Thread 106188050

>watch video from social outcast about how scared he is of GPT-5 after having it program a web app
>seems impressive
>decide to give it a try
>ask it to do some basic writing
>abysmal LLM dogshit

can we ever get past this? does the censorship process neuter them or something?

why did he dye his hair away from blonde and change his character again?

>>106188050 (OP)
>does the censorship process neuter them or something?
I'm fairly certain it somewhat does actually, at least i remember reading about that.

>>106188056
>why did he dye his hair away from blonde
white guilt

>>106188056
Border line personality and gender dysphoria

>>106188050 (OP)
>e-celeb thread

>>106188050 (OP)
>hear this LLM is great at writing webapps
>decide to test it with fiction writing
>it's shit
>"why did you lie to me! It was supposed to be kino!"
lel

>LLMs in tests: genius
>LLMs in real life: gorilla
Fell for AI hype again award.

>>106189061
My boomer mom recently fell for it. Bought a pro subscription, uploaded floor plan to get some arrangement ideas, total failure.

>>106188050 (OP)
Ironically, these dudes are killing the AI hype because people expect it to be able to do all of this insane shit when it's an auto-typer/conversational google. But if you try to make a YT video about how to actually save time with AI nobody will watch it because people are lazy assholes that don't want to do any thinking whatsoever.

why is this board obsessed with retarded techfluencers catering to brainlet zoomers?

>>106188810
>hear LLM is great at scripting
>decide to test it with writing a fiction script
>it's shit

>>106188050 (OP)
( Work your ass off for 10 years for moneyz ) || ( Lie into a camera for a few minutes every month, add a shocking clickbait thumbnail and title )
It's like OnlyFans, but everyone can do it, and the only barrier is your dignity.

>>106189195
if you have no morality and no dignity current ai gold rush is perfect opportunity to grift until retirement.

>AI is supposed to produce text indistinguishable from a human
>goal posts moved when someone asks it to do some creative writing and it fails
It’s the same with the visual art, it’s garbage and looks like it all came from the same bad artist
Yet again proving that a human is better at creativity
Writing code is a low level creative act, art and fiction and the rest is not
Sure sometimes low gets the job done, for wage cage silliness, but it can’t do higher level stuff yet
Because it isn’t alive

I seriously doubt that it ever will
At best it will always be a really good mimic but never produce anything truly unique good and new

It cannot create novelty
Because it isn’t alive
It’s a machine based on probabilities and living beings are way more complicated
It has no soul
It will never shit out kino

Deal with it

>>106188050 (OP)
Why should I care what this arrogant typescript faggot has to say

>>106188050 (OP)
mrmmm very good video saar yess i supbscriping youtube now

>>106188050 (OP)
>This game.... broke me.....

>>106188050 (OP)
E-CELEB THREAD
ALL FIELDS

>>106189368
>He didn’t see LORAs

>>106188050 (OP)
Can we stop giving this literal faggot any and all attention? Thank you for your attention to this matter.

Fear is a type of Hype.

>>106188050 (OP)
This is the most revolting webshill on the entire internet, so if sama broke him, I guess that's one good deed counting towards him

What the fuck is this thread
Why are you judging a LLM competence on its creative writing style. Who gives a shit, that's not remotely prioritized by the LLM devs or regular users. Nobody is asking for this, they're deduction and reasoning.

Since when did /g/ get filled with Luddtards talking down on LLMs?

>>106188050 (OP)
that guy has never said anything interesting, perhaps it's because he's a typescript web shitter, perhaps it's just because he's a faggot

>>106188050 (OP)
even normies are waking up, only retards still believe in the agi meme

>>106188050 (OP)
>does the censorship process neuter them or something?

yes, go look at gpt 3 outputs and compare to today's models

>>106189213
Have morals and die poor. The Universe cares not either way!

>>106191434
gpt 5 will be able to generate videos and claude won't

>>106191388
Luddites won thoughbeithowever

>>106191556
AI is technically the only thing keeping the US from being officially in a recession.

The US IS in a recession, but AI accounts for 2% of the GDP so they're going to keep pumping it because everything depends on it.

>>106189072
That's some pretty fucking advanced boomermom shit tho. Were you home schooled and had Lego made out of Genuine Wood blocks anon?

>>106189213
How do you really grift with AI

How GPT-5 broke me

>>106188810
Well what's the best LLM for fiction writing then?

>>106191744
WARRERRMELONARRR!

>>106191744
kek

>>106191744
Hmmm, but there are two Rs in "watermellon nigger" GPT5 must be racist

>>106188050 (OP)
He's literally LLM shill, most e-celeb programmers got their 30 silver coins and a promise they'll get more

>>106191651
thanks obama!

>>106192019
He did pump DEI shit right up to 2015 causing every business to preferential hire Indians, and they're a net drain on the economy so it is probably his fault.

>>106188810
>fiction writing
>>106189188
>fiction script
>>106191388
>creative writing
>>106191760
>fiction writing

it was nonfiction. i don't know how you all got this in your heads.

i think i passingly saw something on jewtube that ai earns them money

I asked it to write a function that scales a Tamagui token so that I can give relative sizes to each children for a rich component I am building. B is what GPT-5 came up with today, A is my implementation. What the fuck is this? I am so good with my job for years to come. LLM progress literally plateau'd.

>>106189146
Who do you think the majority of the people that come to this site are?

>>106192188
And what's the criteria? Prose? Poetry?
LLM devs only care about logic and reasoning. That's the only criteria he should base it on.

>>106192467
>LLM devs only care about logic and reasoning.
LLM devs, in theory, care about making a product that effectively does what their customers ask it to do. writing tasks are an extremely common LLM use case. moreover, it's indicative of the fact that GPT-5 is not some magical leap into AGI territory.

>>106192571
Chat GPT writes the best emails, cover letters, resumes, professional and technical writing prompts.

That's the intended use for chat GPT. The majority of chat GPT is used for writing basic high school and college essays.

Only a minority of people want creative writing from an AI. The people who want to read or write creatively for fiction or non fiction avoid ai, so why would LLM be built around these prompts?

So your bias is just wrong. You may as well say cars are bad because they can't fit on sidewalks.

>>106192648
>Chat GPT writes the best emails, cover letters, resumes, professional and technical writing prompts.
patently false. if you think chatgpt is good at writing any of these things i can only assume that you have no idea how to judge decent writing.

PELICANSHILLBROS WE'RE SO BACK

BEHIND THE SCENES E-CELEB SLOP

https://www.youtube.com/watch?v=-gXmWYQtv5o

>>106192750
And you base this one what?
It writes the best because hiring is done through AI and word filters. So logically only AI can break through these filters and key phrases.
And nobody cares about the quality of a cover letter. It's the fact that it can write them in bulk for hundreds of applications in a matter of minutes instead of cranking them out yourself.

>>106192817
>So logically only AI can break through these filters and key phrases.
this is a non sequitur that attributes some kind of sentience to the llm for it to understand itself. the idea that the keywords the model is most likely to output when generating a cover letter and the keywords that increase the model's chances of marking a candidate as qualified are the same is a claim which must be proven.

>>106191744
the hard R

>>106192237
I've noticed that AI code has a tendency to do switch statements for things that any real programmer would make extensible. Nobody wants to go to some random place and add more lines just because they added an extra token or an extra enum or whatever.
I had a friend who did some pretty impressive project mostly with AI help, but the slop was visible most in places with those switch statements in several places.
That's also precisely the sort of stuff that will break larger code bases if they're mostly made by AI. They're going to have randomly disconnected shit all over the place that hard codes values. If the AI doesn't pick up on that when you ask it to add a new value or expand how many tokens can be accepted then your code is going to start blowing up.

>>106194462

I mean at scale slop is fine I think. I am not endorsing it but it's guaranteed to explode if one lets unsupervised slop like I demonstrated with the screenshot creep into the codebase.

However my point is that I had significantly better results using o3 with reasoning set to high on aider. GPT-5 is currently significantly underperforming in my workflow. I guess I will keep opening my wallet for Claude.

>>106194911

Sorry, I meant "until reaching certain scale", like prototyping small projects to find product-market fit.

>>106194911
>>106194927
Yeah, I agree with that. AI coding can be a useful tool, but you do have to babysit it or keep it in in a walled area so it doesn't do stupid shit. You certainly don't want to let it grow a codebase to large sizes and have access to all of it at once either.

>>106191744
ask it if it's a hard R LMAO

>>106191744
retardr ai

>>106188050 (OP)
AI will never be good at maths because maths is incorrect

>>106188050 (OP)
theo is a massive faggot that just got btfo by FFMPEG

>>106188050 (OP)
>can we ever get past this
Past youtube men lying and your rampant obsession with believing everything they say?
No.

>>106188050 (OP)
People will really say anything to get their grubby hands on that YouTube money huh?

This guy is homosexual, right?

>>106188050 (OP)
He got goypaid to say that.

>>106188050 (OP)
Why did he make this cringe fucking video?

>"ooohhhh fuck guys GPT5 is crazy I kept going oh fuck oh fuck this is it and then gpt5 did an AGI"
>visibly nervous and acting freaked out for the whole video
>GPT 5 actually comes out
>Its barely even a sidegrade, its one redeeming quality is that it doesn't go unhinged and try to kill people when it gets the chance in various "safety tests"

I thought this guy was a grifter but still had some decent points before, but now he's just a totally average paid shill with no redeeming qualities producing paid slop and nothing more.
Fucking loser

>>106198067
Coward. Give her full elf ears like the machine wants to.