AI is already getting boring
One of your browser extensions seems to be blocking the video player from loading. To watch this content, you may need to disable it on this site.
Try again
Issued on: 26/04/2026 - 16:28
04:50 min Share From the show
Reading time
4 min
It could end white-collar work. It could end poverty. It could end humanity. From AI's boosters and doomers alike, bumptious predictions abound. But in workplaces, the technology, like so many before it, is becoming mundane. Are things really different this time?
Trains, electricity, the telephone. The great expectations and fears that marked their arrival eventually dissolved as they became another humdrum part of life.
But AI's different. That's the impression you get spending time at tech trade shows, at AI policy conferences or – God forbid – on LinkedIn. Its critics say it could end us, or at least end our ability to think for ourselves. Its proponents say it could end disease and usher in an age of riches for all.
The people building AI entertain both the worst- and best-case scenarios, painting alarming pictures of its power.
Dario Amodei, CEO of Anthropic, says we're close to "a country of geniuses in a data centre."
Sam Altman, CEO of OpenAI, says that "more of the world's intellectual capacity could reside inside of data centres than outside."
Elon Musk, CEO of xAI, says AI and robots could "saturate all human needs."
Meanwhile on the frontlines – the London office of architecture firm TP Bennett – they're just using it to "create custom precedents based on a story arc."
Director Simon Mason leafs through a wall of images that they've generated with AI for a project in Oman.
"We see it very much as just another part of what we do now," he says, "It's similar to the invention of AutoCAD from hand drawings ... It's just a natural evolution."
Last month, Anthropic published research on labour market disruption that says more than 80% of tasks required for architecture and engineering could theoretically be done by AI, though actual usage remains a fraction of that.
Chris Fulton will believe it when he sees it. He's Digital Director at architecture firm ADP and sits on the Royal Institute for British Architects' expert advisory group on AI.
He's held "races" between teams using AI and teams not using it. "You don't see different levels of efficiencies between the two teams. You just see differences in the output."
"The reality is that if you give people access to those tools, they may do an individual task quicker. But all that actually does is just move a bottleneck somewhere else in the process."
For AI-generated early concept designs, "You still need to judge them, critique them, make a decision about which of those hundreds of generations that you've just created is going to be good enough to put in front of a client."
He pointed to research by AECFoundry, which showed AI "visual reasoning" was only able to identify doors in a floor plan 32% of the time.
"There is no magic to these tools," said Fulton. "On a Tuesday, it might give you a really helpful answer. On a Wednesday, it might lead you to make a choice that would get your firm sued for malpractice. So that doesn't feel divine to me."
The story isn't unique to architecture. Anthropic's own paper shows that actual AI adoption remains much lower than theoretical adoption, even for markets where AI is most integrated, like computing, maths and administration.
We can't know yet
As with every new technology, it could take decades for us to see where the dust will settle, says Tom Standage, Deputy Editor of The Economist.
He's authored books about social media going back to ancient Egypt, and the telegraph, which he dubs "The Victorian Internet."
"You very often get this pattern," he says. "We see it with the telegraph in the 19th century. On the one hand, some people think it's going to lead to world peace. On the other hand, a lot of people think it's a scam."
Could AI be more like the internet, perhaps? After its introduction in the 1990s, it gave humans superfast access to vast amounts of knowledge. AI condenses this knowledge and makes it even more easily accessible.
The internet is still causing upheaval to this day, with various governments mulling whether to follow in Australia's footsteps and ban social media for kids.
That's normal, says Standage.
"We get these unexpected consequences with the internet, sure, but they’re unexpected reactions by humans to the internet."
As for AI bosses' "simultaneously optimistic and pessimistic" rhetoric, Standage sees another historical echo: nuclear weapons pioneer Robert Oppenheimer.
"Oppenheimer wants to be remembered as the father of the bomb, so he goes around apologising for it. By reminding you of its power, he's really telling you how awesome he is. And I think that's what the AI bosses are doing."
Recent coverage of Anthropic's new Mythos model – whose ability to exploit software vulnerabilities has raised alarm – has veered into the apocalyptic, with opinion pieces describing it as "superhuman", "terrifying" and capable of "breaking everything."
But those aren't the day-to-day concerns of people working with the technology.
"They are talking about existential risk to humanity," says Fulton. "I am talking about much more prosaic and very boring risk: that these models aren't very good. A lot of the time they are not clever. They don't reason, they don't think, they don't understand, they don't get context."
Beyond the claims of those selling AI and the increasing banality of those using it, there is one thing that might be different this time.
"People worry it might have its own agency," says Standage. "It might behave in ways that we weren't expecting and that we don't approve of."
"Railways are unlikely to do that."
Like every technology revolution, we can't quite know what we're building until it becomes baked into everyday life. And the norm for some won't be the norm for others. France has high-speed rail, America has traffic. Australian kids can't use TikTok, British kids can (for now).
But for the first time, the thing we're building might not always do what we tell it to.
By:
Peter O'Brien
Advertising
Related keywords