Good morning. Quite a high proportion of journalistic output on the sudden explosion in generative AI is now prefaced with an announcement that the author has asked ChatGPT to write on their behalf. I am much too motivated by self-interest to follow that approach, which seems to me less like turkeys voting for Christmas than turkeys slathering themselves in butter, turning the oven up, and hopping on in.
At the fringes, there are already ominous signs of the outlook for content drones like me: witness, for example, the news that BuzzFeed has published a series of (quite bad) travel guides bylined “Buzzy the Robot”. Meanwhile, there have been reports of illustrators replaced by AI image generator Midjourney, and nearly half of business leaders in a US survey said that they expected layoffs as a result of the use of ChatGPT before the end of the year.
Last week, a report by Goldman Sachs predicted that 300 million full-time workers could lose their jobs to automation in the US and Europe alone. At the same time, Goldman suggested that those losses could be offset by the creation of a whole range of new occupations connected to the emerging technology – and a global productivity boost could ultimately be the result.
The truth is, all of this is so new and so unpredictable that nobody really knows. Today’s newsletter, with the Guardian’s technology editor and author of the brilliant TechScape newsletter Alex Hern, can’t tell you whether you’re going to be replaced by a robot. But it might help you get a feel for the risks and rewards. Here are the headlines.
Five big stories
Health | Thousands of children experiencing “unacceptable” long waits for NHS treatment face a “lifelong” impact on their health, the president of the Royal College of Paediatrics and Child Health has warned, as shocking figures reveal that nearly 15,000 paediatric operations were cancelled over the last year.
Finland | Finland’s prime minister, Sanna Marin, has lost the battle to stay in power after her centre-left Social Democratic party was narrowly beaten into third place by conservative and far-right rivals in Sunday’s elections. The leader of the conservative NCP, Petteri Orpo, is expected to open coalition negotiations on Monday.
Sex trafficking | Rishi Sunak is to announce new measures to tackle grooming gangs on Monday, claiming that “political correctness” would not get in the way of a crackdown. On Sunday, home secretary Suella Braverman was accused of “dog whistle” rhetoric after singling out British Pakistani men over the issue.
Russia | A prominent pro-war Russian military blogger has been killed in a blast in a St Petersburg cafe. Some 30 people were injured in an explosion that killed Vladlen Tatarsky, real name Maxim Fomin. A St Petersburg woman previously detained for taking part in anti-war rallies was arrested, the Interfax news agency said.
Music | Ryuichi Sakamoto, the Japanese musician whose remarkably eclectic career straddled pop, experimentalism, and Oscar-winning film composition, has died aged 71. Read Alexis Petridis’ tribute.
In depth: ‘We’re looking at a world where unskilled labour is still vastly useful in a vaguely bleak way’
In trying to describe the impact AI can already have on a business, Alex Hern suggests considering the arrival of five precocious Harvard graduates on an internship programme. “That’s incredible, right?” he said. “Loads of businesses would kill to have five free Harvard graduates working for them. But they are still 21 years old. They have some genuine specialised knowledge, but lots of bravado. And you would want to babysit them until you understood what they were good at and what they weren’t. And that’s roughly where we are today.”
Systems like ChatGPT are astonishingly plausible, and very often fulfil their assigned tasks effectively – but they also make weird and unacknowledged mistakes that it takes a human to notice. Witness the macabre sausage-pile excuses for hands in so many images generated by Midjourney or DALL-E, or ChatGPT’s sadly untrue belief that I am editor of the Evening Standard and author of ‘The Atheists Guide to Christmas’. That means humans are still indispensable, even if their roles change. The question is whether that’s a problem that can be ironed out in future iterations, or a fundamental feature.
Here are some other ways to think about what might happen next.
***
We don’t know what we don’t know
The man who invented the garlic press in 1950 probably had a pretty good idea of what it would be used for, and – novelty experiments aside – he hasn’t been proven wrong since. Generative AI is not like a garlic press. “The term of art is ‘capability overhang’,” said Alex. “When you make these things, you release them and then you work out what they can do. They have emergent capabilities that you didn’t expressly set out to give them.”
Some of these are easy to figure out: even if you train “large language models” like the one ChatGPT is built on with general text, you can easily give it a maths question, and see if it can answer it. “But others are harder. It’s going to take a while to work out which domains it is very good at, versus able to give a show of being good at, versus able to get very good at if you ask the right question.”
***
Is AI a steam engine or a car?
In the most radical scenario, the future of generative AI is that it will quickly prove to be a kind of iPhone for everything: a technology which utterly transforms the landscape across every field, making the state-of-the-art archaic, and leaving office assistants and brain surgeons equally redundant.
But it might be too soon to assume such universal transformations. “A lovely analogy that’s often brought up is the invention of the steam train,” Alex said. You might assume that the doughty old horse was immediately obsolete – “but actually the number of horses being used increased by an order of magnitude, because it suddenly became much more valuable to be able to transport goods to the railway station. But when cars were invented, they could go almost everywhere a horse could go.”
The current generation of AI models “feel more like steam engines than cars,” Alex said. “They are extremely good at doing huge chunks of stuff humans do. They’re not good at obviating the need for people – so the hope is that that means that they massively boost the economy, and create work that only people can do.”
***
‘Human work’ isn’t necessarily fulfilling work
If you find this comforting, not so fast: while we tend to think of “human work” as creative and nourishing, the future isn’t all ballet and portrait painting.
“One thing that only humans can do is read a sample of generated erotic roleplay texts to make sure it doesn’t constitute child abuse,” he said. Another is sausage-hand image corrector. Alex points to a Reddit post by someone describing how Midjourney transformed their job as a 3D artist for a games company from creator to AI facilitator: “The reason I went to be a 3D artist in the first place is gone. I wanted to create form in 3D space, sculpt, create. With my own creativity.”
That’s obviously a fairly marginal case, but it may be the thin end of the wedge. With the caveat that broad predictions are bound to be unreliable, Alex suggested one way this could shake out: “We’re probably looking at a world where unskilled labour is still vastly useful in a vaguely bleak way. But we’re going to see deskilling in a lot of industries.”
That would follow the pattern of how technological change has disrupted labour markets in the recent past: whereas waiters and doctors do non-routine tasks and have therefore been relatively safe, this EU/US report from last year says that “typical automation technologies have decreased demand for middle relative to low-paid and high-paid occupations, resulting in a process of job polarisation.”
***
Getting this right isn’t really about the AI
The question about how to deal with this kind of scenario is probably about regulation and political priorities, not the AI itself. The same EU/US report argues that governments will need to invest in new training, regulate the role of AI in hiring decisions, and encourage the development of the technology in directions that benefit society as a whole and don’t just maximise profit.
For example, said Alex, “if I can go to a solicitor and they produce their advice through ChatGPT without disclosing it, and it meaningfully reduces its quality, it’s very important whether or not that solicitor then gets struck off.”
Whether that will happen is another question. In the UK, “it’s hard to see the current government wanting to intervene to stop an innovative company doing what it wants,” Alex said. “If you said, for example, that you need to be clear that you’re offering a worse service using AI instead of humans, that would have a very different effect on employment.”
***
Some versions of the future are just too wild to plan for
A lot of the above, Alex noted, “needs an asterisk. It assumes the outcome isn’t that each version of GPT is so good at generating good data to train the next one that you get a flywheel effect, and we have a superintelligent AI in five years.” And, to state the obvious, a world where an AI can write Proust-level novels which also features 90% unemployment has bigger problems than whether it understands what it’s writing.
Rather than talking about knotty concepts like consciousness, the word that gets used to imagine what might happen in such scenarios is “agentic”: a world where AI is not just a tool to be deployed but can play an active role in changing the world.
“Preparing for that world is hard,” Alex said, and it is unlikely either in five years or 50. “I think it’s much more likely that we end up in a world that looks broadly like ours, but with quirks at the edges, where some number of jobs have been changed, a smaller number don’t exist, everyone is a little bit richer, and some diseases are cured.”
As for the scenario where an elite of trillionaire overlords are extracting wealth via the use of AI from a vast and obsolete underclass: “The AI would have to be so powerful, and so deeply weird, that it’s changing things that are even more fundamental than the structure of employment.” In other words, it’s hard to see it settling for being a fatcat’s plaything.
What else we’ve been reading
Pjotr Sauer’s piece about his friend Evan Gershkovich, a Wall Street Journal reporter arrested on a bogus espionage charge in Russia, emphasises the importance of continuing to press for his release as the story recedes from the headlines. “Evan did everything he could to tell the story of modern Russia,” Pjotr writes. “It is now our turn to keep the light shining on him.” Archie
The Guardian’s Leila Latif reviews the third and final season of the docuseries Surviving R Kelly, which – unlike its harrowing predecessors – offers hope and a sense of victory in the wake of the singer’s convictions for sexual abuse. Hannah J Davies, deputy editors, newsletters
What could be more nosily fascinating than Saturday magazine’s photos and stories of nine singletons, couples, throuples, and families in their beds? Nothing much. Nobody looks quite as peaceful as 70-something Kate and her whippet. Archie
Zoe Williams interviews Michael Bublé, and finds him extremely corny, but also very nice: “He skates this line between panto and passion, concert and royal visit, joke and sincerity, and it works because, whatever it is, he really means it.” Archie
Jamie Fisher has written a brilliant piece for The New Yorker (£) on the complicated legacy of Elliott Smith, and the fans totally in thrall to their late musical hero. Hannah
Sport
Premier League | Two top flight managers lost their jobs on Sunday, with Chelsea removing Graham Potter after a disappointing six months in charge and Leicester City parting ways with Brendan Rodgers after a winless streak of seven matches. Barney Ronay wrote that Potter was “the ultimate slow-burn process manager, thrown into a chaos of panic-capitalism”. Meanwhile, Newcastle beat Manchester United 2-0 (above) and West Ham beat Southampton 1-0.
Women’s Super League | Katie McCabe’s 74th minute goal secured a 2-1 victory for Arsenal against Manchester City, putting the hosts level with their opponents and within three points of league leaders Manchester United in a thrilling title race.
Formula One | Max Verstappen won the Melbourne Grand Prix but led criticism of the sport’s governing body after the race was stopped three times because of incidents on the track. Questions were raised as to whether the stoppages, which closed the field up and were followed by dramatic standing restarts, were employed to improve the spectacle.
The front pages
The Guardian leads with, “NHS delays ‘risk harming thousands of children’”. The Mirror reports home secretary Suella Braverman has claimed thousands in expenses on her London home, with: “Guess who doesn’t have to worry about energy bills”. The paper adds that the claims are “within the rules”.
The Times previews a speech by the prime minister under the headline, “Child abuse gangs ‘fed by political correctness’”. The Telegraph has the same story: “Ethnicity of grooming gangs cannot be ignored, police told”.
The Financial Times leads with, “Oil producers spring surprise output cut of more than 1mn barrels a day”. Finally, the Mail splashes with, “Millions of drivers stuck in parking app hell”, as more pay and display meters are scrapped in favour of cashless alternatives.
Today in Focus
Cotton Capital: the bee and the ship – examining the Guardian’s links to slavery
In episode one of a new Guardian podcast series, Maya Wolfe-Robinson explores the revelations that the Guardian’s founding editor, John Edward Taylor, and at least nine of his 11 backers had links to slavery, principally through the textile industry
Cartoon of the day | Edith Pritchett
Sign up for Inside Saturday to see more of Edith Pritchett’s cartoons, the best Saturday magazine content and an exclusive look behind the scenes
The Upside
A bit of good news to remind you that the world’s not all bad
2023 marks 100 years since the Flying Scotsman locomotive was built. It is something of a mechanical celebrity, and its anniversary is being commemorated with an exhibition at the National Railway Museum in York, special excursions, and visits to heritage railways. It was one of the original anti-car icons, giving it – on reflection – much environmental credibility too.
The Flying Scotsman is not the only relic of the battle between railways and roads: Andrew Martin offers six of the best British heritage railways. With his list spanning from the Highlands to Yorkshire and Dorset, chances are there’s probably one near you.
Sign up here for a weekly roundup of The Upside, sent to you every Sunday
Bored at work?
And finally, the Guardian’s puzzles are here to keep you entertained throughout the day – with plenty more on the Guardian’s Puzzles app for iOS and Android. Until tomorrow.