Many of us are surely now entering the ‘unbridled enthusiasm’ phase of our Olympic journey. In just a few days of competition, the scepticism has subsided. Now we’re obsessed with Pommel Horse Guy and pretending to high-dive off the sofa (or is that just me?). The story is invariably the same for the organisers. The pain and struggle of getting the event off the ground soon gives way to sporting mania.

But there has been a precipitous decline in the number of cities that are willing to go through the early suffering to get to the fun part. London competed with eight others to host the games in 2012. This time around, Paris was one of only two bids.

The main reason for the reticence is, of course, the fear of spiralling costs. But as we learnt this week, Paris looks to be on course to keep close to its budget and may end up with a positive games legacy. The future is even rosier for Los Angeles in 2028.

A less predictable downside of hosting the games is the outrage you might trigger when you invite a troupe of drag queens to participate in your opening ceremony, as Paris just found out. It was a classic case of an “anti-woke” backlash: anger based on a misunderstanding got passed between social media users until no one had any idea what they were mad about anymore. Here’s how to understand the newest Olympic event: the 4 x 400m culture war relay.

Hopefully it will be a few more years before we see AI participation in a sporting contest. So far, generative AI is struggling on some fairly basic tasks thanks to a tendency to ‘hallucinate’. This is the technical term for when an AI simply makes things up because of flawed data.

There is also such a thing as an ‘intersectional hallucination’, and these lead the AI to believe things that the average human would instinctively know couldn’t be true. You don’t need me to tell you, for example, that a child can’t claim a pension or that they couldn’t be a practising doctor. You know that it doesn’t make sense to say that someone is a husband but isn’t married. But AI doesn’t always realise this because of gaps in its synthetic data. Find out how these hallucinations happen here.