0 comments
can AI stop waffling with such chest?
Is there anything more irritating than getting a response from an AI chatbot that you know to be complete bollocks?
Ask your shitty Alexa Echo voice assistant any question more complicated than ‘what’s the weather looking like?’ and there’s a great chance it’ll reluctantly say ‘sorry, I don’t know that.’ Ask an AI chatbot something it doesn’t know, and it’ll rustle together an ad-hoc answer expressed with all the overconfidence of one of those dweeb kids on University Challenge.
Both examples of modern tech failings are exasperating for different reasons, but with the former, at least you don’t come away completely misinformed about something you’ll embarrass yourself with later.
That’s the risk we all take when using LLMs for research purposes. Because they’re designed for everyone – not just researchers at MIT – they’re hardwired to placate users and congratulate them for half-gleaned assertions. They’re impressive and useful if you take the right measures, but they’re also little kiss arse machines.
People have apparently enjoyed the ego massaging so much that chatbots now have this tendency to blag baked in as a default setting. Researchers at OpenAI say that LLMs aren’t inherently deceptive, we’ve just trained them to value confidence over caution, to be agreeable above all else.
Putting this observation to the test, researchers at Carnegie Mellon University set out to understand what actually goes on under the bonnet when the likes of ChatGPT, Claude, and Gemini spin dodgy truths with all the conviction of a Trump infographic on Truth Social.
They gave each model a mix of tasks spanning trivia questions, image recognition, and prediction exercises, then asked them to rate their own confidence before and after answering. They wanted to see if they’d check themselves or continue firing back with chest when presented with their pretentiously-packaged fumbles.
Turns out that chatbots, like Piers Morgan, have no shame. There’s no hedging of bets or going rosy cheeked and staying quiet for five minutes. They remained as self-assured as ever, because they haven’t been taught what humility looks like. Non-answers or an admission of doubt are equivalent to hitting the eject button in a space shuttle. It has to and will produce something.
You could argue we’ve moulded LLMs to fit our modern day behaviours, then… chatting emotionally charged, ill-placed, pseudo-intellectual waffle with all the self-aggrandisation of an inscription on a museum wall.
ChatGPT also fucked my Thai yellow curry right up. Read more
👀 chronically online
If your algorithm isn’t full of clips of Clavicular, basically the modern harbinger of mogging (and a massive dickhead), I beg we swap. If you’re blessed enough to be unaware, ‘mogging’ is the verb of being in the presence of someone else, and outdoing them on the attractiveness scale. It’s toxic, sad, and bang on-brand with chronically online, brain rot culture. The Guardian has reported a livestream trend has now kicked off on ‘Omoggle,’ a spinoff service where strangers are ranked on facial symmetry, jaw structure, eye shape, and ‘sexual market value’ up against looksmaxxing influencers with chiselled jawlines and empty heads. The internet always finds new ways to shit on the self-esteem of the young ones. Maybe we do deserve a social media ban… maybe I’m just chopped. Read more
Clavicular instantly ENDED STREAM after he got BRUTALLY MOGGED in his first ever match on Omoggle 😭 pic.twitter.com/XKmC5Y3Thc
— Clavicular Updates (@Clav0Updates) April 30, 2026
Gen Z reports early cognitive decline. Here’s what to know about the brain rot epidemic and what to do about it – fast company
We undoubtedly spam the term ‘brain rot,’ but is the shit we consume for hours actually hurting us biologically? Research says young people are becoming increasingly worried that their lapses in memory and dwindling attention spans are traceable to veg-out sessions on TikTok. The good news is experts spoke to Fast Company and clarified that you’re probably not speedrunning a path to early dementia, but are more likely just suffering from overstimulation, poor sleep, and tiny dopamine hits that struggle to find a home elsewhere. The fixes are depressingly familiar: sleep more, scroll less, read books, don’t take your phone to the bog, touch grass, don’t write newsletters that require research into nonsense trends like mogging. Damn it! Read more
