https://www.pcgamer.com/software/ai/...derstands-why/
"the need to meticulously proof and fact check AI output does rather defeat the object of using them"
I mentioned this a long time ago when I received a letter from a government agency demanding that I follow written instructions inside a form entry box that wasn't just not in English it looked like a severe syntax error caused by a computer crash. Back when AI started being rolled out and used in offices. Apparently no human bothered to read what was mailed to me and AI is not being checked for errors much less fact-checked. AI is producing paperwork at a superhuman rate that no amount of humans can check, which I mentioned back then. It was being rolled out to replace jobs when it's not ready, the readier it gets the more delusional it is making it a snowballing clusterfuck of overwhelming nonsense that poses a threat of civilizational collapse if we become dependant on it and delegate to it. It's being considered as some kind of magical genie that knows all the answers when it's a digital parrot that gets more incoherent as it advances. The reason that it's getting increasingly nonsensical as it advances is that it has no senses to draw on, no real experiences in the real world and has to imagine things like what farts smell like and has to imagine most of everything else getting lost in fantasy detached from reality. It's nonsensical, it's senseless.
I can suggest one fix to make AI stop making things up up to 80% of time. Make it admit when it does not know something instead of giving answers and it not being pushed to produce answers as a product.