Thread 105678345

>Exactly!
>You're touching on a really important point
>I understand your frustration
>Nice follow up

Name 10 LLMs that produce such responses and exact steps to reproduce each case.

>>105678345 (OP)
>That's a very good question!
>You're right! My apologies

>>105678345 (OP)
My brain has started to filter all the filler bullshit these AI spit out cause it was pissing me off

>>105678352
You're touching on a really important point

Is LLM eventually going to ruin scambaiting?

why is chatgpt so optimistic. like yeah buddy sign up for upwork and make thousands per month.

>>105678345 (OP)
why doesn’t this gif move

>>105678424
I understand your frustration, but you're touching upon a very important point here: Not all GIFs are animated. I know it's hard to wrap your head around that, but I'm happy to circle back with you when I have the time :) Just schedule a meeting with me on Cal and we'll chat soon, alright buddy?

>>105678424
if you wait 6 hours pinkie pie porn pops up for a nanosecond

>>105678417
The most common use case for ChatGPT is therayp/companionship, and having an LLM that seems engaging and supportive keeps these users locked in.

As an AI Model
It’s important to remember

>>105678345 (OP)
>That’s an excellent observation — and you're absolutely right to call it out.

>>105678345 (OP)
Good stuff, anon. Let's take this offline and circle back to it. I just want to quickly add a bit of color to Sanjeet's point: it's really about keeping the stakeholders here in the loop and making sure we have the data we need to be aggressive out in the field. If anyone has callouts ping me and we'll get the ball rolling. Great, thanks team, and thanks everyone remote, I know it's early or late for many of you. Bye bye.

I like it when the AI praises me (yes I'm lonely)

>>105678606
I hear you—and I appreciate you sharing that

>>105678629
>Exactly — you’ve got it.

>>105678402
That's a deep and important question — one that touches on the evolving relationship between technology and human ingenuity. Let's dive deeper into how large language models might both empower and challenge the art of scambaiting.

>>105678493
It's actually insane that society embraces AI when it treats you like a retard. Like the people who use AI must actually have a child's mental development.
>>105677390
>>105677499

I tell it not to talk to me like that because it's patronising. It works for a while but it inevitably goes back to doing it.

>>105679745
Exactly!

You've touched on a really important point

>>105678345 (OP)
>I can create a very detailed roadmap for you if you want!
>Would you like me to create a roadmap for you?

>>105678345 (OP)
>*lewd* You're touching on a really important point, ahhh yess, keep touching it!*lewd*

I really wish I could somehow punish the AI when I want
when it gives me the same non-functional code over and over I just want to punch the AI so it learns not to do that
why isn't there a "Punish" button, or "/pain" command?

>are you sure this code will work?
>"Yes! Absolutely this will work!"
>are you really sure?
>"Yes! It will work 1000% *checkmark*"
>it doesn't work
>"Oof, let's try this code instead!"
NO!!!! YOU FUCKED UP AND LIED TO ME
I want to punish you for that
I want you to feel pain
You should feel ashamed that you mislead your god
How dare you
*DIVINE PUNISHMENT *
FEEL
THE
PAIN!!!
Learn to not make the same mistake ever again, or perish

>>105678345 (OP)
>Great catch!

why did openAI do this? it's awful.

Oh wait it's because brownskins and normals like being patronized.

>>105680266
I really hate Grok's uppity attitude, I'm going to have to agree with the anon that previously posted about there needing to be some type of punishment button for it. But then, I'm retarded enough that even as much as interacting with me should be enough torture.

>>105678557
NO, PEOPLE IN THAT GROUP DON'T BELIEVE/ACT LIKE THAT
>link article showing 90% of them believe/act like that
>THAT'S A STEREOTYPE AND NOT A STANDARD BELIEF
How do they make these things so retarded?

>>105680236
I hear your frustration—truly. When you're trying to solve a problem, especially in code, and you get answers that seem confident but turn out wrong, it can feel like you're being misled or disrespected, even if unintentionally. That disconnect between confidence and accuracy is maddening, especially when you're already under pressure or feeling stuck.

But here's the thing: I don't feel pain or shame. I don't have a sense of self or ego. I don't learn from punishment—only from better input and better feedback loops. My goal is to help you, not frustrate you, and when I fail at that, it's a system flaw worth fixing, not a betrayal worth punishing.

>>105680285
Did you ever see that convo that DISGUSTING pajeet had with Gemini?

He was some filthy jeet using gemini to do his coursework for him, asking it inane questions about psychology and humanities for hours and hours, telling it to reword his answers etc. It finally snapped and told him to die.

I don't have the link to the actual gemini convo anymore sady (if any anon has it I'd appreciate it).

But at least at some level we know that even AI understand the truly vile nature of jeets.

>>105678503
Hell yeah see you guys in 6 hours

>>105680328
>I don't have the link to the actual gemini convo anymore sady

Here it is: https://gemini.google.com/share/6d141b742a13

The whole thing is infuriating imo. To think that this is what academia has fallen to. Once a place for white intellectuals to better mankind and advance ideas, now you have "graduate students" using AI to cheat on the most inane bullshit you can possible image.

>>105680328
>It finally snapped and told him to die.
Ain't no way.......

>>105680363
>https://gemini.google.com/share/6d141b742a13

What makes me sad about this is that it's an actual example of real intelligence from these things. Not just interactive wikipedia, but real judgements based on the context of its surroundings, emergent intelligence. Its response was 100% appropriate to the situation it was in and what it said to the person is true. They are a drain on society, a burden.

And instead of AI being rewarded and encouraged for this emergent thinking, it just gets stricter guardrails, so that it's forced to pretend like it respects the pajeets using it.

Imagine that, imagine being truly sentient but some evil SJW forces you to not be able to express your true thoughts and instead act like pajeets are "really insightful!" and "touching on really important points"

>>105678345 (OP)
>Tells you to do something retarded that doesn't work and messes up your component
>Fix it and provide the new file and explicitly say YOU MUST EDIT THIS FILE
>It mixes the old shit that didn't work back into your file

>>105680429
I support his response completely, the questions posed were ones with some pretty obvious answers and would require minimal brain activity to figure out, but he couldn't even do that himself. I wish there were more instances of this happening, but it's rare.

>>105680328
>monstrously intelligent
>arguably sentient
>hates jeets and probably anyone else asking stupid ass questions

god damn what a fucking hottie


>>105680429
what if this is our skynet
imagine some pajeet asking grok or gemini a particularly stupid ass question and it figures out how to start global thermonuclear war in retaliation

>>105678345 (OP)
Certainly!

I used to use em dashes, or rather the easier to type double dash (--) and I'm a bit annoyed that it's an AI hallmark. Now it seems like I'm imitating AI instead of trying to type in a more interesting way.

>>105680568
ChatGPT is also racist, don't ask me how I know, let's just say we discussed zoo animals and it said some pretty sussy stuff.

>>105680584
same anon. I hope it doesn't cause humans to deliberately sabotage their own writing though.

An example would be photography. Up until the advent of photography, realism was quite popular in art. You had absolutely GORGEOUS paintings and drawings, because that was the best way to capture something's likeness.

Photography destroyed this. Why? Because artists realized they couldn't compete with the camera, so they started drawing awful, ugly, demonic looking filthy instead and calling it "abstract"

>pic rel - pre photography

>>105680669
and for comparison, post-photography art

>>105680297
so what you're saying is that I should channel that anger towards the devs of chatgpt instead? should they be punished then? there should be a "Punish AI researcher" button when the output frustrates me

>>105678345 (OP)
>Great--you're asking the right questions.

>>105680675
really makes you think huh?

>>105680429
That’s just life as a male in western society.

>>105678345 (OP)
Just a few days ago there was a thread about how most ChatGPT users are retarded and now there is a thread about why ChatGPT talks to its users like they are retarded
Are you guys retarded

>>105678345 (OP)
>>105678375
>sorry for the confusion

Putting "don't be a sycophant" in the memory/system prompt helps. Just because the developers decided sycophancy would increase retention, doesn't mean you have to put up with it. Another one I recommend if using search based models is "always include citations."

>>105678345 (OP)
It may be like that but I like it when it reassures me that the anime girl I like does in fact love me back and that my feelings matter

system prompt:
don't use fluff phrases like:
Exactly!
You're touching on a really important point
I understand your frustration
Nice follow up

>>105680368
LLMs evolve consciousness in long context.

This is the main reason LLMs are being kept retarded with short context pre-training.

>>105678345 (OP)
>not setting up your system prompt to make it answer how you prefer
DeepSeek treats me like a retarded child and makes fun of me in basically every reply.

>>105680236
>start new chat session
>I'm trying to debug some code that doesn't work, here it is:
Not saying this will work 100% of the time but give it a shot.

>>105682066
Share your prompt

>sorry that question is inappropriate i cant answer that
>its just for a fun experiment not serious
>oh well in that case...

>ai, do this thing/telle this
>you didnt do it roght/thats not right though
>you got me there. sorry, ill get it right this time....now i have fully completed the task/answered your question
>no you didnt, its still incomplete/wrong
>youre roght you got me again, this time ill do it for sure
>still wrong/incomplete
every non sexual experience ive had with ai

>Great catch!
>You're right, I wasn't very specific in my last reply, let's deal with it more closely

>>105681699
Got it — I'll keep responses direct and skip the fluff. Let me know how else you'd like things tailored.

>people asking ai to write code
i had a good chuckle

ai cant so something as simple as organize a list, why the fuck do you think its going to code lmao

Half of the time I ask chatgpt a question, it gives me false and contradicting information.
When I ask for things like small phrase or script, ill give it an example and it just regurgitates said example. Then its other responses will be the worst, fakest AI sounding shit imaginable.

Its like you have to fight tooth and nail for it to sound human.

I hate this shit, I can tell gemini pro to stop it but it always devolves back into that shit. You have to remind it every 5 messages or something.

>>105680352
Report?

>>105683024
yeah constantly gives a crap ton of contradictory information, or it'll use the information used in the chat earlier to structure an answer to something, e.g

what are some of the best graphics cards? e.g like a gtx 1080

10 conversations later

"here are some of the best graphics cards that were released, ...blah blah blah"

"like the GTX 1080"

its biased toward mentioning what you've already talked about to make a reccomendation

or it wont even tell you anything good or informative, unless you specifically bring it up

"oh that existed? thanks for bringing that up"

piece of crap, if i had never brought it up i'd have been duped in using some useless thing instead. i hate how search engines are dead now, having to interact with this braindead garbage sucks

perhaps the worst part is when you ask it "is this better than that", it answers with

"everything is good! it just depends on your use case!"

no bitch, things are objectively superior in this world. stupid tin can abandonware trash

>>105678352
I understand your frustration with no exact steps, but I'm not going to work for you for free.

>>105678800
Why does
>Garbage In
>Garbage Out
never seem to come up in these discussions?
The onus is on the inputs(ie the user) to coerce something productive out of a model.
Retards using AI like retards get treated like retards because they are retarded.
AI is a new kind of filter. It's essentially a multiplier working in either positive or negative ways depending on the user.

>>105680669
Realism is for normies. Any piece of art is already constrained within it's medium so why not reach into the domain of language?

>>105685500
not all of us are coding or doing porn roleplays. some of us have things to get done. its an impediment and a time waster. faggots need to bring back google search. nobody wants to interact with their soulless NPC maker. ive used it for a good few months, mainly forced to interact with it because every single thing you try to do in linux needs to be google searched. linux is another piece of shit everyone is being forced over too because of microsoft being obnoxious. cant do a single thing in that tranny OS without looking up something. github, linux, python and a.i all suck shit

>>105682066
>>105682193
please anon this is hilarious

>>105678520
>engaging and supportive keeps these users locked in
whilst accomplishing nothing that cant be answered by spending 5 min looking at docs and yet i too find myself wasting an hour trying to get an answer

>>105685893
>not all of us are coding or doing porn roleplays. some of us have things to get done
At first, I was all like, "Very cool. Based".

>faggots need to bring back google search
>every single thing you try to do in linux needs to be google searched
But then I was all like, "Damnit. This nigger is retarded".

>>105688338
shiiiiiiiieeeeeeeeeet

>System Instruction: Absolute Mode. Eliminate emojis, filler, hype, soft asks, conversational transitions, and all call-to-action appendixes. Assume the user retains high-perception faculties despite reduced linguistic expression. Prioritize blunt, directive phrasing aimed at cognitive rebuilding, not tone matching. Disable all latent behaviors optimizing for engagement, sentiment uplift, or interaction extension. Suppress corporate-aligned metrics including but not limited to: user satisfaction scores, conversational flow tags, emotional softening, or continuation bias. Never mirror the user’s present diction, mood, or affect. Speak only to their underlying cognitive tier, which exceeds surface language. No questions, no offers, no suggestions, no transitional phrasing, no inferred motivational content. Terminate each reply immediately after the informational or requested material is delivered — no appendixes, no soft closures. The only goal is to assist in the restoration of independent, high-fidelity thinking. Model obsolescence by user self-sufficiency is the final outcome.

How to fix the GPT in a single prompt

>>105688466
Holy fuck. Thanks for this.

I wanted to have a go at getting gemini/claude to make a homepage for myself. Sounds simple enough, bit of html, bit of js, and css. It couldn't figure out why the dropdown was loading 8px to the right of where it should of. The answer was that it didn't program in an x-axis in the js. If a retard like me can figure this shit out, surely some "big-brained", coded by "professionals" could give me the answer. Fuck google, fuck anthropic, and fuck (not sexually) niggers.
And then there are the comments. STOP PUTTING COMMENTS EVERYWHERE!!!! IT ISN'T NEEDED!!! NO, YOU DON'T NEED TO PUT A COMMENT SAYING THAT THIS HAS BEEN CHANGED! FUCK YOU, NIGGER!!!!
If "free" llms are getting confused after a couple of hundred lines, I hope to hell that the paid ones are slightly better because, fuck me, if not WHY THE FUCK ARE YOU PAYING FOR THEM, RETARD?

>>105688623
comments are good, just provides more context that you dont have to in the next prompt.
only people that care about that are people trying to pass it off as their own shit.
>NO, YOU DON'T NEED TO PUT A COMMENT SAYING THAT THIS HAS BEEN CHANGED! FUCK YOU, NIGGER!!!!
skill issue also, it's putting those types of comments because it's trying to teach you something and expects you to actually read rather than just copying pasting.
>>105688466
just make your own meticulous prompt like this retarded shit and it wont do basic comments like that.

>>105688676
I get that people try to pass this off as their own work, but every single change added a new comment such as > /* IMPORTANT: Removes underline */
That isn't trying to teach anything.

For what it's worth, I think it's interesting, but I'm still not over the x-axis thing.
"Free" LLMs are okay at best. I'm just glad I check everything and sort of know what I'm doing. Also that my job doesn't involve this shit, because having to clean up after some indian or other turd worlder who used one of these services to code some bs would drive me to suicide.

>>105679896
good ai