“Oikophilia”: the new driving force in politics

Heroes
The family behind Britain’s biggest toy shop chain, The Entertainer, who are handing over ownership of the business to its 1,900 employees. Gary and Catherine Grant opened their first store in Amersham, Buckinghamshire in 1981, and the company now spans 160 shops and more than 1,000 concessions. Gary says it wouldn’t have felt right to sell the family business “just for money”.

Villain
David Lammy, who has referred himself to the environment watchdog after going fishing with JD Vance without the required licence. The two men went angling for carp in a lake at Chevening House, the Foreign Secretary’s official residence in Kent, despite not having a rod licence – a breach of the rules punishable by a fine of up to £2,500. Worse, Lammy didn’t catch a thing.

Villains
Starbucks customers in South Korea, after the firm had to bring in new restrictions on “excessive personal equipment usage”. So-called cagong – people who sit for hours in coffee shops working or studying – have been bringing in full power strips, desktop computers and printers. One even set up a three-sided partition to create a private cubicle.

Villain
Clive of India, after the Labour peer Thangam Debbonaire called for a statue of the controversial British Empire administrator to be removed from outside the Foreign Office, saying its depiction of “subservient Indians” was damaging international relations. Clive appears to be guilty of “that most common of modern crimes”, says Douglas Murray in The Spectator: “being found to have lived in the past”.

Villain
The Environment Agency, for suggesting that people could save water by deleting old emails and pictures because data centres require “vast amounts” of H2O. Gary Barnett, a technology analyst, called the advice “really, really silly”, noting that people would need to delete tens of thousands of emails to make up for a single extra second in the shower.

Prime ministers rarely have a truly relaxing time on holiday, says Matt Chorley in The New Statesman. Margaret Thatcher got so bored on one trip to Salzburg that she set up a meeting with Helmut Kohl, who was staying nearby. When the German chancellor cut their seemingly unnecessary chat short, citing a crisis or “some other reason to get away”, Thatcher took herself off for a walk, only to stumble upon Kohl tucking into an ice cream. David Cameron was photographed awkwardly wriggling out of damp trunks with a Mickey Mouse towel tied around his waist, and once had to leave his Cornwall break due to the lack of phone signal – he’d been climbing to the top of the nearest hill to take calls from President Obama. Then there was Winston Churchill, who, while holidaying in the French Riviera in 1934, hurtled down a bright blue water slide headfirst with such ferocity that his trunks fell off. “It was all caught on camera.”

Is the AI hype overdone?

Much of the “euphoria and dread” about artificial intelligence is based on one big assumption, says Cal Newport in The New Yorker: that the more data you use to “train” these large language models, the more they improve. At first this “scaling law” held true. OpenAI’s GPT-3 was “leaps and bounds better” than its predecessor; GPT-4 marked another astonishing improvement. So when OpenAI released GPT-5 last week, many hoped it would usher in the next giant leap in AI capabilities. Instead, while the model is an improvement on previous incarnations, it has left the industry largely underwhelmed. And this has fuelled the “creeping fear” that the AI scaling law “wasn’t a law after all”. That – whisper it – AI “might not get much better than this”.

The tech titans now put their faith in so-called “post-training” techniques, where you tweak the model once it’s already trained. But sceptics say this is like the difference between building a new car and souping it up. Sure, you can improve performance, but “no amount of tweaking” will turn your Toyota Camry into a Ferrari. And much of the AI optimism – not to mention the hundreds of billions in investment – is based on getting a Ferrari, in the form of all-powerful “artificial general intelligence”. If the sceptics are right, AI tools will instead make only “gradual” advances in the coming years. Certain fields, such as programming and academia, will change significantly; some may disappear altogether. But AI won’t “massively disrupt” the job market or drastically transform our lives. Of course, no one really knows, and some huge development could be just around the corner – “the whole enterprise of teaching computers to think remains mysterious”. But a little less hubris wouldn’t go amiss.