I have ChatGPT Plus and Gemini Pro (?) for testing purposes right now and it's shocking and scary how confidentially it is hallucinating like I asked it simple stuff like "did XY die in the season finale of Dexter" and both were making up stuff that didn't even happen.
I can't image how many people think it's only telling the truth like I hear coworkers prais it for doing competitor analysis so quickly and I am like ... how do you even know if what the results are is even true and not made up cuz it's trying to literally please you.
It also told me I need a visa (again, for testing purposes) for Oman when it's visa free for less than 14 days and when you tell it "you are wrong!" its always like "that's right! You don't need a visa ... Thank you for pointing it out "
And as more and more content is produced from this junk, just think how much worse it will become when the new garbage is absorbed into the LLM's. This is how the internet dies.
It can't even do painfully easy things properly: https://www.telegraph.co.uk/news/2025/03/14/ai-fails-reading-clocks-calendars-study-finds/