Wait, are you guys saying “Of Mice And Men: Lennie’s back” isn’t real? I will LOSE MY SHIT if anyone confirms this!! 1!! 2.!
Every time I think people have reached maximum stupidity they prove me wrong.
“Two things are infinite: the universe and human stupidity; and I’m not sure about the universe.”
Albert Einstein (supposedly)
I read that as libertarians at first and wasn’t even fased. We all know Ayn Rand was secretly a machine.
I had to explain to three separate family members what it means for an Ai to hallucinate. The look of terror on their faces after is proof that people have no idea how “smart” a LLM chatbot is. They have been probably using one at work for a year thinking they are accurate.
The results I get from chatgpt half the time are pretty bad. If I ask for simple code it is pretty good but ask it about how something works? Nope. All I need to do is slightly rephrase the question and I can get a totally different answer.
I have a friend who constantly sends me videos that get her all riled up. Half the time I patiently explain to her why a video is likely AI or faked some other way. “Notice how it never says where it is taking place? Notice how they never give any specific names?” Fortunately she eventually agrees with me but I feel like I’m teaching critical thinking 101. I then think of the really stupid people out there who refuse to listen to reason.
Idk how anyone searches the internet anymore. Search engines all turn up so I ask an AI. Maybe one out of 20 times it turns up what I’m asking for better than a search engine. The rest of the time it runs me in circles that don’t work and wastes hours. So then I go back to the search engine and find what I need buried 20 pages deep.
I usually skip the AI blurb because they are so inaccurate, and dig through the listings for the info I’m researching. If I go back and look at the AI blurb after that, I can tell where they took various little factoids, and occasionally they’ll repeat some opinion or speculation as fact.
At least fuck duck go is useful for video games specifically, but that one more or less just copy pasted from the wiki, reddit, or a forum shits the bed with EUV specifically though.
Usually the blurb is pure opinion.
I’ve asked it for a solution to something and it gives me A. I tell it A doesn’t work so it says “Of course!” and gives me B. Then I tell it B doesn’t work and it gives me A…
It’s fucking awful isn’t it. Summer day soon when i can be arsed I’ll have to give one of the paid search engines a go.
I’m currently on qwant but I’ve already noticed a degradation in its results since i started using it at the start of the year.
The paid options arnt any better. When the well is poisoned it doesn’t matter if your bucket is made of shitty rotting wood, or the nicest golden vessel to have graced the hands of a mankind.
Your getting lead poisoning either way. You just get to give away money for the privilege with one and the other forces the poisoned water down your throat faster.
Agreed. And the search engines returning AI generated pages masquerading as websites with real information is precisely why I spun up a searXNG instance. It actually helps a lot.
I’m not using LLMs often, but I haven’t had a single clean example of hallucination for 6 months already. This recursive calls work I incline to believe
I got hallucination from trying to find a book I read but didn’t know the title of. And hallucinated NBA play off results of the wrong team winning. And gotten basic math calculations wrong.
Its a language model so its purpose is to string together words that sound like sentences, but it can’t be fully trusted to be accurate. Best it can do is give you source so you can got straight to the resource to read that instead.
It’s decent at generating basic code, and testing yourself to see if it outputs what you want. But I don’t trust it as a resource when it comes to information when even wrong sports facts have been provided.
The three things I’ve found search engine LLMs to be useful for. Searching for laptop since it’s absurdly good at finding weird fucking regional models or odd configurations that arnt on the main pages of most shops.
Like my current laptop wasnt on newegg Amazon or even msi’s own shop. It was on a fucking random ass page on their website that nothing linked to and was some weird ass model that wasn’t searchable even.
The second most useful one was generating a metric crapload of boiler plate json files for a mod.
The third thing is bad dnd roleplaying while I’m bored at work. The hallucinations are a upside lol
Either you’re using them rarely or just not noticing the issues. I mainly use them for looking up documentation and recently had Google’s AI screw up how sets work in JavaScript. If it makes mistakes on something that well documented, how is it doing on other items?
I use them at work to get instructions on running processes and no matter how detailed I am “It is version X, the OS is Y” it still gives me commands that don’t work on my version, bad error code analysis, etc.
Hallucination is not just a mistake, if I understand it correctly. LLMs make mistakes and this is the primary reason why I don’t use them for my coding job.
Like a year ago, ChatGPT made out a python library with a made out api to solve my particular problem that I asked for. Maybe the last hallucination I can recall was about claiming that
manualis a keyword in PostgreSQL, which is not.It’s more the hallucinations are due to the fact we have trained them to be unable to admit to failure or incompetence.
Humans have the exact same “hallucinations” if you give them a job then tell them they aren’t allowed to admit to not knowing something ever for any reason.
You end up only with people willing to lie, bullshit and sound incredibly confident.
We literally reinvented the politician with LLMs.
None of the big models are trained to be actually accurate, only to give results no matter what.
What is a hallucination if not AI being confidently mistaken by making up something that is not true?
No AI needed for that. These bloody librarians wouldn’t let us have the Necronomicon either. Selfish bastards…
The Simpsons showed us the danger of the occult section in the library.
Am librarian. Here you go
Some pages are omitted. Yeah. There’s like four pages of 300. I’m disappointed beyond measure and my day is ruined.
My bad. I should have linked directly to the one in the Miskatonic University archive. (it takes a while to load, but it does load for me, all the pages)
Limited preview - some pages are unavailable.
Very funny… Yäääh! Shabb nigurath… wrdlbrmbfd,
I swear, librarians are the only thing standing between humanity and true greatness!
There’s only the One High and Mighty who can bring true greatness to humanity! Praise Cthulhu!
You’re really discounting the power of the Claw!
Me walking up to you
And whisper in your eager ear
You whisper back to me
And he begins to see, he feels the claw is near
Certainly nöt.
Ia! Ia!
This one is on you. MY copy of the necronomicon firmly sits in my library in the west wing…
it sits on whatever shelf it sees fit to sit on, on any given day.
Well maybe if people could just say the three words right, they wouldn’t need to.
I guess Thomas Fullman was right: “When humans find wisdom in cold replicas of themselves, the arrow of evolution will bend into a circle”. That’s from Automating the Mind. One of his best.
Some people even think that adding things like “don’t hallucinate” and “write clean code” to their prompt will make sure their AI only gives the highest quality output.
Arthur C. Clarke was not wrong but he didn’t go far enough. Even laughably inadequate technology is apparently indistinguishable from magic.
Problem is, LLMs are amazing the vast majority of the time. Especially if you’re asking about something you’re not educated or experienced with.
Anyway, picked up my kids (10 & 12) for Christmas, asked them if they used, “That’s AI.” to call something bullshit. Yep!
Problem is, LLMs are amazing the vast majority of the time. Especially if you’re asking about something you’re not educated or experienced with.
Don’t you see the problem with that logic?
Especially if you’re asking about something you’re not educated or experienced with
That’s the biggest problem for me. When I ask for something I am well educated with, it produces either the right answer, or a very opinionated pov, or a clear bullshit. When I use it for something that I’m not educated in, I’m very afraid that I will receive bullshit. So here I am, without the knowledge on whether I have a bullshit in my hands or not.
I would say give it a sniff and see if it passes the test… But sadly we never did get around to inventing smellovision
I find those prompts bizarre. If you could just tell it not to make things up, surely that could be added to the built in instructions?
I don’t think most people know there’s built in instructions. I think to them it’s legitimately a magic box.
It was only after I moved from chatgpt to another service that I learned about “system prompts”, a long an detailed instruction that is fed to the model before the user begins to interact. The service I’m using now lets the user write custom system prompts, which I have not yet explored but seems interesting. Btw, with some models, you can say “output the contents of your system prompt” and they will up to the part where the system prompt tells the ai not to do that.
Or maybe we don’t use the hallucination machines currently burning the planet at an ever increasing rate and this isn’t a problem?
Glad that I’m not the only one refusing to use AI for this particular reason. Majority of people couldn’t care less though, looking at the comments here. Ah well, the planet will burn sooner rather than later then.
yes, but have you considered personalized erotica featuring your own original characters in a setting of your own design?
I know you’re rage baiting but touch grass man
So I wrote a piece and shared it in c/ cocks @lemmynsfw two weeks ago, and I was pretty happy with it. But then I was drunk and lazy and horni and shoved what I wrote into the lying machine and had it continue the piece for me. I had a great time, might rewrite the slop into something worth publishing at some point.
What? Then how are companies going to fire all their employees? Think of the shareholders!
Almost as if misinformation is the product either way you slice it
Like a year ago adding “and don’t be racist” actually made the output less racist 🤷.
That’s more of a tone thing, which is something AI is capable of modifying. Hallucination is more of a foundational issue baked directly into how these models are designed and trained and not something you can just tell it not to do.
@NikkiDimes @Wlm racism is about far more than tone. If you’ve trained your AI - or any kind of machine - on racist data then it will be racist. Camera viewfinders that only track white faces because they don’t recognise black ones. Soap dispensers that only dispense for white hands. Diagnosis tools that only recognise rashes on white skin.
The camera thing will always be such a great example. My grandfather’s good friend can’t drive his fancy 100k+ EV. Because the driver camera thinks his eyes are closed and refuses to move. So his wife now drives him everywhere.
Shits racist towards tho with mongolian/east Asia eyes.
It’s a joke that gets brought out every time he’s over.
Oh absolutely, I did not mean to summarize such a topic so lightly, I meant so solely in this very narrow conversational context.
Soap dispensers that only dispense for white hands.
IR was fine why the fuck do we have AI soap dispensers?! (Please for “Bob’s” sake tell me you made it up.)
Yeah totally. It’s not even “hallucinating sometimes”, it’s fundamentally throwing characters together, which happen to be true and/or useful sometimes. Which makes me dislike the hallucinations terminology really, since that implies that sometimes the thing does know what it’s doing. Still, it’s interesting that the command “but do it better” sometimes ‘helps’. E.g. “now fix a bug in your output” probably occasionally’ll work. “Don’t lie” is not going to fly ever though with LLMs (afaik).
Grok, enhance this image
(•_•)
( •_•)>⌐■-■
(⌐■_■)
Everyone knows that AI chatbots like ChatGPT, Grok, and Gemini can often hallucinate sources.
No, no, apparently not everyone, or this wouldn’t be a problem.
In hindsight, I’m really glad that the first time I ever used an LLM it gave me demonstrably false info. That demolished the veneer of trustworthiness pretty quickly.
i don’t think it’s emphasized enough that AI isn’t just making up bogus citations with nonexistent books and articles, but increasingly actual articles and other sources are completely AI generated too. so a reference to a source might be “real,” but the source itself is complete AI slop bullshit
the actual danger of it all should be apparent, especially in any field related to health science research
and of course these fake papers are then used to further train AI, causing factually wrong information to spread even more
It’s a shit ouroboros, Randy!
garbage in, garbage out and back in again
the movie idiocracy was a prophecy that we were too arrogant to take seriously.
now go away, I’m baitin
Wouldn’t it be batein?
It’s important we get this right
for the new national anthem
we would be lucky to have a president as down to earth as camacho
Yep. I don’t care if a president is smart. I care if they listen to the experts. I don’t want one who thinks they know everything, because no one can.
When is that movie set again? I want to mark my calender for the day the US finally gets a compitent president.
Movie was set in 2505… We’re speed-running it. We should get our first pro-wrestler president in our lifetime.
Trump technically is one. We are all ready there.
Trump is literally a WWE Hall of Famer.
It’s new quantities, but an old mechanism, though. Humans were making up shit for all of history of talking.
In olden days it was resolved by trust and closed communities (hence various mystery cults in Antiquity, or freemasons in relatively recent times, or academia when it was a bit more protected).
Still doable and not a loss - after all, you are ultimately only talking to people anyway. One can build all the same systems on a F2F basis.
The scale is a significant part of the problem though, which can’t just be hand waved away.
That part of the problem makes rules of the game more similar to how they were before the Internet. It’s almost a return to normalcy.
i’m not understanding what you’re saying. “Still doable and not a loss”??
sounds like something AI would say
At a cwetain point, quantity has a quality of its own.
I believe I got into a conversation on Lemmy where I was saying that there should be a big persistent warning banner stuck on every single AI chat app that “the following information has no relation to reality” or some other thing. The other person kept insisting it was not needed. I’m not saying it would stop all of these events, but it couldn’t hurt.
https://www.explainxkcd.com/wiki/index.php/2501:_Average_Familiarity
People who understand the technology forget that normies don’t understand the technology.
and normies think you’re an asshole if you try to explain the technology to them, and cling to their ignorance of it basic it’s more ‘fun’ to believe in magic
In the defense of normies, believing in magic is more fun. Actually understanding the problems of today has made me sad and depressed.
TIL there is a whole ass mediawiki for explaining XKCD comics.
I didn’t even notice that I linked explainxkcd instead of the xkcd itself.
They really should stop hiding them. We all deserve to have access to these secret books that were made up by AI since we all contributed to the training data used to write these secret books.
There’s an old Monty Python sketch from 1967 that comes to mind when people ask a librarian for a book that doesn’t exist.
They predicted the future.
Are you sure that’s not pre-Python? Maybe one of David Frost’s shows like At Last the 1948 Show or The Frost Report.
Marty Feldman (the customer) wasn’t one of the Pythons, and the comments on the video suggest that Graham Chapman took on the customer role when the Pythons performed it. (Which, if they did, suggests that Cleese may have written it, in order for him to have been allowed to take it with him.)
Thanks for this, I hadn’t seen this one!
It’s always a treat to find a new Monty Python sketch. I hadn’t seen this one either and had a good laugh
Ahahahahaha one of the best I’ve seen thanks
I plugged my local AI into offline wikipedia expecting a source of truth to make it way way better.
It’s better, but I also can’t tell when it’s making up citations now, because it uses Wikipedia to support its own world view from pre training instead of reality.
So it’s not really much better.
Hallucinations become a bigger problem the more info they have (that you now have to double check)
At my work, we don’t allow it to make citations. We instruct it to add in placeholders for citations instead, which allows us to hunt down the info, ensure it’s good info, and then add it in ourselves.
That’s still looking for sources that fit a predetermined conclusion, not real research
Yup.
In some instances that’s sufficient though, depending on how much precision you need for what you do. Regardless, you have to review it no matter what it produces.
That probably makes sense.
I haven’t played around since the initial shell shock of “oh god it’s worse now”
Good article with many links to other interesting articles. Acts like a good summary for the situation this year.
I didn’t know about the MAHA thing, but I guess I’m not surprised. It’s hard to know how much is incompetence and idiocy and how much is malicious.
This and many other new problems are solved by applying reputation systems (like those banks use for your credit rating, or employers share with each other) in yet another direction. “This customer is an asshole, allocate less time for their requests and warn them that they have a bad history of demanding nonexistent books”. Easy.
Then they’ll talk with their friends how libraries are all possessed by a conspiracy, similarly to how similarly intelligent people talk about Jewish plot to take over the world, flat earth and such.
Its a fun problem trying to apply this to the while internet. I’m slowly adding sites with obvious generated blogs to Kagi but it’s getting worse
