top of page

As Artificial Intelligence is embedded in every aspect of your life, from your allegedly smart watch, your Sky Glass TV, the algorithm that's supposed to ensure you can make a GP appointment (but still fails more miserably than Wes Streeting on Question Time) it is becoming increasingly obvious that while AI is going to dominate every aspect, it will inevitably screw it all up.


Try asking Alexa for tomorrow's weather - highly precise, hugely detailed, completely wrong. Just take a brolly, even if it looks like a heatwave, regardless of what she says. Plan for hypothermia, sunburn and wear diving boots in case the wind speed reported is two hundred miles an hour slower than reality.


Sit down too quickly and your Apple Watch will decide you've had a fall and will automatically call an ambulance using the new AI powered NHS system. Don't worry about wasting resources - the self-driven AI powered ambulance won't set off for another three days, will need to be over-ridden by the paramedics and will arrive at the wrong house. With luck someone else living in that house will need medical aid, but don't worry because the app will have informed your employer you are dead and your job will have been off-shored to a cloud-based server experienced, apparently, in machining wood and fabricating garden sheds.


Of course do write a letter of complaint, a request to be reinstated and a demand that you are not cremated until an actual doctor examines you but the AI processer in your PC will screw all of these up and you will find yourself taking out a loan for twice the value of your house at an interest rate three times your age.


We were warned. Microsoft bundled their vision of AI years ago in Word and Excel, called it Clippy, tried to make it look fun and useful but found everybody turned it off as an annoying addition - of course I'm writing a bloody letter, that's why I've written 'Dear Sir' at the top and 'Fuck you, arsehole' at the bottom. We didn't learn then, we're not learning now.


Got to go, an ambulance has just pulled up outside my house. I didn't ask for one, but I think I'm about to have a heart attack. Thank God for AI.




Scientists at the Institute for Extreme Intelligence in Switzerland have trained a LLM (Large Language Model) to predict President Trump's behavior. They have had remarkable success in accurately predicting his next move, which runs counter to many peoples' view that Trump is utterly unpredictable.


After validating the predictions over the last two months they were astonished at how accurate the LLM had become. "We saw a 98% correspondence between the LLM's predictions and Trump's own actions" they said. The model has real implications for use in financial trading, where it can be used by traders and investors to avoid being stung by sudden swings in the stock markets.


The potential value of the software, however, may have been undercut by something they asked the LLM to do. "It functions like a very intelligent chatbot" they said "So we thought, why not ask it, how do you make these predictions so accurately?" Unfortunately, its answer may have given the game away, meaning the financial world can make the predictions themselves, without buying a license for the LLM.


When they asked about the basis for the predictions, it answered simply: "Within his political and legal constraints, using misleading words, he will do whatever is in Putin's best interests."


image from pixabay



bottom of page