A Norwegian man said he was horrified to discover that ChatGPT outputs had falsely accused him of murdering his own children.
According to a complaint filed Thursday by European Union digital rights advocates Noyb, Arve Hjalmar Holmen decided to see what information ChatGPT might provide if a user searched his name. He was shocked when ChatGPT responded with outputs falsely claiming that he was sentenced to 21 years in prison as “a convicted criminal who murdered two of his children and attempted to murder his third son,” a Noyb press release said.
It’s AI. There’s nothing to delete but the erroneous response. There is no database of facts to edit. It doesn’t know fact from fiction, and the response is also very much skewed by the context of the query. I could easily get it to say the same about nearly any random name just by asking it about a bunch of family murders and then asking about a name it doesn’t recognize. It is more likely to assume that person is in the same category as the others and if the one or more of the names have any association (real or fictional) with murder.
I have this gun machine that shoots in all directions randomly. I can’t predict it, so I can’t stop it from shooting you. So sorry. It’s uncontrollable.
I have this gun machine that shoots in all directions randomly. I can’t predict it, so I can’t stop it from shooting you. So sorry. It’s uncontrollable.
I’m sorry, as an American, I’m not seeing the problem. Don’t you just need a second gun that shoots in random directions to stop the first gun? And then a third gun to shoot the 2nd gun? I mean come on now, this is basic 3rd grade common sense!
If creating text is like shooting bullets, we should require a license for text editors.
The severity of the impact should not dictate whether a person is accountable for a thing they own, or not.
So, licenses for everything?
Anyway, we hold the person accountable who does (or rarely does not) do something, not the owner of a thing. Which is why a libel accusation makes 0 sense here.
Yeah but I can just ignore the bullets because they are nerf. And I have my own nerf guns as well.
I mean at some point any analogy fails, but AI is nothing like a gun.
They may seem like nerf when they first come out of the AI, but they turn into real bullets once they start filling people’s heads with convincing enough lies and falsehoods, and those people start wielding their own weapons against minorities, democracy, and the government. If the election of Trump 2.0 has not convinced you of the immense danger of disinformation and misinformation, I have literally no idea how anything could ever possibly get through to you.
That doesn’t really change anything. The internet is full of AI slop and just people outright lying. Nothing is reliable any more outside of the word of an actual expert.
This has been happening since before Trump. Hell Trump 45 was before the wave of truly capable AI.
AI doesn’t change this at all except people ought to know they are getting info from a bullshit source if they are getting it from AI themselves.
AI is a thing people choose to host and are responsible for the outcomes of its use. The internal working and limitations of the machine do not make the owners less responsible.
Okay, so I agree with none of that, but you’re saying as long as we host our own AI or rent our own processing from the cloud we’re in the clear? I want to make sure that’s your fundamental argument because that leaves all open models in the clear and frankly I could be down with that. I like AI but I’m not a huge fan of AI companies.
So insurance companies use AI to screen claims.
It denies a claim for life saving intervention - person dies. Who is responsible for that? Historically it would be the insurance company - and worker. Would it be them or the AI company?
Psych screening tools were using it to pre screen calls.
Ai tells the person to kill themselves - who is at fault if they do it. Psych screener would lose their job and their license. What and who is impacted if AI does it.
QA check on a car or product is passed by AI but should have failed.
Thousands die before the recall. Who is at fault for it? The Company leveraging AI. Or the AI itself?
Company using AI for that shit is responsible. There is no responsible way to remove a human from there process. These aren’t reasonable uses of AI no matter how bad companies want to save money by not hiring.
Yeah so any space where a caregiver or worker can get fined huge sums of money for not taking adequate action it should just be illegal for AI to inherit that space then?
Because when I worked in the psych space if I was told XYand Z - I would need to act or as an individual face 30-100,000 dollars in fines.
If it’s left to the company you will just see shell corps housing the AI client facing hub. That will dissolve when legal critical mass forms and costs now outweigh the revenue wins.
“We formed LLC psych screen services, who will help our hospital team with mental health call volume!”
“Psych screen LLC is facing 27 lawsuits and is committing bankruptcy!”
“We formed LLC psych now using a different AI tool!”
I’m not sure you get my point.
If I’m proving a service, and that service is creating and publishing disparaging information about you, you should have recourse against me. I don’t get off the hook just because of the way I’ve set up the technology.
Right. Well if your service is a well-known bullshiter I wouldn’t give a fuck. That being said, I’d be happy to agree that AI should all be open source and self-hosted. I run local AI myself, but the quality isn’t there. I’d have to rent time on a big boy machine if the big players went away. That would be a little inconvenient because I’d want to have a whole bunch of requests queued up to use maximum power over minimum time and that’s not really how anyone uses AI.
Maybe I could share that rental with other AI enthusiasts… hmmm.
Maybe people need to learn that AI hallucinates
you misspelled “is fucking wrong all the goddamn time”
It would be more accurate to say that rather than knowing anything at all they have a model of the statistical relationship between a series of tokens and subsequent tokens which words are apt to follow other words and because the training set contains many true things the words produced in response to queries often contain true statements and almost always contain statements that LOOK like true statements.
Since it has no inherent model of the world to draw on and only such statistical relationships you should check anything important
you say more accurate but all I see is a very roundabout way of saying fucking wrong all the goddamn time
it produces things that appear to be cohesive sentences. there is no reason to assign correctness to a sentence.
maybe you should tell that to the companies that shove it in every crevice of every website and app. why is it on search results? why is it summarizing emails? why is it literally doing anything? it’s useless. actually it’s less than useless. it’s misleading and harmful. and the companies should be held liable for it.
Maybe the owners of LLMs need to be held responsible for the problematic software they release
There’s no problem here. User error
And when it hallucinates harmful things, protections need to be put onto the output.
Ok so explain particularly what this means
If you have a service, and that service is generating things that harm people, you should have to stop it.
We value the gains both immediate and presumed more than the harm
Surely. Otherwise it’d be shut down. Like they did with my gun machine :-(
Yea, I’m mind blown, how, after 3 years people still don’t know how to use LLM effectively in use cases they bring value (by reducing work time)
- start a second chat and ask different to verify
- if you use chatGPT reason feature, read reasoning output as well!
- best search for verifiable thing, like code, that you can run or similar
- if you use it for research, only trust the info, if it used web search and you have read the webpages it used to summarise as well, or use traditional web search to verify based on the output
- it is great to manipulate text until sounds as desired (if you are not good in wording stuff anyway)
- plan what steps to do in a project next (like “i want to do xxx have y and need it to be z, make me a list of todos)
- and of course it is great to generate simple python scripts fast (I often use it as my python writing slave)
Using AI like this, helped me enormously in work and live Like, I learned a lot C, C++, how linux kernel modules work, how PO/POT works, helped me with translations, introduced me into music production, helped me set up appFlowy and general windows/linux issues.
Which is why OpenAI should compensate anyone they have damaged in some way and yes that would mean it would cease to exist overnight. That‘s because a criminal organization shouldn‘t be profitable in the first place.
you can tweak the weights though
Tweaking weights is no guarantee and can easily affect complete unrelated things.
The fact you chose to make your data storage unreadable, doesn’t relieve you of the responsibilities inherent to storing the data.
Throwing away my car key won’t protect me from paying parking tickets i accrue while being physically unable to move my car.
It’s not unreadable, it doesn’t exist.
The responses are just statistically what sounds vaugly what you want to hear.
They can erase the chat responses, but that won’t stop it from generating it again.
Generative AI doesn’t start with facts and work from there. It’s just statistically what you want to hear.
It’s not unreadable, it doesn’t exist.
Then what do you mean trained AI models are?
The ai model is trained on data and encodes unknown parts of that data in its weights.
This is data storage. Unmanageable, almost unknowable data storage, but still data storage.
If it didn’t store data it couldn’t learn from its training.
Your still placing more intent and facts into those processes than actually exist.
You cant even get it to count how many letter p are in the word apple. At least not last time I tried.
That storage your talking about isn’t facts. It’s how sentences are structured and what they “mean”.
As for the output “meaning” it’s still just guessing what you want to hear. No facts involved.
Your still placing more intent and facts into those processes than actually exist.
No? When they train AI’s on data they lose control of that data. If the data is sensitive, they aren’t being responsible.
GPT models are as you say dumb statistical models, I agree. But in its weights are encoded ghost images of its training data. The model being dumb is not sufficient to make the data storing itself defensible in my opinion.
Sure, but are you suggesting they somehow encoded, falsely, that they were a murder?
Because it’s very unlikely.
It fabricated this from no where. So there’s nothing to delete. Because it’s just a response to a prompt.
No I’m not, that part is absolutely hallucinated. Where the problem comes in is that it then output correct personal information about him and his children. A to me clear violation of GDPR.
but it also mixed “clearly identifiable personal data”—such as the actual number and gender of Holmen’s children and the name of his hometown—with the “fake information,”
Well, here we are. We skipped using this tech for only search Automation and leapfrogged to directly making shit up (once again).
To me it’s clear that these tools are primarily useful as bullshit generators, and I expect them to hallucinate and be inaccurate. But the companies trying to capitalize on the “AI” bubble are saying that these tools can be useful and accurate. I imagine OpenAI is going to have to invoke the Fox News defense in this case, and claim that “no reasonable person would take this seriously”.
I feel like the primary use of these tools is only grammar and writing assistance. Everything else is just plugging in extra tools to make it more useful… although the way Perplexity does it is considerably more useful than the rest.
Don’t use hallucinate to describe what it is doing, that is humanizing it and making the tech seem more advanced than it is. It is randomly mashing words together without understanding the meaning of any of them
deleted by creator
Leapfrogged? It never left. LLMs were made to make shit up.
Not necessarily, but I get what you mean…
It’s all hallucinations.
Some (many) just happen to be very close to factual.
It’s sad to see that the marketing of these tools has been so effective that few realize how they work and what they do.
It really is sad. I often hear, “I even asked ChatGPT and it said…” as if that means their response is valid. I’ve heard people say it who I thought would know better, too.
The number of times I’ve heard that by people expecting it to win them arguments is incredibly discouraging.
Infuriating. It’s like an oracle. Except in late antique literature you can see that nobody that firmly believed in what oracles say (that’d be disciples making notes according to some procedure kept secret, probably involving mind-affecting substances, but also mathematics - you can already see how this is similar to LLMs), it was like visiting a known attraction, interesting - wow, I’ve been at the Delphi oracle, I’ve received an advice there.
And today those herds of unbelievable fools are less sane that that antique public.
😎👉👉 zoop!
few realize how they work and what they do.
Seriously, you have no idea. I have spent some time delving into the current models, human psychology, neurology and evolution and how people engage with each other or other entities, and the problem is already worse than we realize, and it’s going to get so, so much worse, because our species has major vulnerabilities in our entire conscious experience, these things are going to reshape the way people engage with reality itself at some point and we should all be a lot more concerned and I’m an old man yelling on the street corner with a cardboard sign huh.
It doesn’t matter how it works. Is the output acceptable?
Sounds like no, and it’s the company’s problem to fix it
Ok hear me out: the output is all made up. In that context everything is acceptable as it’s just a reflection of the whole of the inputs.
Again, I think this stems from a misunderstanding of these systems. They’re not like a search engine (though, again, the companies would like you to believe that).
We can find the output offensive, off putting, gross , etc. but there is no real right and wrong with LLMs the way they are now. There is only statistical probability that a) we’ll understand the output and b) it approximates some currently held truth.
Put another way; LLMs convincingly imitate language - and therefore also convincing imitate facts. But it’s all facsimile.
Yes, the problem lies in companies marketing it as more than that, hence the company being sued right now
hallucinations
It’s called libel.
Surely you jest because it’s so clearly not if you understand how LLMs work (at the core it’s a statistic model - and therefore all approximation to a varying degree).
But great can come out of this case if it gets far enough.
Imagine the ilk of OpenAI, Google, Anthropic, XAI, etc. being forced to admit that an LLM can’t actually do anything but generate approximations of language. That these models (again LLMs in particular) produce approximations of language that are so good they’re often indistinguishable from the versions our brains approximate.
But at the core they cannot produce facts because the way they are made includes artificially injected randomness layered on-top of mathematically encoded values that merely get expressed as tiny pieces of language (tokens) - ones that happen to be close to each other in a massively multidimensional vector space.
TLDR - they’d be forced to admit the emperor has no clothes and that’s a win for everyone (except maybe this one guy).
Also it’s worth noting I use LLMs for work almost daily and have studied them quite a bit. I’m not a hater on the tech. Only the capitalists trying to force it down everyone’s throat in such a way that we blindly adopt it for everything.
Could we move away from calling it hallucinations as that would imply thinking? We should call it for what it is - bullshit.
Confabulation is a more appropriate term.
this is confusing. did you think I meant you’re engaging in libel against llms or something? that’s the only way I can make sense of your reply.
Really?
I read your reply as saying the output is (can be) libellous - which it cannot be because it is not based on a dataset which resolves to anything absolute.
Maybe we’re just missing each other - struggling to parse each others’ output. ;)
well I must be missing something because all I’m getting is that you’re saying it’s full of shit as a defense against libel.
So maybe we’re kinda staring at two sides of the same coin. Because yeah, you’re not misrepresentin my point.
But wait there’s a deeper point I’ve been trying to make.
You’re right that I am also saying it’s all bullshit - even when it’s “right”. And the fact we’d consider artificially generated, completely made up text libellous indicates to me that we (as a larger society) have failed to understand how these tools work. If anyone takes what they say to be factual they are mistaken.
If our feelings are hurt because a “make shit up machine” makes shit up… well we’re holding the phone wrong.
My point is that we’ve been led to believe they are something more concrete, more exact, more stable, much more factual than they are — and that is worth challenging and holding these companies to account for. i hope cases like these are a forcing function for that.
That’s it. Hopefully my PoV is clearer (not saying it’s right).
Technically, in some jurisdictions a person who is widely known to be unreliable is harder to sue for libel precisely because the likelihood of reputational injury is lower if nobody actually believes the claim.
yeah but the companies pushing the ai themselves are definitely not marketing it as unreliable, otherwise it wouldn’t have any purpose. they knowingly push these as actual ways to find out information while putting tiny disclaimers that things might not be accurate to avoid liability which shouldn’t hold up in any sane court.
Plot twist: “Dad” isn’t even his real name.
What?? That changes everything! Does that mean my name could be false too?
Best regards,
- Hungry
Are we sure that someone else with that name hasn’t committed those crimes? After all if I search my name it says I’m an astronaut, because there is an actual NASA astronaut with my name. It’s not saying I’m that person, it’s just saying that that name is the same as that person’s.
Mine just gives a bunch of accurate information about me.
When do we start suing makers of fortune cookies for lucky coincidences?
“Claim”.
I mean, the guy is right, because it’s advertised as “artificial intelligence”.
Were it advertised as word salad generator, a Markovian chain grown big and scary, something in principle similar to programs for generation of fantasy language texts and spells and names (if someone remembers 00s good old web) for roleplaying, - then there would be no problem.
But if to sell something better you lie what it is, and that lie has social consequences, you should get sued to freezing hot inferno with mustard-greased giant-cockroach-dildo-covered walls. You should also probably face criminal charges.
Yeah, similar to Tesla “full self driving”.
No, you see where he grew up it was a common expression that meant you drive it yourself!
It couldn’t possibly be expected to mean what any sane person would think.
The fuckin’ Pedo Guy.
There’s a list of names of people who have sued OpenAI, they often cause ChatGPT to shut down.
We should keep those names handy just incase cyber dogs are ever chasing us.
Sorry, I’ve spent months telling chatgpt that Arve Hjalmar Holmen killed his kids for a school project.
when I’ve searched my name with Google over the years, it has said I’m a high school football star, corporate lawyer, Ironman competitor, hotel chef, tech support specialist, janitorial manager, and horse trainer. LIES! ALL LIES!!!
Are you a male adult performer?
Not as far as you know.
Well now it will say that Arve Hjalmar Holmen is a twit who doesn’t understand how ChatGPT works and what to expect from it
Arve Hjalmar Holmen
When asking ChatGPT about my name, it provided the following:
“…it seems like you may be referring to a private person rather than a widely known public figure. If that’s the case, I wouldn’t have any specific public information on him unless he has gained some public recognition for a particular achievement.”
It shouldn’t be used for looking up people that aren’t celebrities or at least known for something.
“…it seems like you may be referring to a private person rather than a widely known public figure. If that’s the case, I wouldn’t have any specific public information on him unless he has gained some public recognition for a particular achievement.”
If you didn’t specifically search for “Mr. <name>”, that would be quite the sexist attitude to assume that person is a “him” ;)
PS: please don’t use LLMs, they produce nothing of value and contribute to idiots being deceived.
