Some nights your dreams are lovely. You’re at the beach, everybody you like is there, and the whole thing makes perfect sense until you wake up and wish you could go back.
Other nights your brain apparently raided a pharmacy. You’re at work in your underwear while your third-grade teacher, who is now a giraffe, grades your presentation. Nothing connects to anything, and while it’s happening you buy every bit of it.
That’s AI. I use it and I like it, and I’m convinced it’s a dream machine.
When it works, it’s great. You ask for something specific and it hands you exactly that, faster than you could have done it yourself. Then you ask a perfectly normal question on a different day and get an answer with no logic and no connection to reality, delivered in the same calm, confident voice. It never says “I’m not sure.” It just keeps talking.
Google gave us the classic in May 2024, when its AI Overviews told people to mix glue into pizza sauce so the cheese would stick. Google’s head of search later explained that the answer came from a forum post. Another overview suggested eating rocks, which traced back to a satirical website. The machine read a joke and filed it under advice.
Funny when it’s pizza, less funny in a courtroom.
This June, Legal Cheek reported on a federal case in Mississippi where Judge Sharion Aycock found that lawyers on both sides of a dispute with the City of Aberdeen had used AI to draft filings and never checked the output. When the court went looking for the cited cases, several didn’t exist. The two lawyers who drafted the filings were barred from appearing before that court for two years and fined $2,500 and $3,500. The other two were disqualified and fined $1,000 each. With all four gone, the trial was cancelled. Four licensed professionals, and not one of them read the thing before signing it.
The part that gets me is how convincing it is. A dream feels completely real while you’re in it, and so does a bad AI answer. Both stitch together fragments that sound like they belong together without ever checking them against the real world. The difference is that you wake up from a dream and shrug. With AI, you’re the one who has to wake up, and nobody rings the alarm for you.
I’ve written about AI before, from the Amazon layoffs it got blamed for to the chatbot “friends” that aren’t helping anybody’s loneliness. The common thread is people trusting it a little more than it has earned.
So here’s how I treat it. An AI answer is a dream journal entry, not a witness statement. Enjoy the good ones, and if the answer matters, whether that’s a legal brief or a pizza recipe, check it against something real before you act on it.
My two cents: If a dream told you to put glue on your pizza, you’d laugh it off, so give the chatbot the same courtesy.