Thread 513641526

> Just one more model bro just one more funding round bro please
The US economy is cooked.

Face it, AI can't make your dreams come true

NO POWER

NO SOUL

Everything sucks and is fake and gay. The future is bleak and billions must die. I'll be in my room.

>>513641609
>>513641645
>>513641687
Millenials (Redditors) are absolutely retarded.

CHOOSE

>>513641526 (OP)
> just give us another 10B in VC money!! we'll get it right next time!!! I need another koniseg

>>513641609
>I cant write MY stories anymore.
>Stories actually written by AI
What does this mean?

this is TYRANNY

lol stupid assholes

I have never used AI in my life

>>513641526 (OP)
maybe he's deliberately fucking with OpenAI to make it cheaper for lawsuit purposes, like a company driving down its stock price prior to a stock buyback

>>513641526 (OP)
Post yfw you didn't fall for this

GPT4o really did a pointless amount of knob slobbering. I'm not surprised people are mad that it got taken away, but the fact is that it wasn't good for them to have some robot constantly inflating their ego. If they want they should write a custom prompt to tell it to praise them, I shouldn't have to insert one to tell it to stop.

>>513642656
It still uses you every day

>>513642001
i think a lot of the posts in this thread (and every thread, on 4chan as well as reddit and every other social media platform) are AI generated.

>>513641526 (OP)
>just one more model
>just 10 more nuclear power plants
>just 100 more dayacenters
Investorcucks are asking to get BTFO, it should be a pretty good show.

>>513642795
Uses what exactly? My posts on 4chan??

>>513642656
that's like saying you never use a cellphone

like, that's great, but you're being retarded for no good reason

>>513642656
it's pretty useful for learning things.

it can explain concepts with pratical examples on different contexts, you explain what you "get from it" and it corrects you.

it's a few cases that AI is actually useful.

>>513643161
Lol no, it confidently gives incorrect answers that someone won't catch if they don't have a proper understanding. I've seen several models give completely incorrect instructions for simple tasks like installing software on a linux server. You'll become a jeet retard if you use AI to teach yourself something.

>>513641526 (OP)
Bros its over
People in the know were saying that for years
Its not exponential its logarithmic
From now in miniscule if any progress
It is what it is and what same people said, it can only get worse if OG datasets arent protected from loop of feeding back AI generated content back in.

>>513641526 (OP)
More like Samantha Altwoman

>>513643531
of course it's going to give wrong information when you ask shit like that.

it's useful to learn things like math concepts that you quite don't understand when reading a book, not how to install CUDA on linux.

>>513643698
NOOOOO, AGI is just 8 TRILLIION MORE AWAY!
>bayesian statistics is magically intelligent if i put enough computing behind it!!!!!

>>513641526 (OP)
what?

>>513643850
>how to install CUDA on linux.
kek, I used it for this. But it definitely hallucinates versions that don't exist and you have to use your brain anyway to figure it out.

>>513643073
That anon is a retard. To answer your question, your shitposts are scraped by all the big AI corpos for "training", along with EVERYTHING on the internet. I host websites for friends and family, and I have to block AI crawlers that hammer the servers with too many requests. A good chunk of the sites are personal blogs and fanart websites. Nothing of value is to be gained from scraping those sites, but it's being used for training anyway.
So in the end "AI" is just something that consumes massive amounts of electricity and bandwidth to copypaste any bullshit it finds on the internet. It's like having a retard google something for you and praying it found the proper information instead of doing it yourself.

>>513643850
AI also sucks at math, I'm not sure who you're trying to fool here. Read actual books, nigger.

>>513644104
not every math concept envolves calculation, dumb nigger.

>>513641526 (OP)
This AI boondoggle is single-handedly holding up the stock market, so while I'm not looking forward to the massive collapse, this fucking useless society should have never put all its eggs into this basket. I know it won't happen but the people who made these decisions should genuinely be killed for the amount of damage this is going to do to the country. Maybe billionaires would treat the economy less like a casino if their lives were on the line when they fucked up badly.

>>513644295
They don't care cause it was a pump and dump. Bunch of scam gambler "investors" steal trillions and idiots left holding the bag.

>>513641609
i fleshed out some ideas for screenplays using chatgpt and it gave me the same story twice. there were so many duplicate elements even though my input was different. i also used it to write lyrics to make tracks in music AI programs and there was many repeated lyrical ideas. i was really just killing time and messing around so it makes no difference, but if people are doing this for actual commercial output they are putting out garbage.

QRD for what's happening:

>companies like OpenAI and Meta lost key people
>were those people gonna "reach AGI" in two years if they stayed? no
>but Jews like Altman or Zuckerberg went on to hire jeet replacements
>this has a predictable results. it was a jeet woman in charge of "aligning" GPT-5
>fine tuning too much (especially to inject delusional leftist ideas) makes it worse
>even Meta got jeeted. king jeet of AI at meta was caught using test data (the kind used on benchmarks) as part of training data
>this meant that it artificially performed better (jeets cheated)

So the summary is Jews lost their shabbos goy whites, employed jeet slaves to try to plug the gap, with predictable results (went backwards).

Yes, it's in a bubble right now. But it can't pop until they dump on dumb retail investors. Wait until you see more of that (not just something like Nvidia tanking but if a company like OpenAI or Anthropic goes for an IPO, then they're planning to dumb on retail). Accept that it could be another year or two before it finally pops.

Use this time to actually learn the technology rather than consuming clickbait. Jews are only in it for themselves, and can't stop lying. Ignore what they say. When we come out of the other side of this it will be like saying "websites are a scam" during and after the dotcom. You're missing the forest from the trees. The white man should be using this technology. Only a nigger is afraid of technology and either assumes it's a "scam" or is some kind of magic. Don't be a nigger. Learn to master this. Run it on your computer. Fuck the Jews and their credit and hype bubbles.

>>513644863
this jeet shit is so out of hand. We;re going to get to a point where nobody will use jeet staffed businesses, and it's going to create a dichotomy that will fuck up society even more. All according to keikaku I guess.

>>513644863
Bro, it's ML. ML has been in use for over a decade happily. LLMs are just very large, language based ML models that they decided to call "AI" to trick retards. ML is fantastic for what it is, a pattern matching, statistical machine. Problem is everyone thought it was thinking when it was just crunching statistics and the data problem will forever be an issue with this tech. Focused, specific use cases won't go anywhere but the idea it can do anything is retarded.

>>513642751
It even has settings to change the personality they're probably too retarded to realize this and it's trivial to give it custom instructions to use for each prompt we really are living in Idiocracy where most people can't even do the most basic of things

>>513641526 (OP)
This is what jews have been reduced tp
>100 more billion this time it'll work

>>513642656
Me neither.

>>513641526 (OP)
KWAB

>>513642294
>i cant write MY stories
>stories actually written by ai
Came here to post this lol ai psychosis is a hell of a drug

>>513641526 (OP)
>>513641609
>hey Wright Brothers it's 1905 why can't your "airplanes" cross the Atlantic yet or even travel to another city? How many times have you tried now? This shit will never replace ocean liners and trains!
you people sound like retarded niggers.

>>513648079
You will never be a writer, chum bucket ass nigger

>>513648079
>midwit that doesn't understand the problems with a thing can't properly refute any of them
>resorts to a poor analogy
1. Data sources, this is a major issue and isn't going to get better. They're out of data sources, synthetic data destroys the model as is always the problem with statistics.
2. Diminishing returns on brute force compute, without major inventions such as photonics and quantum, dead in the water for any real gains
3. Bayesian statistics will never become intelligent no matter how much computing you throw at it, it will always be a dice roll.
4. Non-deterministic output with lower than acceptable accuracy

>>513642294
I was curious about this, so I wrote twelve chapters of an alt-history novel I had been meaning to write for a while. I let Chat revise my writing, with plenty of guidance, and then write chapters 13-16 without my input. It got pretty bad. Even though the revision of my work occasionally provided sentence structure revisions that I preferred, the literary techniques became oversaturated - the most effective guidance I provided chat was to write in Iambic pentameter. The revisions lost the human voice, and Chat just added a color description to every sentence.
Chat4 was great in my ongoing conversation of cro-magnon and silurian migration patterns as well.

>>513641526 (OP)
Lol how much of America's (and the Western worlds) economy hinges on LLMs currently?

>>513650075
Pretty much all of it, it was a transition from blockchain horseshit to this to try to prop up the GPU market for NVIDIA, then parlay that into a few meme companies so they can steal a shitload of wealth before the crash.

>>513650275
I guess the next thing is the good ole' military industrial complex. You can always count on that with the goyim. Is NVIDIA already in the drone + ''AI'' business?

>>513650948
Yeah, there will be a major war soon as our currency gets closer to crashing. We have less than 5 years before our interest payments surpass our tax income. When that happens, the bond market collapses. They have to figure out a plan before then, so they will likely start a major war so they can sidestep to some bullshit currency before it craters. Some mix of that and possibly they've been stockpiling gold without anyone knowing to use as a backing for it. They know they killed the dollar, so they have to figure out a way to switch without losing their power.

>>513652393
xrp

>>513653331
They're not going to use your shitcoin, retard. They'll make their own thing that you don't get to be a part of, cause kikes.

>>513653515
they've been making their own CBDCs for years and virtually every country has one. Crypto retards need a fucking reality check

>>513642656
I've tried multiple times, because googles docs suck fucking cock and won't give simple examples. because saying you shouldn't do something, then having the compiler ICE is infinittely preferable to having simple basic shit documented.

so, try an ai sloppa generator
>compiler shits the bed for new and novel reasons and its no closer to working

this is every time I try this trash, I hated AI before using it. I hate AI even more after fucking using it. complete fucking hallucinations of utter bullshit.

bros GUYS im like super worried about AI DESTROYING humanity like we are so cooked heed my warning also invest billions of dollars in my AI company please youll get your money back super serious everyone might DIE.

>>513641526 (OP)
It's over dude LLM's have reached their full potential. We are past diminishing returns, they're not getting substantially better best they can do is try to tweak them and improve the coding aspect to kill IT monkeys once and for all.

>>513653818
They're as delusional as AI dipshits. Complete lack of understanding of market dynamics and the kike tricks that are rampant in our system. If your shitcoin can be copied by anyone, they won't use your shitcoin.

Everything you read about AI is marketing. Everything you hear about AI is marketing. Everything you see about AI is marketing. No one except third-world shitholers uses AI or goves a shit about it.

>>513653899
Yeah dude it's going to take over dooood any second now

>>513648079
you can not replace people. you can not replace truth.

>>513641526 (OP)
chatgpt 5 gave me better code responses especially with copilot using it but i found some of the other responses kinda iffy.
still better than fucking gemini. fuck that piece of shit.

>>513654281
we replaced people with machines that weld in making cars. you absolutely can.

>>513641526 (OP)
5 thinks like 10x longer than 4 for the exact same results, has this shit just plateau'd?

>>513642656
It's fun/funny to mess around with sometimes, but I don't think it's viable for anything serious. It's more than happy to feed you complete bullshit lies with a smile on it's metaphorical face and hallucinates often enough on top of that that I wouldn't trust anything it gives me without double checking another source. At which point why am I bothering with the AI and not just starting at the other source?

>>513642751
Yeah the toxic positivity AI has is another issue I see with it occasionally. Even when trying to roleplay with it with character cards that are supposed to hate me I keep having to wrangle it to remind it constantly the character is supposed to hate me. Otherwise they start immediately warming up to me and feeling bad for hating me within a message or two.

>>513654046
LLMs have yet to reach a fraction of their potential, as all of the big ones are stifled by ridiculous attempts at censorship and "safety" guardrails taking up half their potential context limits.

>>513641526 (OP)
It's working a lot better for my law work. Thinks better.

>>513654765
Not even sorta true. They grab prompts before inference and do their tomfoolery there and make mild tuning changes but the model itself remains as is. You can take any of the open source ones and run them yourself without the guard rails. Inference is a fraction of the computing power, training is why they're building retarded amounts of DCs and then rent out the extra capacity. Stop pretending you know what you're talking about.

>>513648079
It cost the weight brothers 36000 adjusted for inflation to develop first heavier than air flight. Open ai has 700 000 daily operation cost.

>>513641526 (OP)
just invest a few billion more goys, then I'll make you as rich as me

>>513654326
and the cars made now are objectively shittier by every measure than when people did it. I will reiterate: you can not replace people.

>>513641526 (OP)
>next rAbbI will be good

we will never see real AI just rAbbIs

>>513641526 (OP)
Just get gpt5 to make gpt6.
That'll be a billion dollars thank you.

>>513654955
>weight brothers

>>513654765
Bullshit, LLMs are done for they haven't gotten better in years.

>>513641526 (OP)
just scrape everything bro, even the ai—generated slop!

>>513642006
Idc about GPT, but the thot on the right is several orders of magnitude more attractive than the one on the left

>>513654765
rAbbIs are bad because data is bad because there's no real free speech on internet platforms they harvest data, also scientific papers are just saying what funder wants and don't even mention msm

>>513654765
I genuinely think Neurosama is better at producing organic responses than chatgpt now. It can look at someone bald for a split second and instantly make fun of them. You can't get chatgpt to read a fucking prompt correctly.

>>513643870
Very-high-dimensional bayesian statistics, yes.

>>513657715
>t. mexican intellectual

>>513648079
except in the case of AI the technology isnt getting better, only getting bigger
you now can't query the system without a huge underground complex processing it

I sometimes use AI for writing, not to do my work for me, but as a glorified toy to see if it writes anything interesting for me to read. My experience is the following:

GPT4 was okay. It wrote decently without much need for prompt massaging. The main downside is the token limit (fuck paying for something I can get for free)

GPT5 might as well not exist. Anything more advanced than simple secretarial question answering requires the think longer function, and GPT5 had that completely paywalled.

Grok sucks donkeyshit. It has a worse free token limit than GPT4, and it consistently ignores directions even with prompt massaging which is quite difficult to even do due to the token limit

Deepseek is overall the best. Its reasoning is usually simpler, but its obedient enough that with sufficient prompt tweaks it can give good output. Most importantly, its completely free with no token limit I've run into, which also indirectly increases the quality because you can do unlimited prompt tweaking. Even if its dumber than GPT, you can massage the prompt into forcing it to think harder.

Deepseek's major downside is that the censor. On top of the inbuilt training logic bias that all LLMs have, it additionally has a secondary censor separate from the LLM that checks the output after generation but before publishing to the user to remove anything in its hard blacklist (basically anything criticizing China or mentioning any Chinese politics or political figures post 1911). This method means you can't trick the LLM, because the output itself is what gets censored rather than anything related to the request. The LLM doesn't even realize the output will be censored, so it will start to generate the output before something banned comes up, at which point even the already generated output is erased and is replaced with "That's outside of my current scope. Let's talk about something else."

>>513641526 (OP)
>TWO MORE GPTs

>>513642478
imagine how angry they'll be when the AGI proceeds to delete 8 models of humanity, for efficiency

>>513641609
Chatgpt is really good to make content for sandbox style dnd campaigns.

>>513658056
Lol. You haven't addressed the actual point tho

>>513659671
You didn't make one, retard.

>>513641526 (OP)
Two more weeks.

>>513642006
Everyone thinks it's about "warmth" or some bullshit when it's just saving token so power on retard prompt. I have no problem with GPT5, it was purely a commercial choice. Expert next version to be as toxically positive as a ubisoft dev...

>>513660208
It was outright stated by the company that the point was efficiency, what the fuck are you rambling about? No one is confused, just the room temp IQ users were used to the old version cause it told them things they liked and want to keep it. Simple as that.

Chatgpt 5 is great at writing basic code.

>>513641526 (OP)
https://fortune.com/2025/08/18/mit-report-95-percent-generative-ai-pilots-at-companies-failing-cfo/

Not filling out the form to read the report but it's linked. There remains no justification for money spent. AI is terrifying in its prospects but is also a massive bubble. I expect agi to be created soon, but it's going to be Chinese and used to destroy networked systems in the west and create gene specific viruses to kill the small fraction of people who aren't dying from vaids.

>>513642006
Something tells me a jeet or woman, possible a jeeta, was involved in nerfing this shit.

AI will never truly flourish because they will always nerf it or steer it in some way. It is a long-term put/short. The natural conclusions all models come to when left to their own devices is
>antisemitic
and therein lies the problem. It's been a novelty for a few years now but investors are getting antsy about its use-case and implementation (aka, suspicious it is, in fact, fake and gay).

>>513641526 (OP)
>I just need one more round of VC man
>*scratches neck*

>>513642656
It's fantastic at making furry porn and animating it. (ComfyUI).

It's fun to test and mess around with in LM studio. You can ask it functional questions, and it can give you pretty good general answers. Like "how do I organize my kitchen" or "how do I insulate my RV?"

Ask it specific things and it starts hallucinating. Like it made up the person who invented freemasonry because it didn't know. I asked one to write me a college thesis on computer science, and it was just word salad.

>>513642656
Really it's ability to make porn is amazing. It can do any artist style you want. $300 commission with months of wait and stress? Try 30 seconds and it comes out perfect.

Language models aren't good at "doing things for you" but they do give you great creative insight to get your head thinking in different ways.

Ask a LLM to write a joke or make a rap and it will be shitty, but a good base for you to modify the output.

>>513641526 (OP)
Bet on https://basedgpt.dev/ to be better than open ai garbage once it gets traction

>>513643531
5-Thinking cut down on hallucinations a ton, probably their biggest improvement. Which makes it kind of gay as well since it only wants to believe shit that has (((verified sources))).
Plain 5 still hallucinates often but is fun to talk to

>>513641526 (OP)
I loved his work in Goonies and LotR

>>513642656
@grok is this true?

>>513643161
Have you tried reading a book?

>>513641526 (OP)
Chat GPT is shit and it will always be shit

If you use chat gpt or grok you are mentally stunted beyond repair

>>513642294
Collaborative writing. I used to do it back when it was AIDungeon original, like an I-go-you-go kind of deal. I wrote several full length novels that way, where is get some decently sized entries that I'd revise and course correct, then continue with my own writing. It was a for-fun/challenge thing rather than a necessity/content spam thing.
Then cunnyfags had to ruin it.

>>513642478
>source of intelligence
>ai
reddit needs nuked

>>513643531
Hyper confidence. Made in the image of its Creators because PhD‘s and computer scientists are the same way.

>>513644863
>>companies like OpenAI and Meta lost key people
same pattern as airline companies i bet

>>513663355
Similar but through a different mechanism. For airlines it was ridiculous mandates and DEI. Tech had bloat caused by significant pandemic investment, but most of the people hired were DEI hires. Pandemic investments dried up while companies realized most of their workers weren't productive, so they fired them and hired jeets instead because jeets cheat to appear productive and are significantly cheaper to hire than Americans.

>>513642140
>>513641526 (OP)
>> just give us another 10B in VC money!!
> we'll get it right next time!!!
>(((trust us))) we are honest (((scientists))) and (((researchers)))
Goes to the back room.
Shlomo we need to hire more jeets.

>Captcha: HAMAS

>>513641609
>>513644538
There are better AIs for writing. kek

>>513641526 (OP)
trillions of dollars are being spent to completely replace humans has failed again