangible in design, she suggests. “We really need to decide where we need something massive like OpenAI, and where we could use something smaller,” says Drage. “We have a small data centre in our office that we manage and that isn’t outsourced to Amazon Web Services. We understand what it does, where the data centre is, how much energy it uses. This turns AI into a real thing that we can touch and understand and make sense of. It stops us thinking that it’s the cloud, the Terminator or some sort of vapour in the ether.
“This is not a book about how bad data centres are. It’s about how there should be more small, local data centres that use the heat generated by them to power a sauna or sports complex. It’s about being smart and reasonable about improving AI infrastructure to stop people from panicking. Normal citizens can make good decisions.”
This involvement contributes to what Drage refers to as “AI literacy”, which allows us to analyse and understand systems and in turn lessen the sense that they are “dumbing us down”. This should extend to the workplace, Drage argues. The Government’s latest sector study of the UK’s AI ecosystem puts revenue at £23.9 billion, reflecting the increasing integration of AI across the economy, from healthcare to professional services. Plenty of workers are now being told to use AI by their employers, but Drage is keen for us to consistently question the real-life value of this.
“You cannot delete the human factor entirely”
“I think people will become more selective once we get more data on whether there is actually increased productivity and increased economic gain from various different kinds of AI at work,” she says. “At the moment, there isn’t much evidence to suggest that AI is improving people’s sustainable productivity. It’s making tasks more accessible — for example I wouldn’t have been able to code something two years ago and now I’ve got a bunch of tools that mean I can take on new tasks. But then I have more stuff to do. What you’re doing, effectively, is just increasing the workload. So there will be increased burnout. You cannot delete the human factor entirely.”
As for how we can integrate AI usefully at work, Drage again advocates for transparency. “Something that I think is especially important at work is just being really clear to yourself and your employer about how much a tool is helping you work less. We need a critical mass of people asking ‘why are we using this?’ and ‘what does this actually do?’.” Sometimes, the simplest questions are the right ones.
