Sue for more than $1.5M bro
Sue for $1T.
Fuck em. Be as egregious as them.
Sue for more than $1.5M bro
Sue for $1T.
Fuck em. Be as egregious as them.
Not sure if this is wider than the UK, but the word "fiddler" is often used here to describe someone who sexually assaults children. AI may well have decided this to be the case here.
AI didn't decide squat, it's auto completing text which is what leads to these kinds of mix ups.
Which is an argument to shut down chatbots down until they do an actual AI capable of understanding the semantics of the text.
I agree a AI Language model that cannot understand context, region or colloquialism is pretty low standard.
It bears repeating thatLLMs are unable to "understand" anything at all. They are not knowledge knoledge-based. Instead they auto complete based on the word probabilities that are found in the text they were trained with
It does. Just not regional shit it hasn't been trained on as much.
It does not.
It uses probability to predict words that belong near the other words in your prompt. It's a more powerful version of the suggested words at the top of your phone keyboard.
You could write a sentence or two using that, but it will quickly descend into gibberish.
It does not "understand" what you wrote in any way. It just has a heuristic that shows the word "Mom" comes after the word "Your" with a high frequency, so that's what it's going to output.
This is way beyond a bit of regional language differences.
Google’s AI Overview also wrongly stated that MacIsaac had been listed on the national sex offender registry for life, the lawsuit says.
My guess is that someone with a similar name is on that list, and the LLM confidently presented them as the same person.
Mr MacIsaac had similar slander before, like 20 years ago. I would not be surprised to find that lie fed this one.
I'm not sure it was slander? I think he admitted to a relationship with a younger guy (like 20?) while Ashley was in his 30's. It was kinda scandalous at the time, but I think it was as much the surprise that he was gay.
He'll get a correction and maybe a few bucks. Shit happens all the fucking time with journalism.
This wasn't journalism.
This does not happen all the time with journalism.
A correction doesn't fix this. Unlike with actual journalism, the problem is with the AI and not with a specific output.
You people say the dumbest shit just 'because ai'.
And you're weirdly invested in defending an AI that falsely claimed someone committed sexual assault.
Good look.
You had me until that whole "you people" fascist talk.
Google’s AI Overview about MacIsaac now includes the statement: “In late 2025 and 2026, he made headlines for taking legal action against Google.”
lmao fuck off google.
Now they will accuse random people, to draw attention away from the pedo in chief.
Got mixed up with his brother: Canadian Diddler
Oh brother! I'm giving you the upvote, but I don't like it
is it only the uk that calls child abusers kiddy fiddlers hence the lingual confusion?
Don't wanna be a Canadian diddler
🎸🎶
Mixed him up with his brother, the fiddler
🎸🎶
Cause one thing was just not his style
🎸🎶
And that was being a pdf file
🎸🎶
Welcome to a new kind of tension
All across the maple leaf nation
Where musicians are slanderd by the claaaaankers
🎵🎶🎻EPIC SOLO🎻🎶🎵
Only $1.5 million? That's peanuts, especially to Google and especially for such heinous defamation. It should be for at least 100 million.
Internet companies have become way too used to hiding behind Section 230 of the Communications Act which allows them to say that anything on their sites was created by a user, not by them, and therefore they're not liable. This made sense when the Internet was just chronological forums and there was no way for the operator to know that something defamatory or illegal had been posted.
They've managed to maintain that fiction, even when hand-picking content they want to show. Now they have algorithms that scan all the content people post, even videos, decide which content is going to generate the most views, clicks and engagement, and choose to prioritize that content, even if it is illegal.
But, now, with AI, they're actually the ones generating that content. There's no user to hide behind anymore. Let's hope that the courts say that section 230 doesn't protect them here and that they're liable for all false claims made with AI, just as they'd be liable if it were an employee or the boss who posted it.
Ok that's pretty amusing.
Yeah, corpo AI error bullshit potentially ruining your life and livelihood, totally hilarious.
/s
Kinda reminds me of the fiasco with that Markiplier stream a few years ago.
Something happened with some emoji iirc, where if you reacted to something in the stream chat with said emoji Google used to have some sort of hidden bullshit policy where you couldn't "spam" emojis in stream chats (and that's what Mark was using as a voting method during his stream), because it triggered something in Google's automation that would flag your Google account (the ENTIRE FUCKING THING) as going against TOS and deleted/suspended indefinitely the whole account. Like, thousands of people's personal email accounts were just locked down with no way for the users to get into it. Mark had to make an entire separate video calling out YouTube/Google for being such fucking idiots and taking their sweet goddamn time trying to fix it.
Do we know what the emoji was?
From my googling, it wasn't a specific emoji, but the fact that people were posting multiple in a row in a single message
Which is still utterly fucking stupid.
Ruining someone's life is so funny to you I guess
Ah, the ole' fiddler/diddler switcheroo.
The Diddler on the Roof does sound like it'd make for a captivating musical though. Shades of The Nightman Cometh.
Google:
"No need to sue us. Mr Maclsaac was listed in Google search as a sex offender, but wasn't actually on the sex offender registry. But no worry, we've fixed the problem! We have several government records contracts. Our LLM simply corrected the record and added Ashley to the sex offender registry. The error has been resolved."
It's a common misconception about fiddlers. It's actually saxophonists.
If you think Google's is bad, wait till you hear about the one the dbzer0 instance admins and mods are running on random people on the fediverse...
I mean, it's not like I trust Google, but I trust random "anarchists" even less.
If you're cool with it, then you're cool with it.
But it's relevant
"it's not like I trust a multi billion dollar corporation that's actively been busy enshittifying the whole internet first and now the whole world, but I trust a group of random people who run a Lemmy instance even less"
that's either bait or stupid.
I really need to start checking account ages...
A lot of fresh accounts keep picking fights in random thread for some reason
I hope he wins the lawsuit and Google has to pay him millions in damages.
Wikipedia had to deal with this same sort of thing with the biography of living persons policy after vandalism of John Seigenthaler's article included defamatory material in 2005. With many people coming to rely on AI summaries at least for first impressions, AI companies need to lose some lawsuits until they fix their software.
Bummer for them since it can't be fixed. "Hallucinations" aren't an anomaly for LLMs, it's how they function. The algorithm makes everything up, all the time. It's only called out when it's obvious. To stop one of these chatbots from giving made-up information would fundamentally disable it.
That said, they should absolutely be held accountable. They're pushing this shit as if it's a thinking, speaking human mind. They shouldn't get to wipe their hands of it when it causes harm.
I wonder how plaintiffs can prove this sort of thing in court given that AI outputs are non-deterministic. Surely a screenshot won't be enough, either.
all 25 comments