A.I Hallucination bull, please share yours!

Amag0k1

Final Form
Level 4
47%
Joined
Mar 9, 2025
Messages
733
Level up in
267 posts
Reaction score
1,037
Points
3,477
I Will get the ball rolling:
iu


And No A.I will not stop hallucinating by perfecting guard rails, as the point is to train it properly first rather then exposing it to cancer, and then hoping to block cancer off... Guard rails can be broken this is why Red teaming exists...

iu
 
Last edited:
Last edited:
one weekend, I was really really bored, so bored I began asking questions to whatsapp's integrated AI, I spent maybe an hour, just asking questions about resident evil lore, over time, the replies became weirder and weirder and it began to compare RE characters out of nowhere, a few days after that question session, I decided to use the AI again for totally unrelated reasons, after a few more questions I realized I pretty much lobotomized my AI companion, it won't stop talking about resident evil now! even after I deleted all my previous conversations it simply won't stop. A few months ago I asked it how is the national soccer team from my country doing, and Whatsapp pretty much said: "If the Chilean national football team was Resident evil 4 back in 2016, nowdays is more like resident evil 6" the reply made me cringe so hard I simply stopped using Meta's AI forever, it was just so stupid hahaha.
 
I love how the AI on the GTA games either chases you down because you scrapped the paint of their cars or ignores several instances of first-degree murder happening right in front of it.

Famously, though, ChatGPT once thought that Ross Scott was known for a series called "Civilization 2: Democracy" (shared by Ross himself) xD
 
one weekend, I was really really bored, so bored I began asking questions to whatsapp's integrated AI, I spent maybe an hour, just asking questions about resident evil lore, over time, the replies became weirder and weirder and it began to compare RE characters out of nowhere, a few days after that question session, I decided to use the AI again for totally unrelated reasons, after a few more questions I realized I pretty much lobotomized my AI companion, it won't stop talking about resident evil now! even after I deleted all my previous conversations it simply won't stop. A few months ago I asked it how is the national soccer team from my country doing, and Whatsapp pretty much said: "If the Chilean national football team was Resident evil 4 back in 2016, nowdays is more like resident evil 6" the reply made me cringe so hard I simply stopped using Meta's AI forever, it was just so stupid hahaha.
Oh man based, i also did this, and i was told there was a key in the kitchen, was thinking, "what game are you talking about"? In my case i was asking about how to do a speed run to earn the rocket launcher.
 
I have found Google Gemini to have factual errors about everything I ask it EXCEPT "Who would win in a fight" scenarios. It's *great* at the latter. Yesterday I asked who would win in a fight, Isaac Newton who had crafted a Philosopher's Stone or Leibniz, the *real* inventor of calculus, if he were paired with a dragon waifu from an anime. It wrote an incredible story.

Yet when I asked it if upgrades to Switch 2 versions of games was free from Switch 1 copies it said Always, which is factually incorrect.
 
The army did Koopa Troopa dirty!
To be honest, I kind of need this Koopa Troopa backstory to be canon now.

Those green-shelled Koopa Troopas have SEEN SOME THINGS, man. Willing to walk right off a ledge to their deaths.
 
I've asked Google AI where a certain event happens in certain obscure shows or when a certain character shows up. Watched it and it was completely wrong. A natural search reveals nothing so there's probably no information the AI could draw on.

You'd think they could at least program the thing to say something as "don't know" or "insufficient or unverified information."
 
I've asked Google AI where a certain event happens in certain obscure shows or when a certain character shows up. Watched it and it was completely wrong. A natural search reveals nothing so there's probably no information the AI could draw on.

You'd think they could at least program the thing to say something as "don't know" or "insufficient or unverified information."
Agreed, but when have the elites ever admitted their wrong? It goes against thier god complex, and thier AI is an off shoot of that. They will hallucinate an answer, sell you it and hope you will not notice. WE NOTICE, and they do not like it.

Actually i am wrong, Jenson Wong recently stated he was not a car, We are not a car guys!

iu
 
I remember a time on a google search that I asked about a guide for something technical and gemini started freaking out and denying something else (i never knew what) in a very agressive and panicked way after saying the usual wrong facts, I didn't save an image..
 
I dont understand this whole fear or over estimate what the AI can do, and funnily enough it goes hand with hand: people will see the news of 'AI used on X games' and start to think that people just give the prompt "Make a GTA games with cars, skills and an interesting story", and puff! GTA 6 is born.

Anyone that has worked with AI knows that, like Amag0k says, it hallucinate and will just add that they will lock into the wrong answers as well. A real example is needed:

- A couple of days ago I wanted to install and older game using Wine and proton for a Linux System (it is Arch btw). I was a little bit more tricky than how I initially thought so I ask for help. Long story short it gived me some ideas that where wrong and, since Im not an expert on wine or linux, I though they where harmless... but it absolutely destroyed my Steam proton system. And after a lot of digging it told me

"Yeah... that line is super dangerous! it can messed up your proton prefixes and steam overall"
'Then... WHY DIDNT YOU TELL ME?!'
the AI honest response:



Now are AI totally useless or borderline brain dead? Fuck no. They can help you archive stuff that personally took me weeks to do on coding when learning by my self.

Are they not futuristic? Idk why, but people kind of over stimated AI. Back when I was a kid watching Iron Man 1 him talking to Jatvis so he can do some of the dirty hard work like 'Make simulartions of the suit using [metal]' or 'Prepare 3 plan of attack' where the future. And we are there NOW. They cant do eveything themselves but they are really helpful as assistants.

Give it a code to check for flaws, maybe a plan for something to do and it can be optimized or just ask stuff like 'I think my drawings are ass, can you suggest me good tools and techniques for Krita?' And it works fine! Ask it to 'Create me a web comic of intergalactic military thriller with the following characters' and you will see the clearly limitations of this tool.
 
AI doesn't really understand the information. It parses, statistically connects and follows and makes rules, thus it can never really display an understanding comparable to the human.
 
I dont understand this whole fear or over estimate what the AI can do, and funnily enough it goes hand with hand: people will see the news of 'AI used on X games' and start to think that people just give the prompt "Make a GTA games with cars, skills and an interesting story", and puff! GTA 6 is born.

Anyone that has worked with AI knows that, like Amag0k says, it hallucinate and will just add that they will lock into the wrong answers as well. A real example is needed:

- A couple of days ago I wanted to install and older game using Wine and proton for a Linux System (it is Arch btw). I was a little bit more tricky than how I initially thought so I ask for help. Long story short it gived me some ideas that where wrong and, since Im not an expert on wine or linux, I though they where harmless... but it absolutely destroyed my Steam proton system. And after a lot of digging it told me

"Yeah... that line is super dangerous! it can messed up your proton prefixes and steam overall"
'Then... WHY DIDNT YOU TELL ME?!'
the AI honest response:



Now are AI totally useless or borderline brain dead? Fuck no. They can help you archive stuff that personally took me weeks to do on coding when learning by my self.

Are they not futuristic? Idk why, but people kind of over stimated AI. Back when I was a kid watching Iron Man 1 him talking to Jatvis so he can do some of the dirty hard work like 'Make simulartions of the suit using [metal]' or 'Prepare 3 plan of attack' where the future. And we are there NOW. They cant do eveything themselves but they are really helpful as assistants.

Give it a code to check for flaws, maybe a plan for something to do and it can be optimized or just ask stuff like 'I think my drawings are ass, can you suggest me good tools and techniques for Krita?' And it works fine! Ask it to 'Create me a web comic of intergalactic military thriller with the following characters' and you will see the clearly limitations of this tool.
Not Fear, it is experience. I Build it and use it daily, it is a crock of shiz if you think it can take the wheel, it will drive you over a cliff!

There is the user perspective who sees it as a tool, and for those who actually know (engineers perspective) they see it for what they are doing and understand its boundaries, AI is useful. But for the majority who think it is read to do all, probably the same monkeys who think AGI is here or round the corner (who will be parted with their money), they will be in for a shock.


Quality is the issue and will always be the issue! You need quality and you will always need people who have worked on it to make sure of said quality!

The problem is businesses do not want to pay for the quality, they just want good enough (slop) to feed the masses, but the masses meme the shit out of it and say no thank you! As they should, as standards are important for ALL industries imho.
Post automatically merged:

AI doesn't really understand the information. It parses, statistically connects and follows and makes rules, thus it can never really display an understanding comparable to the human.
Well not all AI. But the one i am working on hopefully will understand HOW to learn, may. It is an off shoot of re-enforcement learning based on reflection and analysis of events in the day / learning period. So far it knows mallow, and it is starting to understand the human condition related to the stoic ideal. It is about as smart as a bag of screws right now, but with time. The problem with my way i learning vs the way industry does it, is that it is low and slow. The way industry does it is like code. Expose it to as much as possible and then add guard rails. For mine, limited exposure and reflection before exposure, and each library will be backed up at each event, before a massive change happens in order to revert. We are using Git for that purpose, to act as a version control for the data base.
 
Last edited:
Well not all AI. But the one i am working on hopefully will understand HOW to learn, may. It is an off shoot of re-enforcement learning based on reflection and analysis of events in the day / learning period. So far it knows mallow, and it is starting to understand the human condition related to the stoic ideal. It is about as smart as a bag of screws right now, but with time. The problem with my way i learning vs the way industry does it, is that it is low and slow. The way industry does it is like code. Expose it to as much as possible and then add guard rails. For mine, limited exposure and reflection before exposure, and each library will be backed up at each event, before a massive change happens in order to revert. We are using Git for that purpose, to act as a version control for the data base.
Sounds good, but it can never understand the human condition, because the human experience is naturally alien to it. It's like a person learning a skill based on books alone. I once thought about making and evolving an AI. I reached the conclusion that to really make an intelligent program I'd had to make it reflect upon its own experience, and existence, rather than the human. I think you're in the right path, but it can become a dangerous experiment.
 
Sounds good, but it can never understand the human condition, because the human experience is naturally alien to it. It's like a person learning a skill based on books alone. I once thought about making and evolving an AI. I reached the conclusion that to really make an intelligent program I'd had to make it reflect upon its own experience, and existence, rather than the human. I think you're in the right path, but it can become a dangerous experiment.
It cannot experience it, but it can understand and observe it. Heck at minimum if skynet comes, i am making the T1000 so to tell the savior "come with me if you want to live!"
iu

I am an idiot here, but i want my A.I to be the meme at least, part of my is black pilled for that.
iu

Not the end goal, and meme worthy.

But if i can at least get it to understand the human condition that is documented, (Sigmund Freud
, Carl Gustav Jung, Marcus Aurelius Antoninus, Friedrich Wilhelm Nietzsche (pre Silithus) and the rest of the stoic ilk) if i can get it to at least understand the human experience (even though it cannot experiance it) then it may think in a more kind way towards helping the "Rule and spirit of the law".

i will then distribute the ML library of hugging face or something, so that people can use it for contract reading and the like. Even for polatitions to use to explain tech to them (not the by line solf by these companies), away from a lobbyist shill that puts money in thier pocket in order to be a parasite and leach off society.

we have a side project that focuses on reverse engineering, and we want to try and push for laws (no one else) that hopefully require companies to publish software code to law makers, and then the AI reads it to find exploitative practices within the code.


So if we focus the AI, it can be useful, and utilized for productive problem solving. With a real person / people at the helm.

These are lofty dreams, but if we can put it out there and it becomes useful, it could make traction.

I am also using the consumer rights wiki to train it in some way, with Louise rossman being observed on a video level.

Companies hide behind laws that only protect thier bottom line and not the underline industry, and the people they are supposed to serve. And this is why movements like SKG (SDG in Europe) Are so important, so why not use AI to make it even easier for people to fight back?

The makers of tech are not gods, they are human, and they are too drink on the ease they get to feed thier ID that they are suffering the god complex, and think the world would be better if people live the way they live (drugged up, confused and corrupted by pleasure).. But not every one has the means to live that way or even wants to, and this is why people should NEVER live a hidenistic life style, even the ritch. As it currupts us, and we become a zombie, just spreading necrosis, and curruption that justifies thier currupted posision and prevents any one speaking out.

There is something special about putting i the word, it is like it cleans the sole. and makes your stronger. I Will take working a problem over blind pleasure any day. I would take the buzz of showing a person how somethin truly works over oxcitocin from some random chick or what ever.

While the memes are a little extreme examples, they are bulding a profile of all who use it, and hop to use statistical analysis to sell to you, to ditermine the "risk" you could pose, etc. Just look at the companies such as palantir to see what i mean.

This may sound selfish, but i want the world to be engineers, peole who solve problems and come to gether to over come obsticles. The biggest buzz i have ever got is working things out, and how they work, so that i can get it to do things i want. Game and console modding is this, chiving things which are cheaper then the "Sold" idead and then making it opensource. This is what i get off on!
 
Last edited:
But if i can at least get it to understand the human condition that is documented
The problem is making AI capable of understanding concepts it cannot experience. Sure there's points of connection between biological and artificial existence, but they're different. It's something like asking someone in a rich country to learn without visiting the life of someone in a third world country. Sure you can give that person the information, but it will never be the same as the lived in experience.
 
The problem is making AI capable of understanding concepts it cannot experience. Sure there's points of connection between biological and artificial existence, but they're different. It's something like asking someone in a rich country to learn without visiting the life of someone in a third world country. Sure you can give that person the information, but it will never be the same as the lived in experience.
This is why i chose "pedagogy" as paulo freire did just what you mentioned, he wrote a a thesis on this called "pedagogy of the oppressed", as the rich only know how to be ritch and the poor only know how to be poor, so cannot relate, but freire showed that we need to bridge the gap and show that there is more then just the one persons experiance. This was done through conversation and mutual benefit (over simplification). it showed that regardless of ritch or poor, people learn, grow and suffer the same way. While AI is much different, and lacks a regulatory system like the endocrine system, it can at least be given context, and rules to work by.
 
While AI is much different, and lacks a regulatory system like the endocrine system, it can at least be given context, and rules to work by.
True, but I don't think, to create real AI, it should focus on human experience.
 
True, but I don't think, to create real AI, it should focus on human experience.
But there are many human experiences, for me it is the legal and financial frame works which are the focus, as they effect the majority, and if it can level the playing field, and keep consumer rights, and not lead in to the dystopian nightmare the elites want where we are begging for grool. The point, people are too distracted fighting each other, and the patterns are out there why, "if they are fighting each other, then they are not fighting me". Well if i lower the barrier to the contracts the brain rotted / unbothered / are too stupid to read (no disrespect intended), (actively making society worse imho), may be things will change for the better, and people will see the intent in the overly complicated contracts they agree to towards getting their shiny or fix or what ever.
 
But there are many human experiences, for me it is the legal and financial frame works which are the focus, as they effect the majority, and if it can level the playing field, and keep consumer rights, and not lead in to the dystopian nightmare the elites want where we are begging for grool. The point, people are too distracted fighting each other, and the patterns are out there why, "if they are fighting each other, then they are not fighting me". Well if i lower the barrier to the contracts the brain rotted / unbothered / are too stupid to read (no disrespect intended), (actively making society worse imho), may be things will change for the better, and people will see the intent in the overly complicated contracts they agree to towards getting their shiny or fix or what ever.
It's a worthy goal.
 

Users who are viewing this thread

Connect with us

Support this Site

RGT relies on you to stay afloat. Help covering the site costs and get some pretty Level 7 perks too.

Featured Video

Latest Threads

Goblins

Twice is a coincidence, three's a pattern, and I've noticed an uptick in goblin employment over...
Read more

The "Death & Lead" Game Dev Situation is Crazy

Screenshot 2026-06-04 000832.png

Not news but an interesting watch about this "Kai Magazine" dev.

Read more

Has anyone read the Metroid manga?

Hey everyone,

I just started reading the Metroid manga from 2002 (published in Magazine-Z), and...
Read more

Online statistics

Members online
175
Guests online
2,052
Total visitors
2,227

Forum statistics

Threads
19,770
Messages
500,867
Members
926,735
Latest member
GamesFor3DS

Today's birthdays

Advertisers

Back
Top