AI has an Awful Image problem
Stay informed with free updates
Simply sign up to the Artificial intelligence myFT Digest — delivered directly to your inbox.
The techies of Silicon Valley are currently buzzing with happy talk about how AI is changing everything. They may well be right. But whereas most of them assume this will be for the better, the US public appears to think it will be for the worse. AI dread stalks the heartlands amid worries over job losses, child safety and the environmental impact of giant data centres.
The chasm of perception between industry and society in the US is highlighted in Stanford University’s latest AI Index report. Whereas 73 per cent of AI experts think the technology will have a positive impact on people’s jobs, only 23 per cent of the public think the same. It’s a similar story with the economy and medical care. Most alarmingly, those likely to live longest in the AI-defined future appear the most alarmed by it. Among US voters aged 18 to 34, AI’s net favourability rating is minus 44, according to an NBC poll.
In one instance, distrust of AI has even spilled over into violence. A 20-year-old man was arrested on Friday after throwing a firebomb at the San Francisco home of Sam Altman, OpenAI’s chief executive. Federal officials said the man was in possession of an “anti-AI document” warning of the dangers of humanity’s extinction.
The mixed messaging of the industry has not helped. On the one hand, AI executives promise to cure disease, tackle climate change and usher in an era of radical abundance. On the other, they warn of job disruption, cyber and bio threats and existential risk.
As with previous economic upheavals, such as the industrial revolution and globalisation, AI’s economic gains are likely to be generalised while the pains are localised. It is also easier to identify the jobs upended today than to imagine those created tomorrow. The internet created demand for software coders, for example, but AI may lead to their replacement by prompt engineers. The Luddites get a bad rap for resisting the industrial revolution. But they were rational economic actors: it took decades before ordinary workers benefited from the revolution’s productivity gains. Still, we are in the early days of the AI transformation. Three factors may yet reshape the public debate.
First, AI is unusual in that it drastically cuts the cost of intelligence; it is an invention for inventing. Not only can the technology boost productivity, it can also transform creativity and scientific discovery. As James Manyika, Google’s senior vice-president for research, technology and society, tells me: “AI is the Industrial Revolution plus The Enlightenment.” Much of the talk at the moment is about AI’s “denominator” effects, reducing the costs of producing goods and services. But it will increasingly focus on its “numerator” effects, enabling innovation and new types of business to be created, driving additional revenue.
Second, users are increasingly realising the benefits of AI for themselves. The adoption of AI has been faster than previous technologies. Within three years, generative AI has reached 53 per cent of the population in the countries surveyed, according to the Stanford study. Thousands of start-ups are applying that technology in beneficial ways, to explore new ways of developing drugs and personalising education, for example. The increasing use of AI agents is also stimulating a burst of individual creativity and economic opportunity. That might yet lead to the emergence of “political superintelligence”, in the phrase of Andrew Hall, a political economy professor at the Stanford Graduate School of Business. AI can itself be used to better inform the debate, test solutions and improve governance.
Improbable as it may seem today, a more responsive regulatory regime is likely to develop in the US at the federal level, too. The Stanford report found that Americans had lower trust in their own government to regulate AI effectively than in any of the other 29 countries surveyed. There is clearly strong demand for more protective legislation, reflected by the 145 AI-related bills passed in US states by 2025.
Finally, if AI really can boost productivity then that will itself change the terms of the debate. Dario Amodei, Anthropic’s chief executive, suggests that over the past few decades economic growth has been difficult but creating methods of redistribution has been (relatively) easy. But, he tells me, AI may yet flip that calculation.
Existing tax systems have been designed to underpin generous welfare states in many countries. But in a world of abundant growth, and increasing concentration of economic wealth, the mechanisms of redistribution will have to be rethought. The logical answer would be a radical shift in taxation from income to capital. We are not there yet but we may be one day soon.
