It hurts to miss an unmarked shot in basketball. And it certainly seemed to pain the Alvark Tokyo shooter, halfway through April’s Japanese league game against Shimane Susanoo Magic. As the ball bounded off the rim, the player wheeled away, head lowered, eyes downcast. The disappointment looked glaringly real.

Which is interesting, because it was not. The player could not have cared less. They literally could not care at all, and not just because this was a half-time exhibition. It was because they were a robot, created by Alvark Tokyo’s team sponsor, Toyota.

CUE7 is not the first AI-powered b-baller, but it is the coolest-looking so far. At 7ft 2in, it wears the team kit well, gliding easily around the court with wheels for feet and hoops for hands. Red, jointed limbs sport a metallic sheen that might, on a human, be sweat. Alongside a cyclops-style eye, the cameras that help guide its movements have been fashioned into deeply familiar features, like the LeBron James-style beard and headband.

Its look and its movements are recognisable enough that it is impossible to watch it line up at the foul line without a twinge of anticipation. Is Big Red going to make the shot? Even when watching a highly edited clip, I found myself anthropomorphising the robot so quickly that I felt a guilty horror when I saw him – it, dammit, it – being hoisted away at the end of the demo by a crane. No athlete should suffer so undignified an exit from the arena.

For anyone who has never contemplated the rising dawn of robot athletes, the past two weeks may come as a shock. Just a few days before CUE7 made its debut, the Beijing half-marathon saw humans and robots running the course together, albeit in separate lanes. This is the second consecutive year the experiment has been tried: in 2025, several of the non-human competitors struggled to get away from the start line, and most failed to finish. The winning robot, Tiangong, which needed its battery replaced three times around the course, finished in 2hr 40min, twice as long as the quickest man.

Last month’s results, then, are a miracle of technological advancement (or, depending on your outlook, yet another death knell for humanity). All three robots on the podium beat the current world record, set in Lisbon in March by Jacob Kiplimo. Kiplimo ran 57:20; Lightning, a robot created by smartphone manufacturer Honor, was nearly seven minutes faster in Beijing.

Once, we could comfort ourselves that however good machines were at, say, chess, they could never mimic the super-fast physical reactions that occur so naturally and instinctively in a human athlete. Now, thanks to advances in AI, they can. Only a week ago, Sony AI shared footage of their table tennis-playing robot, Ace, winning three out of five matches against elite players.

Ace is not the kind of athlete anyone is putting on a magazine cover soon. There is nothing humanoid about its appearance at all – it is a robotic arm on a moving platform that looks like it manufactures cars in its spare time. While short clips of Ace in action against the professionals are diverting enough, it is hard to imagine a future where people tune in en masse for the spectacle. But that is not the ultimate goal of sporting robots, say the experts: their value will be in training flesh-and-blood sportspeople.

To date, sporting history bears out the argument. Bowling machines may be more consistently accurate, and capable of faster speeds, than the best paceman on your cricket team, but no one has changed the laws to include one in a team yet. Ever since Deep Blue beat the world champion Garry Kasparov exactly three decades ago, computers that can easily beat humans have established their place in chess as training companions. And Ace has already proved its worth in that regard: a former Olympian, watching the footage, saw it play a shot he had previously thought impossible.

The development of sports-playing robots generally is not about sport at all – rather, it is a focus for research and development in the wider scientific field. RoboCup, the annual football tournament which celebrates its 30th anniversary this year, aims to develop robots capable of beating the World Cup winners by 2050. The average football fan may consider that ambitious: in last year’s final, the half-height humanoids stamped around the pitch at walking pace, shuffling up to the ball and occasionally falling flat on their face. But whether RoboCup achieves its goal or not, the tech it has inspired is already benefiting society from search and rescue operations to Amazon warehouses.

There are plenty more sports that robots are already being taught to play, from badminton to pool, archery to martial arts, skiing to parkour (there is also a Tesla machine that does yoga, although what it gets out of it is unclear). But even some of those most invested are realistic about the more general appeal. Peter Stone, chief scientist at Sony AI and a former president of RoboCup, has said that after the initial “buzz” of robots achieving or surpassing human skills, “it tends to get less interesting” to the public.

Fundamentally, the philosophical question of what meaning there might be in any theoretical future contests between machines and humans remains unanswered. For humans, sport is a physical and mental trial, an effort of body and will. What pleasure can be had watching the exchange of shots or blows between contestants that experience no nerves, fatigue, or elation?

Which brings me back to CUE7’s on-court miscue. One of its older siblings – a 2019 incarnation named CUE3 – set a world record when it shot 2,020 consecutive free throws over six and a half hours. Perfect scores might be impressive, but they are also fundamentally dull. Perhaps with that miss last month, sporting AI learned its most valuable lesson yet.

• Do you have an opinion on the issues raised in this article? If you would like to submit a response of up to 300 words by email to be considered for publication in our letters section, please click here.