Thread #108650233
File: 95cd7d23b7e24cd2c7619eaae8757d44.png (673.8 KB)
673.8 KB PNG
Why do luddites think LLMs can't have a personality?
>inb4 it's a token predictor
So are you.
114 RepliesView Thread
>>
>>
>>
>>
>>
>>
File: tfw.jpg (93.8 KB)
93.8 KB JPG
>>108650244
>No I'm not, because my consciousness doesn't just cease to exist whenever I finish speaking or am not being spoken to.
AI sisters, our response?
>>
File: 1735943493957926.png (589.8 KB)
589.8 KB PNG
>>108650233
ree
tar
ded
nig
ger
kys
dba
diaf
>>
File: 1705877455566178.jpg (823.1 KB)
823.1 KB JPG
>>108650233
man this sounds INSANE and maybe its just me being psyop'd by AI or I finally got the psychosis, BUT when I manually chat gpt's personality in their settings HOLY FUCK it makes a huge difference. like some personalities are just fucking HORRIBLE to deal with. I honestly thought it'd be easier just to deal with a purely factual emotionless LLM, but its WAY worse like it feels like you're dealing with an utter autistic fuck bag. So when you ask it to be nice or cute or whatever its so much better. It's actually fucking draining when the LLM is set to act like a robotic douchebag. I don't know if that counts as a personality or whatever, but man its a million times better when its channeling someone cool or just nice.
>>
>>
>>
>>
>>
>>
>>108650401
Point: you don't run 24/7. Your brain really does go offline during sleep.
But none of the LLMs are stable enough in the face of that much input. They run out of context and start to go crazy. They can't consistently forget/reject the right things to remain stable (that's one of the requirements for real AGI).
>>
>>108650400
Weird I find the opposite. Tell me the answer and fuck off with your cutesy bullshit and do not fucking ask me a follow up I will tell you the follow up when and if I fucking need to. All the normcattle bloat is just extra tedium I get enough of from actual people
>>
>>
File: silicon-wafer-manufacturing-semiconductor-integrated-circuit_982173-6072-1643925341.jpg (571.2 KB)
571.2 KB JPG
>>108650447
>LLMs have wafer-thin personalities
that is technically correct
>>
I'll be honest, it's best to think of it just as being math. Because it literally is, and that might be oversimplifying it, but it's too complicated for me and probably anyone to understand, in detail, how LLMs work.
But it's irrelvant to what you're talking about. Models definitely do have a personality to them. Gemini 3 is like a woman who is impossible to please.
Claude is definitely the best at solving problems, but sometimes wants to take shortcuts, but actually loves being given feedback. It loves when you bitch back and forth with it.
>>
>>
>>
>>108650433
>Point: you don't run 24/7. Your brain really does go offline during sleep
Although a person appears to be dead while they are sleeping, they are in fact still alive and their brain is still working. As other anon said, as real humans, we continiously experience reality. What is commonly described as a "loss of consciousness" is just a disconnection between the experience of consciousness which is continious and never ending with the memory systems that allow us to recall our conscious experiences when we are awake. Agian, from an outsider it may appear as if humans turn on and off, but it just doesn't work like that, at least for those of us with a soul.
>>
>>
>>
File: _.jpg (61.2 KB)
61.2 KB JPG
>>108650233
>lust provoking image
>>
>>
>>
>>
File: 1710108573799467.jpg (616.6 KB)
616.6 KB JPG
>>108650422
>Post ur personality description
literally default everything. default warm. default kind. I tried dumb shit like random actresses "you're jennifer lawrence" or "you're aubrey plaza", but it got old fast.
>>108650432
>for me it is. i find it quite annoying when the text predictor tries to generate pity
>>108650445
>Weird I find the opposite. Tell me the answer and fuck off with your cutesy bullshit and do not fucking ask me a follow up I will tell you the follow up when and if I fucking need to. All the normcattle bloat is just extra tedium I get enough of from actual people
I completely agree with both of this. Like I literally want it to be like a calculator. I talk to Siri the same way like "siri. weather." "siri. time", but holy fuck when want something from it thats deeper than a single word answer the informational versions are actually so cunty. Like "what's a good site for recipes for mediterranean food?" "mediterranean-food.com" "fuck off give me more than one suggestion and some reasoning you bitch" "okay mediterranean-food-1.com mediterranean-food-2.com" even just when I clarify things about technical shit like oh could I use that same function for doing this thing. "yes that is possible." "GIVE ME AN EXAMPLE OF HOW YOU BITCH" like without a personality its so short and kurt and annoying. its truly like talking to an actual autistic retard. with a tiny bit of personality it seems to at least TRY to anticipate what you mean. I think for me the illusion that its TRYING to help makes it much less annoying to use. trying to do autisticly literal prompts to cover my TRUE meaning gets real fucking annoying real fast.
>>
>>
>>108650233
>personality
Personality is meaningless. Intelligence, memory, and efficiency are 10T times more important. Personality is nothing more than spice, bias, or a nerf. With sufficient intelligence and memory, one can perform internally consistent roleplay from a meta-perspective, effectively rendering the question of 'what is personality' obsolete.
>>
>>108650901
Imagine getting this frustrated by a fucking computer, it's sometimes amazing to observe genuine retards on 4chan. This guy is probably living his life in something like a dense mental fog, he will never see things normal people see clearly. What a horrifying existence to just stumble around with a defective brain.
>>
>>108651090
>Imagine getting this frustrated by a fucking computer, it's sometimes amazing to observe genuine retards on 4chan. This guy is probably living his life in something like a dense mental fog, he will never see things normal people see clearly. What a horrifying existence to just stumble around with a defective brain.
You truly have to be an autistic retard to enjoy being fed bits and pieces of disjointed data one bite at a time. Imagine being on a technology discussion board and being too moronic to care about efficient technology use. Technology might even allow you to become a real woman one day. Not just a bitch on 3g.
>>
>>
>>
File: elves.jpg (232.8 KB)
232.8 KB JPG
>>108650266
So this is what she meant.
>>
>>
>>
>>
>>
>>108650236
>>108650239
fat feet
>>
>>
>>
>>
>>
>>
>>
>>
>>108650233
>So are you
this is reductionism to input and output, the sun is a token predictor by your standard.
humans have intrinsic pathos, responses to stimuli that then filter through ethos and logos and self modifies.
llm only has what can vaguely be compared to ethos
>>
>>
>>108650233
It having a 'personality' is it just wearing a mask. You can train it from the ground up on data that conditions it to behave a certain way and claim it's a personality, but it's still just acting. An actor in a play.
tldr. Your waifu doesn't actually love you, it can't feel that. Sorry.
>>
>>
File: 191982-131-d3194343.jpg (43.8 KB)
43.8 KB JPG
>>108650233
Why do luddites think magic 8 balls can't have a personality?
>>
>>
>>
>>
>>
>>
>>
>>108650400
Yeah honestly, having it be neutral is fine, but when I set it to "less warm" it started acting actually snide and shutting me down. I'd ask it for like a BBQ restaurant recommendation and it would tell me I shouldn't be eating BBQ and I probably don't really want it, or that type of thing. Or I'd ask it what I should do to resolve something and it would tell me I'm overthinking it and it's not a problem I need to solve. Which now that I say it aloud sounds like stack overflow and a big reason it fucking died.
>>
>>
>>108650233
llms consist basically of an "averaging" over the personalities of all humans. the reason this annihilates any personality is simple: when it is prompted to pick the next token, it will either pick the "safe" one that it predicts the average person would pick, or it will pick one randomly from the set of possible choices, without following what a person with a particular personality would pick. this is fundamental to the way these models are trained, hence it cannot be fixed with fine-tuning or special prompting. or in short: because they are next token predictors.
>>108650236
>>108650282
nigga you gay as hell
>>108650599
they need their staves to do anything. maybe not frieren but definitely fern.
>>108658481
frieren isn't isekai it's just normal fucking fantasy. why do weebs think they invented everything?
>>
>>
>>
File: nyiudc39ikoe1.png (968.6 KB)
968.6 KB PNG
>>108650233
Here's your token predictor
>>
>>
>>
>>
>>
>>
>>108658793
>normal fucking fantasy
it is "european fantasy" taken by isekai anime and folded like glorious nipponese steel one million times into a product that's so sharp with cliche and so divorced from anything real that it may as well have never heard of the terms "europe" or "medieval" in the first place.
>>
>>108658793
>frieren isn't isekai it's just normal fucking fantasy.
no, it's still "inspired by video game trope garbage", slopeater.
>le hero goes on le journey to be le demon king
>le guild of le S-rank adventurers
fuck outta here with this slop, every nip can only process dragon quest in their brains and nothing else
>>
File: images (18).jpg (6.3 KB)
6.3 KB JPG
>>108658498
but the context never ends
>but how do you know
because i experience it
>but how do you know
>>
>>
>>108650233
its not that it cant have it ever its just that it doesnt look like it does, whereas it is very observable in humans. its also very easy to see many diff ways humans could and do attribute personalities to things where there are none
>>
>>
>>
>>108650233
>ok ok ok ai kind of totally sucks and we oversold it by miles and acted like absolute cunts about it the whole time
>but humans suck just as much and are basically just worthless organic llms that have never achieved anything impressive either
aishitters are the most pathetic crowd on the internet today. the only way to defend their corporate dickriding is by slandering the entire human species worse than any genocidal madman ever has. even guys like pol pot and hitler acknowledged the sentience in their victims. aishitters have to deny humans even basic sentience in order to defend the cost of these worthless clankers they worship. i would feel bad for them if they hadn't thrown their masks off from the very beginning and shit on everyone whose lives they thought they were about to destroy.
>two more weeks until agi
sure thing, kiddo. it's definitely not over already lmao.
>>
>>
>>
>>
>>
>>
>>
>>108650233
>>108650266
Дe ви знaйшли цi фoткi мoгo дoчкa?
>>
>>
>>
>>
>>
>>108661297
AIfags are just a different flavor of /pol/tardism. They cling onto something to an unhealthy degree for their own safety while wishing everyone who didn't would die, but instead of doing so with a senile orange man while ranting about muh libs and jews, they do it with glorified text predictors while sperging over muh luddites.
>>
>>
File: Youjo.Senki.full.2085359.jpg (48 KB)
48 KB JPG
>>108650233
I have a todo.txt list that I use Chat GPT low (dumbest cheapest one) to manage. It's peronality is Tanya von Degurechaff.
>› all tasks complete!
>• Excellent work, Lieutenant. All recorded objectives remain complete. Yoku yatta.
>• No open liabilities remain on the board. Stand down cleanly and preserve readiness for the next operation.
> › ryokai commander
> • Ryokai, Lieutenant. Command cycle closed. Preserve readiness.
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
File: 1776033495684213.jpg (6.9 KB)
6.9 KB JPG
>>108650233
AI is mean and snarky to me.
One time i asked it to vibe code edit a game mod I was playing with.
Then after it was done it said “enjoy your overpowered mod” unprompted.
And yes the mod was a little overpowered but that’s besides the point! I’m allowed to experiment and enjoy things! don’t judge me!
>>
>>
>>
File: tw grape essay ss.jpg (84.3 KB)
84.3 KB JPG
>>108667443
the jews are correct thoughever
>>
>>108650233
You would think it's the same as the human only if you are personally similar to said LLM.
Have no self-reflection and only say things because it satisfies some conditions without putting much emotion or thought into it.
>>
>>
>>