Sorry for the long story, here.
Yesterday, I fixed the washing machine. (Long story. Turned out to be the clutch, which is more of a pain to fix -- requires more disassembly -- than most other common fixes.) During the process, I jostled it and something metal fell out. A cotter pin sort of thing.
My mother checked on me while I was working on it and I mentioned the piece that fell out. She went and started researching about it. (Not something I asked her to do. I would have done the research myself had she not done so. But it didn't hurt anything to have her research. Or so I thought.)
She comes back 10 minutes later and says "this thing is a super important piece that holds something together somewhere. Better figure out where it fell out of." I looked her in the eye, knowing her penchant for reading Google AI overviews and asked "is that from the AI overview?" "No" came the answer. "Oh, some random forum or something?" "Yeah."
She leaves, I go back to the washing machine. (There was plenty I could still do even without an answer to the mystery of the random piece falling out.)
10 minutes after that she hollers from the other room. "Nevermind. It's packing material that isn't needed after the washing machine is installed." She'd found Reddit post where someone had a similar experience with the same exact kind of part falling out of their washer of a very similar model. Mind you, my washing machine is decades old, so I'm a little surprised it would still have parts that were only for packing/shipping, but whatever. Not really implausible at all.
Fast forward to today. She's telling me how she told her friend about the mystery part, and she let slip that "when she checked the references" she found on Reddit that it was a useless piece contrary to what it said first.
"References."
Random forum posts don't have "references." Google's AI overview pretends it can give you "references" telling you where it got the "knowledge" it used to generate an answer. (But it's bad about misinterpreting references, and I have to imagine when it provides references, those are complete guesses. Everything I've heard is that "knowing" where it got the information that went into whatever answer it generated just isn't how LLMs work.)
I brought up that she'd told me the first thing she saw that said it was super important wasn't an AI overview. And she admitted it had been.
And, honestly, I find it hard to describe how absolutely livid I am about it. I've asked her many times not to parrot AI overviews to me. (Whatever she wants to put in her brain is no business of mine, but I don't want her polluting my brain with that bullshit.) She regularly ignores my wishes and continues to read me AI overviews. It pisses me off, but I've learned to swallow it.
But I can't express just how huge a betrayal it feels like to be told AI bullshit and then lied to about whether it was AI bullshit. I'm a lot calmer about it than I was at first, but an hour ago, I was shaking.
I was very calm with her about it, though I'm sure my face was super red. I asked her pointedly not to lie to me about where she gets her information, but I don't expect any change in her behavior moving forward. π
