When an MIT professor called Joseph Weizenbaum created one of the first chatbots in the mid-1960s, he received two shocks in quick succession. The first was how readily people anthropomorphised the rudimentary program, which he called Eliza. “What I had not realised is that extremely short exposures to a relatively simple computer program could induce powerful delusional thinking in quite normal people,” he wrote in his 1976 book Computer Power and Human Reason.
But something else troubled him even more: how swiftly people began to envisage what technology like this could do one day, without pausing to ask what it should do.
In 1966, for instance, researchers at Stanford University wrote a paper that envisaged using a computer program similar to Weizenbaum’s as a form of psychotherapy. “Because of the time-sharing capabilities of modern and future computers, several hundred patients an hour could be handled by a computer system designed for this purpose,” they wrote.
Weizenbaum was appalled. He thought there were “some human functions for which computers ought not to be substituted. It has nothing to do with what computers can or cannot be made to do,” he added. “Respect, understanding and love are not technical problems.”
READ MORE
In his book, which turned him into a self-described “heretic” in the field of computer science, he argued that the most important issues were “neither technological nor even mathematical; they are ethical. They cannot be settled by asking questions beginning with ‘can’. The limits of the applicability of computers are ultimately statable only in terms of oughts.”
Half a century later, a strange hush has descended over the labour market in many countries – not least the US, where artificial intelligence has expanded from the nascent field it was when Weizenbaum was writing in the 1970s to a sector that has propelled the stock market to record highs.
Yet, notwithstanding the splashy headlines every few days about tech companies making AI-driven lay-offs, very little is actually happening in the labour market overall. The rate at which people are getting fired is low. The rate at which people are getting hired is low. The rate at which people are quitting their jobs is low. It’s as if everyone is holding their breath and waiting for what they have been told is coming.
[ ‘More harmful than helpful’: Young people sour on AIOpens in new window ]
Last month, Mat Honan, the editor in chief of the MIT Technology Review, wrote an unusual, meditative essay titled The Era of AI Malaise that captured the mood. “It’s buying things while we sleep. It’s discovering the structure of proteins. It’s telling children to kill themselves. It’s telling children to kill themselves,” he wrote. “Will I have a job tomorrow? Will the market crash? Why does OpenAI need a bunker? Do I need a bunker? Maybe I should have a bunker.”
And while we wait, we forecast. Endlessly, we forecast. Among government departments, universities, think tanks and AI labs, a cottage industry has grown up in producing technical mapping exercises, many of which use large language models themselves to compare their capabilities against all the tasks that supposedly go into each person’s job, in order to predict which are most “exposed” to automation.
Never mind that these predictions vary a lot depending on which model is used. And never mind that “exposed” doesn’t tell you anything about whether your job might get better as a result of new AI tools, or worse, or disappear altogether.
And so we are told that 40 per cent of the jobs in the world are exposed to AI; or maybe it’s 300 million; or perhaps only 92 million jobs are really in trouble.
Weizenbaum would not be surprised by this obsession with forecasts. Back in the 1970s, he already saw that people were becoming distracted by the “unnecessary, interminable and ultimately sterile exercise of making a catalogue of what computers will and will not be able to do”.
The metaphors we use are part of the problem. It is often said that a “wave” or “tsunami” of technological change is heading our way (indeed, I have written in these terms myself in the past). Mustafa Suleyman, chief executive of Microsoft AI, wrote a bestselling book called The Coming Wave. According to the book by Quinn Slobodian and Ben Tarnoff, Elon Musk once described his “Doge” efforts to overhaul the US government as akin to cleaning a beach of “needles and faeces and trash ... [But] how much does cleaning the beach really matter if you’ve got a thousand-foot tsunami about to hit?”
Metaphors are powerful because they shape our sense of what seems reasonable and unreasonable. What does the metaphor of AI as a tsunami imply would be a sensible course of action? To forecast – certainly. To map out the contours of where it is going to hit hardest; who it’s going to affect; who needs to prepare. What else? To get ready to mop up afterwards: to compensate those whose livelihoods have been swept away.
[ AI is moving fast. Official Ireland is notOpens in new window ]
And what does it imply would be an unreasonable response? To think you have any say over the wave itself: its pace; its trajectory; what it should be allowed to do and not allowed to do. Nobody wants to look like the fool who thinks you can hold back the tide. These metaphors also cast technology company chief executives in the role of earnest weathermen, wanting to let people know what is coming, and expecting to be thanked for the warning.
But technological change is not analogous to a natural phenomenon. How technology changes the world is shaped by institutions, laws, consumer demand, regulation, management strategies and the balance of power in different workplaces. This is not an aspirational statement, it is a simple matter of fact.
Why are there more radiologists now than there were in 2016 when Geoffrey Hinton, the “godfather of AI”, said we should stop training new ones because AI could already do their jobs? Because the expertise and daily work of a radiologist is much more than just reading scans. Because when some types of products or services become cheaper to provide, thanks to a productivity-boosting new technology, demand for them simply expands. Because insurance companies are nervous about insuring fully autonomous systems that – if they develop a fault – can make mistakes at a pace and volume that a radiologist having a bad day never could.
Why are some local journalists in Cleveland, Ohio, being told that AI will now write their stories for them, while I – another journalist – work for an organisation that has said its articles will continue to be written by humans? Because our companies have different business models, different people at the top, different customers to please.
What this implies is that the space between what the machines can and can’t do, and how the world does and doesn’t change, is a vast one – and one that will only be filled in by a series of human decisions, arguments, markets, battles, compromises and preferences.
I have spent the past few years working on a book about people on the front lines of technological change, from truck drivers and warehouse workers to translators and Hollywood writers. One of the things that struck me most was that none of these people were standing paralysed as a “tsunami” washed over them.
As someone who escaped the Holocaust as a boy, Joseph Weizenbaum knew that it is a dangerous thing when people begin to see other people as ‘non-player characters’ – and even more dangerous if they begin to see themselves that way
When change turns from something abstract but looming to something real and present, people don’t just close their eyes and hope for the best. They don’t sit passively in the ruins of their former careers, waiting for the government to compensate them in some way. They take action, individually and collectively, to seize opportunities, or pivot away from risks and also, when circumstances allow, to actively shape the technology itself and the way it is implemented.
At the individual level, I met highly experienced professional translators who were angry and sad that machine translation technologies had stripped their work of meaning and pleasure: that much of the work, now, was to “post-edit” the translations of machines: to check and tidy them up, at twice the speed, for half the price.
But they weren’t just accepting it. Some were moving into adjacent professions; others were changing their own business models to sell their services directly to clients, rather than via agencies. One of my interviewees, Rebecca Porwit, told me she simply refused to see herself as a “loser” from technological change. “It takes away all our sense of agency,” she said. “I’m just not willing to let that happen. I’m going to try and fight for this.”
In the workplace too I witnessed examples of workers and employers navigating new technology in a way that was far removed from the “employer insistence/employee resistance” narrative that seems to be taking hold in some American companies such as Meta. In countries such as Sweden, for example, which has a workplace system based on the principle of “co-determination” between workers and employers, people have already got decades’ worth of muscle memory of how to navigate technological change.
Deep inside a mine below a Swedish pine forest on the fringe of the Arctic Circle, I witnessed how that was playing out in the AI era, in a highly productive workplace now populated by self-driving trucks. In one instance, management wanted to introduce a real-time positioning system to keep track of where everyone was. The miners were worried about feeling surveilled, like Amazon warehouse workers.
So they came to a compromise: the mining company asked the technology vendor behind the system to develop anonymisation tags, so the company could see where everyone was, but not who they were (except in the case of emergencies).
Unlike in many US and UK companies, where there is a growing sense of mistrust between employers and workers over AI, the miners I met in Sweden had a seat at the table when it came to its implementation, and so felt less fearful about it. One told me: “We are not so technology-hostile. We are more friendly to doing new things. We see the possibilities.”
[ Message from the Editor: Everyone knows AI changes everything. Nobody knows howOpens in new window ]
In the realm of local and national politics, too, people have begun to shake off the feeling of paralysis and ask questions that begin with “ought”.
In the US, communities are protesting against new data centres. Individual states have been trying to regulate AI systems, and lawsuits have been proliferating, particularly in the realm of chatbots and child safety.
It should be no surprise that the safety of children has become one of the first issues to cut through. It was the same in the industrial revolution, where some of the early Factory Acts were attempts to prevent young children having to work long shifts in the new mills. Then, as now, even a society bewildered and disorientated by the pace of change was able to stir itself eventually to protect its young.
Indeed, the people who still seem most mesmerised are those who are supposedly at the steering wheel. It is tempting to suspect that tech workers in places such as Silicon Valley see themselves as masters of the universe and everyone else as “non-player characters”, as the bro-culture insult puts it – a term from the gaming world to describe characters in the background in computer games, who are just there to further your own game or add a bit of atmosphere. But the truth is stranger than that.
In Silicon Valley, which is plastered with billboards saying things such as “Stop hiring humans” and “That’s so agentic”, many people believe they are working on humanity’s last invention – that once AI reaches a tipping point, it will embark on recursive self-improvement, getting better and better on its own.
And once that happens, some think the social structure will become fixed: that there will be some winners and many losers. In a piece last month for The New York Times called Silicon Valley Is Bracing for a Permanent Underclass, the San Francisco-based writer Jasmine Sun’s first line was: “Most people I know in the AI industry think the median person is screwed, and they have no idea what to do about it.”
There is something telling about the word “permanent”, which implies a lack of faith in politics to address inequality of opportunity or outcome. That is perhaps no surprise, given the current state of US politics. But I suspect it also runs deeper. In 1992, the cultural critic Neil Postman wrote a book called Technopoly in which he argued that new information technologies were hastening the future Weizenbaum had worried about. Technopoly, Postman wrote, “is what happens when a culture, overcome by information generated by technology, tries to employ technology itself as a means of providing clear direction and humane purpose”.
If anywhere is surrendering to the ideology of technopoly, it is Silicon Valley – a place where people are racing towards “superintelligent” and “agentic” machines that will supposedly be able to solve all humanity’s problems for us, from poverty to climate change, but might also render a lot of people unemployable, or, in fact, just kill everyone. Tech leaders tell us they don’t know for sure which way it will go. All they know is that they can’t stop.
Last month, Pope Leo issued an encyclical on AI titled “Magnifica humanitas” in which he wrote that although “not everyone has the same power to make a difference”, still “no one is without responsibility” to shape the future.
[ Pope Leo’s encyclical on AI is a striking manifesto of Christian humanismOpens in new window ]
Someone else also spoke: Chris Olah, one of the co-founders of AI company Anthropic. Someone, you might have thought, who has rather a lot of power to make a difference. But he pleaded with the rest of the world to step in.
“Every frontier AI lab – including Anthropic – operates inside a set of incentives and constraints that can sometimes conflict with doing the right thing,” he said. “The pressure to stay commercially viable and to stay at the research frontier. Geopolitical pressure. And the older, plainer pressures of pride and ambition.
“No matter how sincerely any of us intend to do the right thing – and I believe many of us do – we will always be influenced by those incentives ... We need informed critics who will tell the labs when we are failing. We need moral voices that the incentives cannot bend.”
In his 1976 book, Weizenbaum wrote about something he called “civil courage”. “Every individual must act as if the whole future of the world, of humanity itself, depends on him. Anything less is a shirking of responsibility and is itself a dehumanising force, for anything less encourages the individual to look upon himself as a mere actor in a drama written by anonymous agents.”
That might sound a little grandiose for a book that is, at least ostensibly, mostly about computer programming. But as someone who escaped the Holocaust as a boy, Weizenbaum knew that it is a dangerous thing when people begin to see other people as “non-player characters” – and even more dangerous if they begin to see themselves that way.
My sense is that the era of “AI malaise” is coming to an end – at least, everywhere except the place that is supposedly in charge of it. Admittedly, trying to shape the application of technology through human debate, compromise and action can be wearying and effortful. Participating in politics, at the level of the workplace, the community or the nation, can be frequently enraging and unsatisfying. But it is better than watching passively as a handful of people race to make the future, when even they don’t seem to think they can be trusted with it. – Copyright The Financial Times Limited 2026
