High-speed pneumatic tubes! Helicopters! And a low bridge.

• Why do trucks keep smashing into this bridge? Julian O'Shea's videos on Melbourne are really well produced: and this one, about an Australian bridge that keeps eating the tops of trucks, is a great example.
   
• Andrew Steele, who's literally written the book on aging, makes a great debunk video of a thread on someone "reducing your biological age". Surprisingly, some of the conclusions aren't "this is rubbish": instead it's "the evidence isn't in yet".
   
• Remember that terrible "food, but as processed squares" startup that did the rounds a while ago? Well, Drew Gooden has eaten the squares.
   
• And while it doesn't quite fit the "surprising twists" thread of this week's recommendations, I will say that the introduction on this video, "Blind students learn lockpicking" is exactly right. The video's a good introduction to lockpicking, but it's also an example of how almost any teaching curriculum can be changed to accommodate anyone. (thanks to Lexi for sending this over!)

• Teen vaping has eliminated teen smoking, at least in the US — but now vaping isn't cool any more either. So while it's impossible to predict the future (maybe there'll be a retro trend for cigarettes in a few years), there's a chance that America has just managed to accidentally end teenage smoking for good.
   
• Art collective MSCHF have incredible concepts: turn up at Art Basel, and install an ATM with a leaderboard that shows everyone's balance. There's a lot to criticise about MSCHF's position at the intersection of "commercial publicity-stunt" and "art", but I can't deny their ideas are brilliant. Here's a clip of the ATM in action.
   
• There is an app for counting things, which is a much better idea than it sounds.
   
• But the star of the week for me, as it has been for so many people, has been ChatGPT, OpenAI's chatbot. If you've not had a conversation with it, then try signing up, it's very much worth a go. A few thoughts, which means hold tight, we're going to nested bullet points:
  
  
  • As text and image generators have improved over the last year or two, I've been occasionally surprised, but never more than that. But ChatGPT unnerved me: it's good enough that it sometimes appears to have proper comprehension rather than just text-prediction. Sometimes.
     
  • I asked it to solve moral dilemmas, and while it hedged its bets often (and clearly ran up against some sort of internal filtering frequently, too), it usually came up with a reasonable response. I asked it to guess the object I was thinking of by asking clarifying questions, and it got "toothbrush" in a very efficient manner. It held a conversation.
     
  • Of course, people are finding plenty of ways around its ethical restrictions. There seems to be a cat-and-mouse game going on as the folks running it add more content filters... and then users get around them.
     
  • In short: give it a try. Just don't pester your friends too much with what you've made; after a while it just feels like someone telling you about their dreams!

And finally: would you like to see the only available footage of the most terrifying waterslide ever constructed, which was fully submerged so the rider has to hold their breath from entrance to exit? Of course you would. Alas, it was dismantled about ten years ago, otherwise I would absolutely try this.
 

Next week: a lot.

All the best,

— Tom