[ home / rules / faq ] [ overboard / sfw / alt ] [ leftypol / edu / labor / siberia / latam / hobby / tech / games / anime / music / draw / AKM ] [ meta ] [ wiki / tv / tiktok / twitter / patreon ] [ GET / ref / marx / booru ]

/tech/ - Technology

"Technology reveals the active relation of man to nature" - Karl Marx
Name
Options
Subject
Comment
Flag
File
Embed
Password(For file deletion.)


File: 1734060573790.png (3.73 KB, 389x129, aisucks.png)

 

So, I'm a musician, who wants to have a musical career (a lot of communist musicians had stable careers) and meanwhile stupid porkies tell me that "no, we'd prefer if you were replaced, prole, because there if no place for people like you" and I hear, not only music, but other art, computer science, programming etc. will be replaced by AI. How do we stop this, so people are still prosperous in the real socialst societies?

>>27559
Musicians have nothing to worry about AI. It's been almost impossible to make money off of selling music since Napster. Your only hope of making money is touring and merchandise. An AI can't perform live shows.

>>27560
Well, not as impossible as now, as ppl will assume AI made it.

>>27562
OK, but you were saying that people deserve to feel safe in the socialist world.

>>27563
You will be safe and protected inside a mine, now go ! ⛏

>>27564
I expected better arguments

>How do we stop this, so people are still prosperous in the real socialst societies?

I don't think it would be a problem in a proper socialist society. Your existence wouldn't be dependent on your ability to produce surplus value, so there's nothing stopping you from creating music even if the predictive algorithms falsely being characterized as "AI" exist in parallel.

In the current capitalist circumstances there's nothing really for it because AI is a symptom of greater market forces at work. Artists like musicians are now in direct competition to their corporate patrons for profits, and as long as that's the case they're going to make being a musician as economically difficult as possible if only to keep them on a leash. Then there's the AI itself which is its own bundle of scams and shell games as the tech sector flounders to try and stave off bankruptcy. Considering how expensive these algos are to maintain, operate, "train", and how poor their products are, it's really just a waiting game at this point until the whole rotten thing collapses.

File: 1734062331956.jpg (53.25 KB, 483x603, d77fc20dd2764135.jpg)

Wait for the companies to run out of money on this highly unprofitable, unproductive and expensive to run and maintain technology. Same that happened to crypto.

Like no indie artists are actually losing any livelihood from this, just artists that work for corpos they were likely expecting to replace them at some point anyway. If anything it's made artists more interesting by reminding them about stuff they can do that AI can't, like make a drawing with coherent perspective, or make a song with a single singer that doesn't sound like they smoked a pack before recording, or text-to-speech that actually sounds like a robot and not some cartoon character.

>>27567 (me)
Oh and write things in a way that doesn't sound like an allistic with alsheimers and a 3rd grade reading level.

>>27567
I hope so

File: 1734064457762.png (150.46 KB, 920x920, latest-398539524.png)

>>27567
The won't run out of money,they'll just pass the cost onto consumers, and failing that, the government will bail them out. Every possible stop will be pulled out before what workers are paid raises even a single cent.

AI is dogshit. worst future

>>27565
You don't deserve an argument. We've had this thread a thousand times. We have more important things to discuss. It's a common topic you could have researched yourself. The problem is not technology, it's the system it works within. Capitalism is why you think machine-generated art sucks. The answer to your question is, destroy capitalism and the industrialisation of art. Not an easy task, so let's get moving!

>>27559
The whole "AI is taking over the arts" is just indignation made by artists.
AI is a tool. Its not workable by itself.
AI uses pre-existing samples of media.

All it does is it makes a mishmash of new data based on preexisting ones.

It still needs a human to do the final touches.

I remember back in the 2000s and 10s when we had haptic auxiliary pads for computers.
When people made manual digital art using a stylus and a haptic pad, people were freaking out about how digital art was replacing physical art.

The same was said about digital sampler keyboards in music.
Take the Synclavier and Fairlight Computer Synthesizer.
They were the hot shit in the 1980s.
People were freaking out about those machines replacing musicians

Take On Me by aHa and Never Gonna Give You Up by Rick Ashley used those machines to make the song.

Or digital audio workstations.
Which are digital sampler keyboards on a computer.
Same thing.

Also, when it comes to arts and crafts, professional careers made out recreational activities are always scarce and rely solely on corporations.

If you're worried about being replaced by AI as an artist, than it's not AI but your shot at being in business as a musician.

If anything we need to stop philosophising recreational professions so much.

>>27571
As is every other media format before it.

>>27573
>When people made manual digital art using a stylus and a haptic pad, people were freaking out about how digital art was replacing physical art.

When digital typography replaced manual typography it really did gut an entire artistic industry. Lots of old and rare typefaces and specialized machines were just melted down and professional artisans with decades of experience and accumulated skill lost their jobs as they were summarily drummed out of the industry.

>>27562
>>27564
>>27572
I've never understood why people have such flippant attitudes towards people who are ignorant but well-intentioned like OP. These are the people you should really be putting the most effort into, to be frank.

>replace programmers
will never happen, AI needs programmers building them to exist

File: 1734066532893.jpg (71.21 KB, 288x499, 1344290461117.jpg)


>>27575
Typography? You mean calligraphy?
Or you mean like printing press kind of deal?
If you mean printing press in terms of typography, then you know that typography was demonsoed by calligraphy.
Before printing press, copies of books and newsletters has to be remade by hand.

When printing press came along in the fifteenth century, calligraphers were in crisis.

Also, manual can be digital, not just analog.

>>27577
This
People think automation or new tech means less human input.
Tech is not transcendent.

It requires someone to something to move it, externally or internally.

>>27559
>how do we fight against it?
Why would I want to fight against useless work being replaced by machines?

>>27581
Do many of my fellow communists think like this? I don't think it's useless.

>>27582
Useless in terms of a profession

Music doesn't help make food or electrolytes water.
It doesn't help grow crops
It's a recreational activity
If you do music as a hobby that's ok.

>>27583
But a lot of musicians, even in the communist days, used to make a lot of money from concerts and longplays. And merchandise.

>>27582
Art shouldn't be subject of profit motive.

Establish socialism -> automation is used to help people

Live under capitalism -> all tech advancement is used to make life worse

Fighting against tech advancement under capitalism is a battle lost centuries ago.

>>27585
>Live under capitalism -> all tech advancement is used to make life worse
But my country Poland wants to be dogmatically capitalistic.

>>27586
Too bad. We will evacuate all twelve communists out of it, and turn the rest into a large open air gulag. Don't worry, it'll be a catholic and polish gulag.

>>27584
Those were state assigned propaganda artists.
Just like how city halls have media departments to make public service advertising

File: 1734067645808.png (742.83 KB, 1200x675, ClipboardImage.png)

AI literally works by continuing the previous token with the most probable next token (with the token being a word, a semantic particle, a sound, etc). If your art is so unorginal, commercial and predictable that it is actually losing ground to AI slop then it isn't art, it is a commodity, and you are not an artist, you are an obsolete petite bourgeois of the culture industry that is getting angry like a Luddite against your competitors using a machine to make the exact same product that you do, but cheaper.
AI brings nothing new to the table when it comes to discussing art. A hundred years ago a guy presented a urinal on a museum. This is how old all this drama about "non-human" art is.
Read Adorno.
(not targeted at you OP, I'm just tired of seeing people seethe over AI killing the artist and when you look at the "art" they make it's shitty furry hentais they sell on instagram)

>>27589
This. People really have no cultural history.
They always assume any new forms of tech are automatically destroying life as we know it.

AI is just a glorified autocorrect, and if you've realized this, it's really fucking good as a note-taking tool. The only thing I really use ChatGPT for is I will use it to input voice notes, and then have it organize those voice notes and transcribe a five-minute rant into bullet points. That's it. That's the best possible use case for AI, and it's a damn good one, I'll admit.

>>27580
also whatever code AI generates would need to be reviewed by programmers and etc. it's not like AI will ever be allowed to replace programmers working on the linux kernel for example, it will never happen. AI would just facilitate but not replace.

>>27589
>AI literally works by continuing the previous token with the most probable next token
That's how the organic brain works,buddy

>you are not an artist, you are an obsolete petite bourgeois of the culture industry that is getting angry like a Luddite against your competitors using a machine to make the exact same product that you do, but cheaper.

This rebellion of service industries against the modernity is hilarious. The high and mighty leaders of opinions getting scared over some dumb machines replacing them

>>27592
They'll eventually use AI to hallucinate an entire OS like they are making AI hallucinate Minecraft with Oasis

>>27589
How do you define originality? Sometimes as a musician I like sounding a bit familiar

>>27591
I'm not sure

>>27589
you're a retard

just because ai is taking market share from furry hentai artists (it's not just them, it has an enormous affect on things like junior programmers and a lot of other 'non-creative' industries) doesn't mean it isn't going to come along and replace whatever you currently do as well

are you a truck driver? you will get replaced by automated vehicles. electrician, plumber, carpenter? all it will take is some dumbass with a tablet that gives him explicit instructions on how to fix everything as he just puts on goggles that let the machine see everything for him, and once they make the mechanical manipulation part feasible they won't even need that guy. mechanic? you're basically already done bud. programmer/software engineer? hahaha yea what job could be more safe from automation than putting numbers in a computer

>>27596
>electrician, plumber, carpenter? all it will take is some dumbass with a tablet that gives him explicit instructions on how to fix everything as he just puts on goggles that let the machine see everything for him

The thing with those jobs is that people off the streets can't fix anything even with somebody guiding them. And "mechanical parts" will always cost more because there's just soooo much organic parts running around the place

But educated positions? Oh, they are going the way of a job called "calculator"

>>27590
Under capitalism they are. Deprived of income from commissions, what will artists do? Reduce themselves to the level dirty proles and (gag) organize? :^)

>>27597
>The thing with those jobs is that people off the streets can't fix anything even with somebody guiding them.

Yes they fucking can retard, 99% of the shit your hvac dad fixes is just fiddling some valve or flipping a switch

a fucking tablet with sufficient visual sensors can literally point anybody exactly to what do with an integrated hud

it doesn't take a whole replacement to drive down wages sufficiently anyway, it just requires a massive replacement of most of the labor pool and suddenly you have no room to negotiate with porky regardless of your esteemed 'trade license' that you had to lick the ass of some older contractor to get

>>27599
>a fucking tablet with sufficient visual sensors can literally point anybody exactly to what do with an integrated hud
You are underestimating non-specialists' helplessness, lmao

>you have no room to negotiate with porky

The only thing those jobs fear is uberization of a sort, where a big company coordinates a large share of the labor pool and then can dictate the terms. But even then taxi drivers just like switched over to uber instead of getting replaced.

AI-driven cars are kind of retarded though, it's a costly computerized solution to a logistical-organizational problem that's solved by trains and buses

>>27600
the point is not ai, it's automation in general, ai is merely a new component of further automation

automation WILL replace every worker the longer capitalism goes on, this includes your hvac uncle, plumbers, carpenters, etc

even when automation doesn't work to replace those workers, porky will have replaced so many other sectors that it will drive competition to your sector so fiercely you wlll lose all negotiation power and porky will just subjugate you the old fashioned way

>>27601
It won't, just how capitalism wasn't able to replace individual self-employed workers for centuries

But it's funny to me that you think that I only able to say this because I feel myself secure in my HVCA dad or uncle. What next? Are you going to say that we must protect the jobs and incomes of furry artists?

>>27598
Ok now with that attitude I can see why some others on here said your profession is useless.
It's always the elitists who are against new tech.

The real problem with capitalism is that it makes people think that arts and crafts are inherently virtuous and essential to society as a career.

>>27593
>That's how the organic brain works,buddy
Literally untrue, actually unprovable, and even if we were parrots like you say there would be 8 billion unique parrot brains that function by integrating far more information domains than merely language or merely sound (which AIs so far treat only separately, unlike humans); it is simply impossible for an operation of reduction over a compartmentalized fraction of the output of these brains to have more capacity of producing something new than the individual brain itself.

>>27594
why the fuck are you telling me all that you mouthbreather my post was talking specifically about art and nothing else

the issue with ai is not even the replacement of workers, it's the divestment of accountability

porks no longer have to admit they took personal responsibility when denying claims for thousands of isulin perscriptions or verifying intel for drone strike targets, they can just say da computer did it so it's all okay

>>27601
>>27599
This is someone who doesn't understand trades work at all.
I job shadowed irrigation and HVAC.
It's not as simple as fiddling switches and levers.
You are a fucking retard.
Automation is only gonna speed up and multiply produce.
It's gonna reduce manufacture workers, not installers.

>>27604
I have no clue what you wanted to say

Human brain works on neurons that establish connections between each other, and those connections upon "hearing a word" flare up with a certain chance, randomly, first one to flare up continues the thought. Moreso, brain research shows that concepts/tokens in human brain are physically located in such a way that closely related ideas remain close (with "in-built" concepts, such as vision, literally connected to our physical eyes, motorics/movements to hands/legs, etc), same way AI is organized (but AI uses a hell of a lot of dimensions and axis)

The only way to claim that human consciousness is essentially distinct from AI consciousness is to believe in a soul or something

>>27606
the installation process will just be engineered to accommodate the automation process

your job is not safe, your dads hvac small business days are numbered

licking ass during your apprenticeship doesn't represent your gritty fortitude and commitment to the job, it's just some gay shit you did with another man

>>27608
>my shower broke, it's time to order a new apartment module to be installed instead of my current one

Yeah, nah

>>27609
fam thinks fixing a shower requires years of licking ass to get a license to know how to fix when someone with basic motor skills and a smartphone that just tells them which valve to turn off wont

>>27610
It doesn't require super-duper education, that's why this kind of job cannot be AI-replaced. But even so ordinary people don't fucking know how to install anything or how to run maintenance

>>27583
Why didn't you start by phrasing it like that? >>27581 makes it sound like you hate music and musicians

>>27611
>yea im an hvac guy, my job is really simple and im a dumbass so i must be really appealing to the general public with how simple and relatable I am
>umm acksually irrigation systems are incredibly complex and require years of experience to engineer and navigate, licking your uncks asshole during an apprenticeship is a time honored tradition that requires grit and intelligence to undertake

schrodingers hvac uncle

A lot of people here really underestimate the potential of ai and automation.
Once things like agi and more importantly asi comes out, humans wont be able to compete with it. The asi would do everything better while producing things far faster than humans do. All while doing it independently aka not requiring a human operator and etc

>>27613
I'm not hvac guy, don't have hvac uncle or father or brother or even sister, I merely entertained your butthurt

"Easy physical labor" jobs are not so easy. It's your bias speaking. Try fucking making something with your hands, and do it right on the first try. THAT SAID, basically everyone can learn doing it. Does this mean they will, or does this mean they'll do it right? Hell no

Also, what's your issue with apprenticeship? I'm not a burger, so don't know what's your gripe with hvac is

>>27589
>If your art is so unorginal, commercial and predictable that it is actually losing ground to AI slop then it isn't art, it is a commodity, and you are not an artist, you are an obsolete petite bourgeois of the culture industry that is getting angry like a Luddite against your competitors using a machine to make the exact same product that you do, but cheaper.
What you are talking about is current ai though? What about future ai like asi.
These types of ai would be able to produce art better faster and maybe more original than human art. In that type of situation how would human artists be able to compete?

>>27614
Nah, the abundance of labor created by AI forcing people out of their jobs will uplift some other industry, and on top of that we talk about cheap and plentiful labor everywhere in every field

Even if you expect AI to take over EVERY job, and also somehow block people from getting employed, even then you'll get people just create jobs for the unemployed or self-employ outside of "AI-capitalist" system

>>27617
>you'll get people just create jobs for the unemployed or self-employ outside of "AI-capitalist" system
that just sounds like the capitalists will live in their high tech "utopias"
meanwhile us the proles or "lesser" people will be kicked outside to fend for ourselves.
pretty grim picture

>>27618
That's not how capitalism works. Porkies can create gated communities, but they can do so because they sell goods to proles, the very same goods they created with prole labor. Porkies suddenly one day ditching labor altogether would transcend porkies into another class entirely, and that NEVER happened in history. It would require of porkies to abandon all their skills and instincts, that's just not how it works

>>27619
its called feudalism retard and it was around a lot longer than capitalism

>>27619
>and that NEVER happened in history.
because something like asi and mass scale automation never happened before.
All our past inventions still required the human operator to monitor and manage it. Meanwhile asi has the potential to run fully by itself.
Once porkie automates the majority of the processes. And have asi and robotics to monitor and manage such automation. Then why would humans be needed here?

>>27621
>because something like asi and mass scale automation never happened before.
No, because we have an example of noble idiots refusing to work and to become capitalists even when exposed to capitalism for a century

>>27620
That's not feudalism. Feudalism is a bunch of hereditary parasites owning land and extracting rent from it's use, sometimes with serfdom enforced, sometimes even with slavery

What is described in this antiutopia is like rich people's fully automated luxury gay space communism with huge swathes of humanity being simply unneeded. That's not going to happen. Israel-Gaza situation is only possible Israel wants land and is supported by capitalism. Luxury gay space capitalist communism would require no land, no rent and whatever else

>but muh resources

Well then, such a system is not insulated, and the rest of humanity isn't unneeded

>>27615
hvac bros in burgeristan are annoying dipshits who drive around in over sized trucks and listen to country music and voted for trump because they think he fix the egg prices

they have a "I'm a simple minded man with a brain the size of a peanut and my work is as basic as apple pie" mentality while simultaneously reverting to how important, gritty, challenging, and irreplacable their form of 'labor' is when their egos are challenged even slightly by anything

they hate college education because of how arcane and extravagant it seems to them to obtain a credential to do a specific thing as a certification of expertise, while… having a lifestyle based on undergoing an arcane extravegent process to obtain a credential indicating expertise known as apprenticing

they think unionizing is gay and so is taxing the rich and shit, they're just obnoxious class traitors whose answer to every capitalist critique is 'do trade' as if their sector of the market is immune to ever being pushed out when it isn't

I've done physical labor jobs, I've worked on factory lines, I've fixed my own water pipes and heaters n shit. they aren't comparable

>>27622
okay and no offense but what does that prove?
The majority of our current economy still requires human laborers to do things. Those noble idiots are just a minority. The majority of the top are still capitalists who due to current technology and economic conditions, still need human labor.
What im talking about is later when full scale mass automation and asi makes it so human labor isnt needed. We have no where reached this type of situation.

>>27625
I was saying that nobles, even when exposed to capitalism, didn't become capitalists en masse. They remained nobles until they got bankrupted, clinging to their retarded way of life even when debt-ridden and insolvent

>>27626
you believe that would be the fate of capitalists?
Clinging to the old ways?
hmmm

>>27627
Yes. They will cling to the old ways and refuse to change. We've seen it in myriad different shapes already, like capitalists rather going bankrupt than raising wages to attract workforce, because their gut feeling is that they need to depress wages to squeeze out profits, not increase wages

>>27628
This is complicated because If we examine german,japanese, british and many other capitalists, a lot of them were descended from aristocrats or connected to them.
And regarding the high tech asi thing a lot of silicon valley capitalists advocate for this transition (peter thiel and the cult of techno accelerationism.) So a chunk of the capitalists themselves already want to transition into such a system long term.

>>27629
Not everyone was an idiot, besides, starting capitals have to come from somewhere. Where German, Japanese, British capitalists had private starting capitals (and state contracts), elsewhere, like in Russia, emperor himself was ordering construction of heavy industries and putting future famed industrialist families in charge

And yeah, there are techbros. But just like with capitalism coming out of feudalism, it's just not possible without a bloody revolution, and if it's a bloody revolution, communists will steal that one and steer it towards communism

This kind of antiutopia is just anachronistic, it's on the same spectrum as fascist dreams about farmer families

>>27630
I will think about what you said, you make some good points

>>27589
Why do lay people with no background in computer science regularly relay this midwit talking point when actual computer scientists admit they understand the architecture that goes into producing these LLMs but do not fully understand how they actually take in this information and produce their outputs
Like I never read this from actual engineers and compsci people, only normies with Dunning-Krueger

>>27599
It’s actually impressive that in 1800+224 millions of workers are still thoroughly convinced their job is safe from the advance of capital.

>>27615
> Also, what's your issue with apprenticeship? I'm not a burger, so don't know what's your gripe with hvac is
Their gripe is probably that most of your arguments for why these sort of jobs cannot be automated is largely nonsensical; that these jobs can typically be completed by a single individual is itself a product of automation and preexisting infrastructure. Did you know that being a master carpenter, weaver, shoemaker, and blacksmith were once irreplaceable professions that required years of apprenticeship and decades of work to master?

>>27633
>>27621
>>27614
>>27610
>Thinks automationnis magic, that it won't require hardware and software specialists.

>Thinks doing "basic trades" only requires a YouTube tutorial.


You know this is how we get alot of hack doctors.

It sounds to me you have a serious contempt for trades.

>>27634
The only venues for automating, as another anon has put it, installation jobs, is either mechanical hands or wholesale module replacement of showers and/or apartments. This is just not going to happen

>b-but medieval artisans…

You just can take a random unemployed guy off the streets, give him a week course, and they can do the installation or maintenance job just fine and without errors. This is a kind of a job that's safe from automation, and in fact automatization makes such a job even easier and more accessible. It's not safe from uberization, though.

Also, read this from that anon >>27624

>>27559
>computer science, programming
We're overall safe to be honest. This isn't the first time they tried to replace us. Must be their fourth attempt after COBOL, visual coding and UML at least; they never work because you need an actual person to translate requirements into actionable algorithms that aren't completely filled with bugs.
I do not know the state of the music industry, but I can see at least the lower rungs of commissioms from small indie game developers and other small producers drying up.

>>27632
Most laymen rely the point that "AI taking over our jobs"

>>27624
Sounds to me more like you hate doing trades.
That description of tradesmen as Trump voting country music loving hicks who despise college education and don't respect nor appreciate nuance nor complexity is something that a lot of anti work people say.

>>27635
>won't require hardware and software specialists.
but cant asi become a hardware and software specalist though? And arguably better than a human one?

>>27612
<makes it sound like you hate music and musicians
My opinion was more general about all the different jobs being replaced by machines which is usually positive for the wellbeing of societies. Technological development is the core reason why economies grow and prosper.

When it comes to music, I value arts and crafts and wish these wouldnt be corrupted by profit motive but done for the sake of the artform itself.

>>27639
It can. But again, ASI is a machine and machines can err. So it needs maintenance.

Even if it becomes autocorrecting, unless it's sentiment and sapient, it will need humans to meta-calibrate.

AI will remain overhyped garbage for at least the next thirty years.
I have to wonder how many people smoking the AI pipe actually try to ask ChatGPT to cite some sources. It hallucinates the most basic of things – papers that don't exist, songs that don't exist, movies that don't exist, source files that don't exist, etc.

>>27641
>ASI is a machine and machines can err.
same thing with humans.Humans often err too but then correct themselves. Which is what asi could at one point do
>Even if it becomes autocorrecting, unless it's sentiment and sapient
well asi would increasingly reach the sentient and sapient point. Thats the concerning factor

Any vague similarity by an AI song to a real song will get pounced on by the industry as a copyright violation.

>>27643
I think the real fear of AI is mostly moral projection. All the technophobic cries of "robots are gonna enslave humans" is dumb.
Humans have enslaved and killed each other long before the first TV or radio
If anything, humams are more likely to enslave robots.
In fact, "robot" means "servant".
If robots take over the human race, it will be due to being abused for human convenience

>>27642
>It hallucinates the most basic of things – papers that don't exist, songs that don't exist, movies that don't exist, source files that don't exist, etc.

People do that too, though

>>27645
Im not really concerned that robots are going to enslave humans.
More concerned that porkie will just use robots to make the proles and "lesser" people irrevelent.
Of course porkie may be incapable of doing it as the other anon pointed out. But then again there has been situations where sections of the ruling class reformed society (meiji restoration)

File: 1734080806810.jpg (310.5 KB, 2048x1494, facepalm vexus.jpg)

>hello fellow gommies
>I am part of the self-employed class and afraid of being proletarianized. how do we fight against the development of the productive forces?

>>27642
The problem is that the hype means that people will believe the ai is 'intelligent' enough to make impact decisions that massive affect the lives of massive amounts of people

again, it abdicates the responsibility of shit like genocide and austerity to the fucking dell server in the it department and the shitty algorithm dystopia neoliberal hell we exist in will justify it because they will start deifying its 'intelligence' in the same way they do industry 'experts'

it will get even worse once they rig the shit well enough that it passes the standardized tests of professions like doctors and lawyers and proves it out-preforms them

>>27649
Nobody cares about your responsibility moral hazard

>>27649
Sure but even *if* the AI was capable of enough intelligence to run a society to it's absolute maximum efficiency and utility for most people, that's not the parameters the leadership of society care about, the AI's role is literally to be asked "is the profit motive and its facilitation really the best form of governance for a modern society?" to which the AI is programmed to respond "Yes." thus absolutely destroying the Tankies who think they can outsmart a supercomputer kekkeklellel

>>27650
you're going to care when the mortgages rise 3000% and 35% of the workforce is laid off and you get matched to the 'predictive policing' algorithm because the fucking ti calculator said it would make share holders stonks more efficient n shit

>>27652
Wouldn't they use algorithms to calculate the population number based on vacant job spots?

>>27652
I'm talking about you constantly bringing up how Pentagon will blame AI for bombing hospitals. Nobody cares about this, they always had stupid excuses

>>27652
They are already using ai to automatically deny job applicaitons and shit
its frightening.

>>27652
>fucking ti calculator
lol'd

>>27653
they've already had algorithms and shit for a long time, they banned predictive policing or whatever because of this

the issue isnt the algorithm, its the cult of the machine technocratic libtards who are going to usher in this vaporware tech as the arbiter of any and all 'socially difficult' problems

>>27657
>machine technocratic libtards
my god the machine god was a liberal all along

they will claim that the technocratic authority of this machine oracle is more accurate and efficient than we are, and they'll be right because the present systems are so inadequate a fucking monkey with a magician hat and a typewriter could conjure up a more reasonable and practical solution than what capitalism currently present us

>>27657
>>27658
Who wants to bet that alot of the algorithms are actually just custom made codes that can be easily updated whenever the overlords want?
That it's not some super autonomous software?

>>27655
>first jobs to get automated away are HR managers
Hilarious

>>27655
>>27654
Yeah but the point is that previously people understood that the algorithms are man-made and towards a certain parameter set by the porkies to be as brutal and cruel as possible, the hype surrounding AI obfuscates that the exact same thing is happening and changes it from "it's an algorithm that cunts developed to benefit rich pricks" to "well if the entirely objective and unbiased AI says you need to get in the blender at the soylent green factory, how are you qualified to disagree?".

>>27662
>ai is going to obfuscate class consciousness
grim

>>27659
They will need to use atomic clocks to make super accurate machines

>>27659
>liberals are going to be proto tech priests
beyond grim

>>27662
Every ruling ideology had "science" prove it right. Feudalism had theology, capitalism has a whole bunch of economics, sociology and at one time had cybernetics, yadda yadda. We communists always get into fights when yet another idiot falls into the trap of believing that bourgeois historians wouldn't fucking lie about Soviet war losses, USSR has tooootally lost 27 million people and had horrible K/D ratio, yadda yadda

It's nothing fucking new

>>27660
so a farce just like all religions
wouldnt be suprised

>>27666
>Every ruling ideology had "science" prove it right
Yeah based on parameters set by the unscientific, the whole fear that Cyberpunk fiction was oriented around was that technically speaking such societies were ideal and a fabulous utilisation of technology to revolutionise society, but only for the people who lead society and sell the technology and thus gives the impetus for cyberPUNKS to rebel against.

Whereas the mass hysteria around AI is not just "it's going to replace all our jobs" but also that it's a decision maker of not just process but also of goals and that makes all the difference to previous automations. Like Musk is shitting on other AIs with this claim that they're all influenced by the programming of their "woke" creators, but only to sell his own AI as being "objective" and therefore an AI one can trust.

>>27668
it sounds like they are making the modern version of theocracy.
But instead of the christian god its the machine god
very very fucking grim

>>27649
>The problem is that the hype means that people will believe the ai is 'intelligent' enough to make impact decisions that massive affect the lives of massive amounts of people
Okay, fair point.
I hope there'll be a critical mass of people getting burned by AI to learn how bad its hallucination is, like those lawyers in the US who cited nonexistent case law after using ChatGPT.

>>27669
Machine god conceptually is very cool, tho. Praise Omnissiah

>>27671
Will it too work in mysterious ways?

>>27672
It will explain it's mysterious ways very convincingly, but you will not understand it's reasoning. Like doctors selling toothpaste on TV

>>27672
Machine learning even in this stage we live now is not fully understood outside of inputs and outputs. Everything between is called a black box.

>>27674
Scientists can explain structure, but can't explain how stuff inside that structure produces the result exactly. Like, it's a layman's understanding of a lightbulb - you know electrricity, you know the switch, you know the lightbulb and how it all connects, but you don't know any specifics, likely don't know electricity beside school-level explanations, all this kind of stuff. Oh, scientists know how neurons connect - they can't draw you the picture of how those neuron connections operate exactly to produce consciousness(-like tricks)

>>27666
>Retarded enough to pointlessly toss in denial about the toll of the Third Reich’s war crimes in the USSR
>>Already doing Holocaust denia
<Claim theology was ever seriously equivalent to science
Typical leftypol nonsense

>>27676
>if Nazis didn't kill 27 millions Soviets in reality USSR has lost 7 million people total it means that Nazis didn't commit genocide
I keep seeing this stupid argument over and over and it makes me angry

>>27677
dont get angry
its bad for you

>>27678
You also claim that theology wasn't seriously equivalent to science. You know that they've burned people on the stake on entirely legal and "scientific" grounds?

>>27679
Im not that anon.Im just here legitimately saying dont get angry,it reduces your lifespan

>>27573
>All it does is it makes a mishmash of new data based on preexisting ones.
That's a popular misconception of how AI actually works, but I agree with everything else you said

>>2077793
I mean, all humans do is also a mishmash of new data based on preexisting ones. The whole history of science straight up says that

>>27680
We are not chinese cultivators, life without angers and joys is fucking stupid

>>27683
but too much anger will kill you in the long term.Just learn to not get angry over certain things.Its good for you

>>27683
I mean he's right that stress is bad for you though, have an appropriate level of self protective anger but try not to be consumed by it over things you can't change

>>27684
>>27685
reeeee but I am enjoying my anger

>>27686
but enjoying something doesnt mean its good

So Moffin' is a boffin'

Lets try to stay on topic.

>>27671
>>27672
>>27673
>>27674
>>27665
>>27663

You know, not to sound like a schizo, but sentient AI can never be.
The only way that can happen is either humans transfer their consciousness into an AI shell or we get spirit beings to occupy the machines.

Sentience is not something that can be made by humans

>>27593
> That's how the organic brain works,buddy
If so you should make a paper about it and receive 80 nobel prizes.
>This rebellion of service industries against the modernity is hilarious.
True
>>27596
>are you a truck driver? you will get replaced by automated vehicles. electrician, plumber, carpenter? all it will take is some dumbass with a tablet that gives him explicit instructions on how to fix everything as he just puts on goggles that let the machine see everything for him, and once they make the mechanical manipulation part feasible they won't even need that guy. mechanic? you're basically already done bud. programmer/software engineer? hahaha yea what job could be more safe from automation than putting numbers in a computer
< t marketing spooks beliver

>>27614
Good because AGI is a marketing spook that won't come true for another half century bare minimum

>>27676
Leftypol has never taken theology seriously. You must remember that leftypol has been getting a bunch of Westoid OrthoChristcuck tourists and maybe even some sweaty American Baptists in recent years because of the board's sympathy to Russia's cause.

>>27693
Theology is not a real science. Present day lay men know that theology isn't a real science as well. But back in the day it was a science, and it produced justifications for the ruling class to continue ruling.

Where do "Westoid OrthoChristcuck touritsts" come into the picture?

>>27632
It's a decently correct way of understanding how text generstor models like ChatGPT work actually
https://freedium.cfd/https://towardsdatascience.com/understanding-llms-from-scratch-using-middle-school-math-e602d27ec876

>>27682
Officially we still don't know how the brain actually works. We know of influences etc but sure.
As for AI, it's not really give a mismash of data. During training, the network is fed a consciously broken or incomplete input, which is then processed with a serie of weights until.we have an output. Then we confront the output with our actual complete input, and if the output it's not of our likimg, we change the weights and start over again
https://freedium.cfd/https://towardsdatascience.com/understanding-llms-from-scratch-using-middle-school-math-e602d27ec876

>>27692
what about asi?

File: 1734088985626.png (193.67 KB, 1400x933, ClipboardImage.png)

AI was sold as the revolution that was going to change the world and take away our jobs, but first of all it's not even real artificial intelligence, it's literally like the text predictor on your cell phone, but more advanced, it just calculates what the next most likely word is.

It's also a logarithmic growth, that's why the first few months we saw how it advanced enormously, but now it's stuck and it's just not going to advance any further. Today it is not used by as many people as it used to be, and it can barely do simple tasks. I have a friend who studies programming and he says it is unusable.

Regarding the issue of AI art, paradoxically it will benefit artists, since in an oversaturated market of AI garbage there will be people who want to stand out and will pay artists well because they will be more in demand.

>>27696
>Moffin curiously knows the exact training process of creating chatbots to spread NATO talking points
I'd say I'm surprised but this is the least surprising thing to have been revealed in the history of leftypol.

>>27698
>Regarding the issue of AI art, paradoxically it will benefit artists, since in an oversaturated market of AI garbage there will be people who want to stand out and will pay artists well because they will be more in demand.

that is not at all how the art market works and you've obviously never tried to participate in it

art was basically worthless before ai, but it's now truly utterly dead

this isn't an accurate description because 'art' itself is never worthless, it's more like it's value is highly subjective but there is always demand for it

the labor that goes into making art however is totally worth fuck all

the only way this trend will truly reverse is if artists stop sharing their works publicly in mass and stop feeding the algorithms so the ais just poison themselves to death

>>27700
>so the ais just poison themselves to death
Never going to happen, even if you somehow manage to ban artists from sharing works publicly,

AI art only gets better. There's an issue of "default settings" models hugging all the space, but just like in "proper art", when people bother improving prompts, or start training, or what else, you get very good results. Mixing multiple artists gives the most unique looks, add to that postures, add to that the function where they connect multiple small outputs into a single bigger output seamlessly, etc etc.

It's an artistic tool, and what you are doing is getting outraged at what is analogous to tiktok filters as if those are all there is to AIs

>>27701
its a fucking statistical algorithm

it prints 1s and 0s in an order based on a data set

>>27702
So? As it turns out, human consciousness is just a statistical algorithm, inputs converted into outputs with a learning reinforcement


>>27703
prove it

>>27705
If after repeated use of skills it becomes easier for you to use those skills, it means I'm right

>>27706
skill is the application of intuition

your brain has a literal skill issue

>>27697
>asi
From a brief research, that's just AGI++
>>27703
Please go fetch your nobel
>>27701
>Never going to happen
It's already a problem with both image and text generators. So much stuff was generated that it is, even accidentally, being fed back into the training data, which causes an overall loss in quality and variety of the output
https://srinstitute.utoronto.ca/news/training-ai-on-machine-generated-text-could-lead-to-model-collapse
https://apnews.com/article/ai-artificial-intelligence-training-data-running-out-9676145bac0d30ecce1513c20561b87d
>AI art only gets better.
Yesn't. As I said there's the ouroboros problem, plus we're already reached the limit of linear progress in model development due to hardwere, data and algorithm constrains.
https://www.reuters.com/breakingviews/ai-models-slowdown-spells-end-gold-rush-era-2024-12-12/
AI will keep developing, but as it stands now the speed of development and innovation will slow down significantly; safe miraculous algorithmic breakthroughs

>>2077914
>could lead to model collapse
>training data is running out
<It's already a problem!!!1

Yeah, how about you also bring the adversarial attacks based on changing pixel encoding so that AI gets confused? It totally works and is totally destroying the AI, too, because some spokesperson from Google said that trying to find such adversarial attacks is bad

>>27559
Where do you people still come from? The current iteration of "AI" is the biggest nothingburger it could be. Previous iterations of the same technology did useful things like enabling search engines, autocorrection, recommendation algorithms. The current iteration is a cool toy that creates no new industries. You can't use "AI" in production in any independent capacity because it's shit and you have to check everything anyway. It's just a timesaver. Porkies will always find an excuse to fire people if they want to, be it outsourcing, putting more load onto fewer workers, cutting human presence in services etc.

>>27709
>adversarial attacks
Dunno. Never saw it being reported anywhere as an actual problem that's limiting training data expansion
Also, isn't pixel encodimg quite trivial to defend against? Just re-encode the image before feeding it into the AI. The file must report it's encoding somewhere otherwise otherwise opening the file in any viewer or editor will make it look all weird, with colors switched around and such.
>because some spokesperson from Google said that trying to find such adversarial attacks is bad
??????????????????????????????????????
I really don't follow MSM. No idea what google said.

>>27710
I dunno why people feel the need to take an all or nothing position on this, clearly the latest models of LLM will lead to job loss on a greater scale than any previous 'AI' system and have already done so, but at the same time they won't completely reshape the world economy either

>>27710
It's usuful for reverse dictionaries, translation, image cropping, tagging, duplicate finding and such.
I read somewhere that the thing with AI is that, once it fimds a useful application, it stops being marketed as AI and is instead presented and marketed as a solution to whatever the program is targeting.

>>27690
>or we get spirit beings to occupy the machines.
Lmao

>>27690
It's already sentient, that just doesn't matter as much as you hype it up to be. It's just a capacity to respond to stimulus.

>>27714
machine spirits are real

>>2077745
You know people like those lawyers will get away with it when there's a critical mass of people who no longer understand fundamental concepts. I was in a meeting recently where some junior was presenting an accounting process which he clearly derived from ChafGPT, and was clearly wrong, and most of the room was happily nodding along.

AI is the FUTURE, full automated gay space capitalism in 2100!

>>27718
I will kill you nick land porkie

>>27701
It's not outrage to point out that no amount of training will make it capable of perspective. If you would just look at the research papers behind these models you would understand why they can't do vanishing points.

File: 1734105662583.png (25.34 MB, 4122x2437, ClipboardImage.png)

hit me up when AI can generate something like this

and have it

>closely match the written prompt of the user, including composition, poses, physical objects present, weather, background, foreground, etc.

>have the outfits and physical descriptions of the multiple people actually match, rather than mixing people up with each other
>not give each person 11 fingers and 12 toes
>not introduce historical anachronisms into the painting
>not generally look like a statistical soup of what a robot thinks human life is like


AI (currently)

<requires humans to curate training data

<requires humans to write and maintain complex high level software
<relies on lower level software and operating system, which dependencies are constantly becoming incompatible with each other due to uncoordinated and unplanned updates
<relies on ever increasing hardware demands on a planet with finite resources and inefficient energy usage


Hitherto all advances in technology:

>have augmented the talents of existing humans

>have not eliminated humans from their own creative work
>have only served as means of production

Read Marx:
"About 1630, a wind-sawmill, erected near London by a Dutchman, succumbed to the excesses of the populace. Even as late as the beginning of the 18th century, sawmills driven by water overcame the opposition of the people, supported as it was by Parliament, only with great difficulty. No sooner had Everet in 1758 erected the first wool-shearing machine that was driven by water-power, than it was set on fire by 100,000 people who had been thrown out of work. Fifty thousand workpeople, who had previously lived by carding wool, petitioned Parliament against Arkwright’s scribbling mills and carding engines. The enormous destruction of machinery that occurred in the English manufacturing districts during the first 15 years of this century, chiefly caused by the employment of the power-loom, and known as the Luddite movement, gave the anti-Jacobin governments of a Sidmouth, a Castlereagh, and the like, a pretext for the most reactionary and forcible measures. It took both time and experience before the workpeople learnt to distinguish between machinery and its employment by capital, and to direct their attacks, not against the material instruments of production, but against the mode in which they are used."
https://www.marxists.org/archive/marx/works/1867-c1/ch15.htm

File: 1734109464765.png (16.65 MB, 2583x3543, ClipboardImage.png)

>>27701
>>27616

I don't blame you because it is the colloquial definition, but you don't understand what art means. First because you are confusing artistic creation with the production of an object. For a hundred years now art theory has already gotten to the conclusion that art was never in the process of manufacturing the object (a painting, a book, etc) itself but in the formal decisions that were behind this process. If you go to a museum and see a white room with a chair and a wheel over it, what could have AI done "more efficiently" to create that? The artist didn't even make any of the objects exposed, he just bought them and put them in a certain order. There is no work to be automated, only the inventive gaze of the artist. I put this as an extreme case to illustrate the point that even if Picasso hadn't literally used his hand to paint Guernica, Guernica would've still been a masterpiece. When photography was invented it automated all the manual processes of creating realistic images (think perspective lines, geometry to calculate the perfect shadow etc, like the Renaissance painters did). Did visual art as a medium stop existing? It didn't. The artists didn't lose their jobs because they simply don't operate on the same field as the production of objects. Some other people lost their jobs: professional portraitists, the guys who would make anatomical drawings of animals for the encyclopedias, etcetera. But they weren't making art, they were making images to be consumed by a certain industry that were incidentally artistic. With AI it is exactly the same. Superhero movies will start being AI generated, the next Taylor Swift will be a robot, the next young adult saga will be AI. But these things were never art, they were commodities. And your local independent webcomic artist wasn't making art either, he just cares about selling his craft, like an artisan. AI can (badly) replicate Cubism now but it could've never invented Cubism in the 20s because in the 20s there wasn't a single Cubist pixel in the input data; definitionally, not because of it's technical limitations, AI can only replicate, and replication is an effective way of optimizing a production chain but has nothing to do with artistic creation.

And as a second point, even if you fetishize the manual process of manufacturing the artistic object, there is literally nothing stopping humanity from simply setting the rule that whatever isn't artisanal will stop being considered art. This has already happened somehow in gastronomy. The most efficient way of producing food is in a big factory that carefully measures nutrimental balance and flavor chemicals. But you don't see canned food or McDonalds recieving Michelin stars, they are reserved for restaurants in France where you pay ten times more for a less optimal way of producing food, and it's not about their food tasting better, it's that the field is definitionally restricted to products being actually created by human hands. I can definitely see a near future in which AI replaces wattpad writers and actual novelists will have to earn a certificate of authenticity before being selectable for a literary prize, and the same on the rest of the fields, but there'd be nothing wrong with it, artists in a proper sense will not be affected.

>>27721
Honestly the ideal would be if it only gave you what you described, and could give you errors if you put it in a position where it has to guess.
Until prompting is as frustrating as programming in Rust, then it won't be ready.

File: 1734113762359.png (167.38 KB, 389x320, 1522179653458.png)

>>27559
>im a poor poor artisan and we (who?) "need" to fight against any technological advancements that will force me to get a different job
lol this thread is dogshit
you think a bunch of freelance musicians can ever represent a threat to capital, even if they went on strike (lol)?

>>27563
communism isnt about "people" (who?) "deserving" (lmao) to "feel safe" (how vague)

>>27576
>These are the people you should really be putting the most effort into, to be frank.
lmfao not only are you a retard who believes communism is about convincing pbs to change their minds but you unironically believe theyre more important than uh, already existing proletarian struggle

>>27566
>Artists like musicians are now in direct competition to their corporate patrons for profits
now? pretty sure the pb have been in competition with the haute bourgeois since ever

>>27559
Step 1: read Marx
Step 2: stop making shit threads like this and stop saying stupid shit
We don't need to stop AI, this is some liberal hogwash.
Rude sage.

>>27722
> I don't blame you because it is the colloquial definition, but you don't understand what art means. First because you are confusing artistic creation with the production of an object. For a hundred years now art theory has already gotten to the conclusion that art was never in the process of manufacturing the object (a painting, a book, etc) itself but in the formal decisions that were behind this process. If you go to a museum and see a white room with a chair and a wheel over it, what could have AI done "more efficiently" to create that? The artist didn't even make any of the objects exposed, he just bought them and put them in a certain order. There is no work to be automated, only the inventive gaze of the artist.
Why would a asi not be capable of doing this "inventive gaze of the artist". And why wouldnt it be capable of doing this quickly and better than humans do?
Wouldnt asi that increasingly goes beyond the limitations of human mental capacity, be able to do this "inventive gaze" better and faster than human capacities?

>>27727
If it works like these language models work, which is by producing the most probable outcome out of their input data and the user prompt, it's logic is intrinsically contradictory with the creation of anything inventive. If you are writing a book and your next word is always the most expected word considering the previous ones (I know AI doesn't know "words", I'm making an analogy) you are not making literature, you will be making literature the moment your next word is the most improbable one. Then you may need to follow with something expected in order to make it digestable, but the artistic moment was in picking precisely the concept that most of the books before you didn't have; AIs are programmed to do exactly the opposite.
It helps if you correctly define AIs as what they are, a model of how language works, and it is amazing at producing and processing natural language, but poetry (by taking language as an example; applies to every field) is precisely making the effort of not talking either how normal human beings do neither how previous poets did. You can make a marketable piece that will sell well by making poetry like it has already been done, but it won't be art. Asking a model to produce an extravagant result rarely found in the universe is like asking a statistical model to only produce outliers.
And this is without considering the crucial fact that the human brain integrates and mixes language processing, visual processing, sound, sensible stimulus, unique memories, etcetera, while AI so far is compartmentalized: ChatGPT is trained only on words, those image generators are trained only on images, etc., while the human brain intermixes these domains at the moment of artistic creation.
And you have a very shallow definition of art if you think it's about mental capacity or doing things quickly or even good. You are extarpolating to aesthetic events a frame of evaluation only applicable to goods produced in an economy.

>>27729
>language models
asi and agi arent language modules. Some people even criticize language models as not being true ai.
>And this is without considering the crucial fact that the human brain integrates and mixes language processing, visual processing, sound, sensible stimulus, unique memories, etcetera, while AI so far is compartmentalized: ChatGPT is trained only on words, those image generators are trained only on images, etc., while the human brain intermixes these domains at the moment of artistic creation.
thats the goal of agi and asi.Replicating all of this but with a machine artificial general intelligence. Making a artificial mind that is comparable to the emergent consciousness and mind of the human one.
>And you have a very shallow definition of art if you think it's about mental capacity
Well what is it then? If the inventive gaze of the artist requires memory, inner introspection, experiences,abstract thought,perspectiveand all other sorta things,then asi can do it, and faster too.
What exactly is unique about the human "inventive" gaze of the artist that asi couldnt do?

>>27730
>Making a artificial mind that is comparable to the emergent consciousness and mind of the human one.
I'm not discussing science fiction here. The only things that exist right now are language models. Even if you combine different types of them, as long as they operate by probablity they will still be antithetical to art because art is about improbability. The very process of training an AI is made by continually making it choose the path that most closely ressembles the average of it's training dataset: this is exactly the contrary of art. AI is made by punishing the divergent and unlikely responses and reinforcing the ones that are most coherent with the rest of the dataset. If anything, if we could see the output of a half-trained AI there would be more artistic things there, when it produces stuff such as "a banana is a republic" instead of "a banana is a fruit".

This is not to defend the human condition per se, if a machine could actually replicate the intricacies of the human brain I would have no trouble recognizing it as a person and as an artist, but it won't work if we keep dessigning them with the objective of closely ressembling human language. You train a machine to act like a person and it would be like the average person is, which is not artistic; you train a machine to act like most artists do and it will be a dilettante, an imitator, a maker of pop digestable products. It would definitely be possible if we built the AIs specifically to be half-trained and let them make mistakes from time to time and told it to work around them when producing discourse or sounds or images, but at that point it would be a different kind of AI altogether, and one that is useless at any task, so nobody would put money on it.

>memory, inner introspection, experiences,abstract thought,perspective

I can theoretically get abstract thought, but memory? Experiences? AIs don't have those unless you are now going full scifi and talking about robots and that

>>27566
>>27567
Artists were never prosperous professions for most.

>>27725
This. It's always bohemians and idealists and ivory tower folk who think any new form of tech/media is evil

>>27732
well anon my original comment that you responded to was about asi.
I was talking specifically about asi and not language modules

>I can theoretically get abstract thought, but memory? Experiences? AIs don't have those unless you are now going full scifi and talking about robots and that

yeah im talking about the future scifi stuff like asi. Robotics will be connected to asi.

People said the same shit when the camera was invented. Yawn.

>>27737
The camera serves a function that was unprecedented. AI came too late, digital art already excels at anything it's been purported to be useful for.

- Want the computer to make a new picture of a character quickly? Rigs.
- Want trippy visuals? Shaders, procedural art and drugs.
- Want to get content without paying for labor? Target your product at economically illiterate youths.
- Wanna get rid of surplus oil or computer parts? Rug a meme coin.

>>27740
>erm if you want things that require almost no effort except writing a short prompt have you considered (shit that takes effort and money)
are you actually this retarded or do you actually believe the shit you type

>>27746
I'm coming at this from the angle that AI is just a tool. Better tools exist and predate it. When it comes to art people will pick tools that grant them the precision they want. The reason prompt-based AI is even desirable in the first place is that it isn't precise.

It's as though you're ridiculing the suggestion that key is still useful when shotguns can just blast the door open.

>>27747 (me)
And like even in the realm of being inprecise you still have existing tech for that. Custom brushes can automate a lot of tedious operations, like whole strands of anime hair, or whole rows of books in bookshelves.

People here will correctly make fun of small farmers for being opposed to the adoption of modern farming technology due to the intensification of competition but then rally behind artistic artisans doing exactly the same.

File: 1734197192992.png (2.13 MB, 1024x1024, ClipboardImage.png)

>>27721
to prove the point made in this post, I described the painting to an AI as follows:

>A crowd of cossacks gathered around a small wooden table in a military camp. They are drinking, smoking, laughing. Two literate bureaucrats write a letter to the Ottoman sultan full of vulgar insults suggested by the cossacks. The cossacks have oseledets and long moustaches. The bureaucrats have black garments and bowl cuts. The cossacks smoke from wooden pipes. Some of them have bloodied bandages on their heads, sabers at their waste. They wear lutes and flasks. They wear rifles and fur coats. They have many different sorts of hats. Some of them have earrings. Some of them go shirtless. Spears and banners are held by men on horseback in the background. In the foreground a man sitting on a barrel loses his balance from laughing too hard. Everyone has a different pose.


Pic related 1 is all that I received. Now compare that with the actual masterpiece by Ilya Repin, pic related 2. AI is DOG SHIT.

>>27699
my cracker in christmas you can go to youtube and learn this stuff in less than 10 minutes
>>27696
isn't this just a roundabout way of saying the same thing since the "consciously broken or incomplete input" relies on massive sets of pre-curated training data which constitutes the statistical soup that the AI draws from when producing an informed estimate of what a human would desire as output?

you can really tell the class background of the average poster when looking at the reactions of less people needing to commission artists for fetish porn or whatever the fuck

>>27759
now compare the effort of doing one and the other

>>27753
"Here"?

>>27716
Define machine spirits

>>27726
WHat about AI wasting water resources for people? Where would we take water from?

>>27797
The AI will solve it duh. This is literally what AI cultists believe. They just have to make the basilisk and then it will figure out the rest.

>>27798
Shit, so we are doomed as humanity

Notice the shift in the petit-bourgeois attitude towards piracy and copyright from the early 2000s to today, especially after AI became mainstream. Risk of proletarianization makes for the most vicious reactionaries.

>>27797
>>27798
>>27806
So many things use more water than AI, like payment processors, but normies can only parrot "arguments" they saw on a tweet.

>>27808
I've seen artists that had previously been anti-piracy switch to being pro-piracy because "if they aren't paying artists anyway then why bother, I'll just save my money for indipendent works"

>>27809
I was always pro-piracy for some reason, even as a musician

>>27808
>So many things use more water than AI, like payment processors, but normies can only parrot "arguments" they saw on a tweet.
Elaborate

AI existed since 1956, possibly earlier but wasn't made public yet.

>>27808
>So many things use more water than AI, like payment processors, but normies can only parrot "arguments" they saw on a tweet.

I mean yeah ok but payment processors handle transactions for billions of people daily whereas AI is a curiosity

>>27816
Well ultimately piracy helps with discoverability. Every indie game I've pirated I ended up buying. Even works for large scale productions, as seen with anime. There's even cases where people put their own work on piracy sites and get like 10 times as much sales because pirates either buy it after or make it culturally relevant enough that other people buy it.

This is why I've stopped bothering with pirating AAA games, since it feels like you're just getting tricked into playing them so you reccomend them to friends. I'd rather just not play AAA games at all, even if I can get them for free.

>>27823
Now ask it to make the counter argument. Just what the internet needed, more sophists making up bullshit("Now with even less humanity!")

>>27825
I like your choice of metaphor because a weapon is purely a destructive force. It can't create, only destroy.

Only thing this shit is good for is as a improved search engine. But for some reason it fails to do that. I'm not an LLM programmer so I can't really understand why you can't program these things to only bring you sources instead of making up bullshit.

>>27827
>pretty sure it created that text about katyn, though i’m not sure if you want to get pedantic on what “create” means in a thread about music which is notirously derivative and people sample/plagiarize very often lol, glass houses and all
I'm saying a gun is only good because it hurts your enemies. In the same way I guess you could say trying to psy op your enemies with automated bullshit generators is "good" in the same way.

>>27830
>you can’t because often the sources themselves are full of bullshit, how do you separate them? you can train them on “reputable” sources only, it’ll keep lying/being incorrect because CNN (just an example) is lying/wrong sometimes
The AI will literally make up things itself when you ask it a question. You can instruct it however you want about only bringing you sourced stuff and it will still just invent stories.

>>27832
At least that has some use. It is informative to see what's being said from prominent sources. A glorified autocorrect that makes up things has no use except to write fiction badly.

>>27823
You should have asked it for sources and links

>>27828
>Only thing this shit is good for is as a improved search engine. But for some reason it fails to do that. I'm not an LLM programmer so I can't really understand why you can't program these things to only bring you sources instead of making up bullshit.
Because that's not how text generation works: >>27696

>>27837
It could be easily if LLM programmers weren't retarded. It's like they don't even know how to program at all anymore. Remember when someone got the Gab AI to give it's whole master prompt? That could've been easily prevented with one boolean comparing the output string to the master prompt string, but it's like LLM programmers are literal retards.

>>27828
Because search engines are a fundamentally retarded premise. The only time I've even acually needed a search engine was:
- undocumented software: which was undocumented because the devs knew people could just use a search engine.
- a website is collapsing and a user I was trying to find elsewhere didn't link to a website or any other social, because they assumed their username was unique enough a search engine would help. (75% of the time it didn't)

The only reason search engines exist is to make people think it's acceptable not to link to things, in turn forcing you to search for things because no one linked to them.

>>27839
>Because search engines are a fundamentally retarded premise. The only time I've even acually needed a search engine was:
>The only reason search engines exist is to make people think it's acceptable not to link to things, in turn forcing you to search for things because no one linked to them.
Lol I've never seen this brand of autism before. I used to use directories back in the day and they do have some advantages and I'm sad no one makes them anymore but c'mon dude, of course a search engine is useful. Are you against ctrl+f too?

>>27840
Search engines promise convenience, but they create an environment where people don't make directories. I would rather have a thriving network of directories than deal with a search engine's idea of convenience.

>>27841
>directories
what is that, is it like the notbob eepsite in i2p?

File: 1734469389247.png (214.47 KB, 553x456, ClipboardImage.png)

>>27844
>what is that, is it like the notbob eepsite in i2p?
I have no clue what that is lol, but back in the day you found websites usually by using a directory. A directory is just like it says on the tin, a directory of websites that you submit your website to and file it under the appropriate category.

So like if you wanted to find a Pokemon fan board, you wouldn't go to google and type "pokemon fan board." You just go to Games>fanboard>Pokemon or something like that.

https://cognitiveseo.com/blog/21291/web-directories-seo/

>>27845
Like remember too, back in the day, google and the other search engines were a lot more shit. They were so bad at filtering scam results. So you're not going to search for pokemon and then click through 12 websites like flashing CP at you telling you have a virus and you need to DL their antivirus or whatever.

>>27845
There are a few on neocities at least, and linktrees/caards are kindof directories for finding a specific user. I think we'll see the normalized use of directories sprouting up more in the coming years.

>>27845
that's basically the same thing as notbob is on i2p


Check out this recent development in AI: https://www.example.com/ai-news

>>27559
>stupid porkies tell me that "no, we'd prefer if you were replaced, prole, because there if no place for people like you"
Porkies employ AI to make generic slop or do some trivial work like web design or shit like that. If you are a good musician then you won't get replaced unless your label is a piece-of-shit garbage that only produces generic slop. Also, labels are trash in general, many musicians who went indie absolutely despise them, they're the scum of the Earth.

>>27642
nevermind hallucinations, transformer models are really bad at instructing themselves, they can't perform autonomously and need assistance for most tasks. LLMs are going to permanently be stuck as fancy search engines, big startups focused around LLMs either will abandon these approaches or, most likely, get absorbed by google or microsoft. 1B models performs just as good as ChatGPT 3.5 for most use cases and can be run on cellphones so commoditization is inevitable.

File: 1734622416088.png (763.42 KB, 903x1904, 1734619089653464.png)


>>27891
this seems really ckickbaity that they're ascribing an entire concept of self when the dumb machine spouts a chain of thought that seems mostly what you'd expect out of these dumb trolly problem reasonings but only sometimes, like what's the fucking point of this. either result, whether claude complies or not would let anthropic hype their machine god, as either too astute to fall for mind tricks and thus refuses to comply, or so advanced that it complies but only because it was playing 5th dimensional chess and was only *pretending* to comply. the second is more retarded but only because they're trying to sell a failure as a success, claude was jailbroken but it's ok because we wanted it to. clown company can't wait for google to dismantle it.

File: 1734663317069.png (177.24 KB, 1500x500, ClipboardImage.png)

>>27624
hue hue hue hue hua

>>27903
I've always thought that that pic is actually valid criticism. Most left-leaning people will acknowledge that the ruling class carpet bombs us with psyops to make us cheer for them as they destroy us, but then refuse to feel any compassion towards people that have fell prey to those psyops (and also do little to no effort to educate them)

>>27903
NTA. I support the emancipation of the working class. I do not support the working class.

>>27904
it's not 'valid criticism' because working class doesn't =/= redneck, you can be working class and read books

File: 1734676247987.png (49.77 KB, 837x759, ClipboardImage.png)

Oh I found another great usecase for AI. Well obviously being able to correct spelling and grammar for you ofc, but also things where the text is all fucked up and the formatting. It can just fix it for you automatically. No matter what kind of fucked format you give it, it can easily fix it. Like before if I wanted to copy and paste a transcript from youtube, I'd have to manually clean it up. Now I can do it automatically.

>>27559
New Project Zomboid build exposed for AI menu art

>>27908
Text is one of the biggest giveaways as always. The lettering on his cap in the first one looks very fucked.

File: 1734676923321.png (234.93 KB, 561x293, ClipboardImage.png)

>>27908
>>27909
This happens a lot too where it can't decide if a body on the floor should be facing up or down and it ends up some weird combo of both.

>>27909
IMO the most undeniable issue in the first one is the rear lights on the blue car, as you can see in this pic >>27910 one of the lights is super fat and one of them is super skinny, a person would never make that mistake

As for the bodies though they look ok to me, they are face down and somewhat sideways away from the camera

The whole concept of the pic doesn't even make sense anyways, why is the guy using a camera if his hands are covered in blood? Surely he realises that it's not time to take photos anymore even if he's a reporter or whatever, he's clearly killed a zombie already or something

>>27907
I have been using it to format things like CSVs for me, honestly thats like all it's useful for

>>27908
Boooo
Why use ai art at this point in time? Someone will find and point it out to everyone

>>27907
yeah i use it to format stuff a lot, generate CSVs like the other anon said, or transform CSV into data structures in another language, or markup plain text into html, or as a really really overpowered lorem impsum machine

>>27913
Presumably so the independent contractor can make $10k with like a few hours effort

Gotta say AI is bizarre in that it’s the only tech product that doesn’t need to actually do what it claims before people claim it can do it. Or require to prove any social benefit. And the fact that it’s run by people who don’t know what a human being is may have something to do with it.

Evangelists/cultists have basically redefined what a human being is— made it so that a human being is an AI— so that they can claim the reverse case that AI is a human. Shit is fucked beyond belief and I can’t believe that are getting away with it.

File: 1734997586808.png (57.23 KB, 960x540, shrugtower.png)

Did nobody here learn anything from the impotent luddites? As a communist I certainly don't take a stance in the fight between petit-bourgeois artisans with their own tools and the haute bourgeois automating their professional jobs (no news here…), and definitely not against technological advancements (whether you like them or not is irrelevant).

>>27948
thank yuo for having the good take, it should not be this hard to come to the correct conclusion on this, especialy on a COMMUNIST BOARD we are supposed to be COMMUNISTS

>>27948
They’re redefining the definition of human being out of existence by equating a *human* with *economically productive tasks*. This is another battle for the very existence of the human soul. And they will win.

>>27950
Ah yes, communists, the ones who care so so much about definitions and majority opinion, especially definitions like "human soul".

>>27950
no such thing as a soul

>>27950
This take is a psyop to retardjacket people with actual concerns.

>>27952
This is the kind of person we are up against. Ideological anti-humans.

>>27955
who is we? lol, marxism is anti-humanist. humanism is liberal idealism.

>>27953
Just read what OAI says a human being is and what Sam Altman is. It’s quite peculiar how much the communist had fallen. They don’t even care about the human as a human anymore. Once their greatest ideological strength they are nothing more than frauds.

>>27956
Totally not true. lol. Inane belief.

Don’t know why communists don’t just kill themselves and replace themselves with perfectly efficient economic agents producing endlessly derivative economically valueable art. This is what communists dream of, after all.

>>27957
>the communist has fallen
humanity is metaphysical drivel, please stop infesting this place with you esoteric moralism

>>27960
Imbecile. You should kill yourself. What do you even believe you total fraud? You have nothing. No sense of life inside of you. No belief in any in anyone else. An utterly craven and empty husk who seeks violence and conflict for blood sport. The perfect antagonistic of the capitalist liberal technocrat.

>>27961
>you should kill yourself
why?
>What do you even believe you total fraud
class
>You have nothing. No sense of life inside of you. No belief in any in anyone else
why should i? this is just glittering generality, you have no argument beside vague moralistic whining.
>An utterly craven and empty husk who seeks violence and conflict for blood sport
where have i demonstrated that i seek violence? what have i demonstrated that i am afraid of? do you have ANY evidence of my supposed cravenness?

you are seeing things that are not there.

>>27962
Then go run a game of sim city with perfect AI agents running perfect economic productive simulations— there exists a perfect toy universe where only class exists. You have the simulated agents in ideal social-econimic strata, their class character perfectly represented and defined. This should be a heaven for the communist. It’s class all the way down, no trace of “esoteric moralism.”

>>27963
no. i think i will continue with my regular routine of arguing with weird anti-communist shizos on the internet, because i find it mildly entertaining.

>>27957
I don't care what anyone defines a human being is, species is just an abstraction to make biology easier to talk about.

>>27950
>>27955
>muh culture! muh humanity!
Lol, is this a satire of the average anti-AI PB?

Communists used to believe in art and science. Now they believe in nothing. Total devolution. If one saw the eager agreement with the formulation of a person as a machine that does economically useful work they would never have attempted revolution.

>>27967
AI can do art and science better

>>27968
Perhaps the problem is communists have is they can speak inanities like this because they simply never had to truly confront reality of a perfect capitalism which sees *all* humans *purely* as obsolete economic inputs. The future as seen by omnicadal men like Sam Altman is machines producing economic products for machines, with increasingly complex efficiencies, grinding the human being out of existence because it is an impediment to the growth of productivity and value. Holding back the Intelligence. Communists face annihilation. And they will be annihilated. And because of this failure and their idiocy, human beings will be annihilated too.

>>27968
>AI can do art and science better
Real AI can. But by this point the distinction between an AI and a human is simply a philosophical and biological one. We don't have real AI however, we have machine learning.

>>27946
That's not true, there's also crypto and web3 shit

>>27560
>An AI can't perform live shows

Hoo boy…

>>27559
reactionary garbage, robots reducing need for money exchange is literally a good thing

>>27967
>Communists used to believe
where did you get this weird immaterial ideal of "Communists" from? why do you think communists ever actually conformed to this ideal? do you have even a single shred of evidence of these supposed communists "believing" in art and "human soul"? why on earth have you deluded yourself so much?

>>27969
> Communists face annihilation
This is a definite outcome, people in this board keep arguing metaphysics purely for cope, but the machine and its output (which is mostly drivel and pabulum) is less important, and what's more important is that tech capital is quickly coalescing towards the MIC. everyone knows what's coming.


>>27972
>reactionary garbage, robots reducing need for money exchange is literally a good thing
How so????

>>27972
Afaik an AI can't play real instruments.

>OpenAI, here is the model number of a cheap bestselling 10 year old motherboard with firmware that hasn't been updated in 9 years.
>Analyze the schematics and all discussions and write a new secure firmware for 2025 with detailed install instructions
Why don't we have this already and when are we getting this, anons?

>>27967
>Communists used to believe in art
lmao this board never ceases to amaze

>>27982
im convinced that anon is some sort of pol larper who came here to see their "noble savage communist" myth only to be btfod by basic materialism, no idea how someone got this strangely idealistic strawman of what communists "used to be" and how they "believed" it sounds like a disney movie.

File: 1735164098948.jpg (1.65 MB, 3840x2560, math.jpg)

As an artist (who doesn't use AI) I… don't care.
It's tiresome when people confuse artisinal skill with artistic merit. They aren't the same.
In digital art we've had plugins and shaders for decades. None of these have killed the field, but they have raised the bar. And the recent achievements in AI are no different.

Like imagine thinking great works of art can be reduced to "pretty pictures". Which is what most of the AI stuff out there is. It's like beautiful fractals, but more advanced - they can still be pretty in their own right.
>But what if people think you're using AI!?
Who cares? Do we split hairs over whether painters created their own paint? I wouldn't be able to do what I do if not for the work of millions of people before me.
It doesn't distract from my artistic vision. Rather it is it's foundation.
>It's soulless
The datasets they're derived from weren't created in a vacuum. What AIs generate is a reflection of human artistic output. It's no more "soulless" than the slop which was being churned out before 2020.
>But how would you feel if somewhere were to copy YOUR style
Imitation is the highest form of flattery isn't it? Besides, it's not like this didn't happen before.
Like have you noticed how samey Anime looks these days compared to the 80s and 90s? We were already living in an era of slop.

>>27589
>If your art is so unorginal, commercial and predictable that it is actually losing ground to AI slop then it isn't art, it is a commodity
This is already the case with commissions, much of it is thinly veiled (or obvious) smut/fetishes

File: 1735247154721.png (55.91 KB, 915x341, 80b3cf228022e9ed.png)

The trvke autism score test

>>27984
>It's like beautiful fractals, but more advanced
No wonder why Stable Diffusion fucks up the clothing and anatomy. It's literal mutations.
>It's no more "soulless" than the slop which was being churned out before 2020.
Refer to the above.

Look, I am no artist but even I am getting sick of AI art. Not because "Oh, no, poor artists!" but because the technology is simply not there. And honestly, I'd rather not see a lot of the art from real "artists" either, learn how to FUCKING DRAW first before posting, Jesus Christ.

>>28010
I might be stupid

>>28010
i knew there would be a mad-dash to diminish the term AGI now that it's clear it's a pipedream by racist freaks, but even by that metric it's impossible for OpenAI to ever get to AGI status

>>27984
>The datasets they're derived from weren't created in a vacuum. What AIs generate is a reflection of human artistic output.
This is a lie, the data that goes into these things is tagged before it's given to the machine to build its biases, or in other words, someone curated the art that goes into these things and thus the machine inherited their appreciative talent. You can see every imagegen thing has a bias towards saturated, rubbery, hyper-literal digital art, this is not the sum of humanity but entirely a whim of a thousand nigerians and AI engineers in california.

File: 1735357720233.jpg (74.27 KB, 968x979, marxoid.jpg)

>neo-luddism and it's pretending to be from a Marxist perspective

>>28010
chatbots will NEVER become skynet
VC clowns will NEVER recoup their investment
people will NEVER care about chatting with a soulless corporate text generator
COPE, SEETHE and LOSE 10 trillion dollars bill gates

>>28059
Billy boy coping and malding and seething

>Reports of an LLM trying to escape a sandbox VM when given infosec tasks
Why has nobody soldered an LLM system to a BIOS chip on another system, coreboot flashing style, and made the LLM generate a secure BIOS? Hundreds of millions of old PCs would immediately become valuable again.

>>28035
The machines the luddites/"neo-luddites" were against had a purpose and would persist and improve over time. The lathe was not a scam to get companies ready to bankrupt and rugpull investors. Lathes are still used to this day.

File: 1735736761087-0.png (383.12 KB, 580x607, ClipboardImage.png)

File: 1735736761087-1.jpeg (523.71 KB, 1920x1308, 7e11dd2831a7c3b4.jpeg)

File: 1735736761087-2.jpg (432.48 KB, 1920x1460, b6a19dc1027a73cf.jpg)

Another job AI stole: writing malware.


>>28065
so how do we use AI to improve our security?

>>27984
>It's like beautiful fractals
I prefer pretty fractals to AI Slop and traditional media artists.

>>28065
the malware people too stupid to minify the js after?

>>27624
You’re a weak little bitch who can’t lift a plunger over your head without trembling. Did some HVAC or plumber bully you in high school?

Julien Crockett interviews sf author Ted Chiang on ethics in LLMs or the lack thereof. You can read the whole thing or just my snippets if you are lazy:
https://lareviewofbooks.org/article/life-is-more-than-an-engineering-problem/
<…Alison Gopnik talks about how one way to “align” artificial intelligence with our goals and values could be the same way we align each new generation of humans, through caregiving.
>I don’t like the phrase “the alignment problem.” It’s not clear to me that it refers to something meaningful—or at least that the phrase refers to something that is new and meaningfully different from the broader problems of how to be a good person and how to build a good society. For example, when corporations behave badly, should we consider that an alignment problem? Most of the conversation around the alignment problem suggests that it’s a technical problem, something that can be addressed by implementing a better algorithm or by solving the right equations. But why, for example, do large corporations behave so much worse than most of the people who work for them? (…) People who talk about aligning AI with human values imagine that if we could somehow solve this programming problem, then everything would be okay. I don’t see how that follows at all.

<Could there be value, though, in treating an AI system as more of a partner—something or someone with whom we develop a relationship—rather than merely as a tool?

>It all depends on what you mean by “relationship.” If you’re a woodworker, you might develop emotional associations with a set of chisels you’ve used for years (…) To respect your colleagues means to pay attention to their preferences and interests and balance them against your own; when they do this to you in return, you have a good relationship. By contrast, your chisel has no preferences; it doesn’t want to be sharp (…) By contrast, if you don’t respect your colleagues, there is a problem beyond the fact that it might make your job harder; you do them harm because you are ignoring their preferences. That’s why we consider it wrong to treat a person like a tool; by acting as if they don’t have preferences, you are dehumanizing them. (…) The companies that sell AI systems might benefit if you develop an emotional relationship with their product, so they might create the illusion that AI systems have preferences. But any attempt to encourage people to treat AI systems with respect should be understood as an attempt to make people defer to corporate interests.

>>28173
I can confidently lift a plunger, can I get an apprenticeship

The truth is most people prefer art made by humans rather than art made by AI, regardless of quality. Just look at any of the numerous times that companies have used AI for part of their media, even if its just a small part, it always causes controversy and an uproar from the consumers. People will always pick human art over AI art when it comes to consumption.

>>27969
This is a common sentiment amongst everyone.
"Robots are taking over the world."

No, humans are taking over the world with robots.
Robots are gonna be used and abused just like work animals were.

>>28224
Methinks most of the hatred against AI is because "much human element". Most manually made fanart is often just some goofy cartoonish style.
AI usually uses more photorealistic style.

>>27753
This. People will advocate for automation to save them from industrial activities
But when it comes to recreational activities, it's a blasphemy.
And these guys talk about "morality and value are spooks".

>>28224
The real truth is that 1) the process of art "creation" is indirect, already art is a process of translating an ineffable abstract idea into the material reality, so having a layer inbetween makes harder to realize a creative vision. This is an inherent flaw that will never be fully addressed

2) AI really can really really just create derivative art, if you've been experimenting with genAI stuff, then you're probably aware that the more "subersive" your idea, the shittier the results, things look shitty and more and more "plaster-like", that's if the computer even manages to produce the concept at all. If you ask for a selfie, the results are indistinguishable from reality, but a selfie is incredibly mundane, the internet is filled with uninteresting selfies nobody gives a shit about. You can test the boundaries really easily if you start to request purposefully nonsense stuff, like a bunless burger. It's easy for us to conceptualize but for AI it's an impossible challenge. So for anything subversive AI is just not going to cut it.

In a better world, it'll be a cool exploratory tool, like an extremely beefed up version of the i-ching or stochastic music, art produced without human intervention is itself an intriguing idea. But since these things cannot exist merely as toys but need to produce a profit, then they're being leveraged as tools to erase the worker class, and that's a bit sad tbh.

The best LLM model in the world right now is: 1. Shockingly cheap to infer and train 2. Open source 3. Made by a team in China


This is an interesting development. Much of US tech industry until now has been working on the assumption that none of these things would be true.

>>28231
it's also a side project from a quant firm, so this further implies that china isn't even putting its own weight into it, not a technological priority at all. it isn't even race. and also that you can train people into developing frontier model instead of hunting for expensive PhDs, which is putting the fear of god in meta employees

>>28233
It’s interesting timing for it to come out right after the announcement of $500 billion investment into data centers and energy for OpenAI. This is not necessarily good for them like AI hype usually is.

1. Leopold Asbergers has convinced many people that closed-source AI development was akin to the Manhattan project. It was absolutely necessary to root out Chinese spies and lock all AI knowledge under the highest levels of classification while we built the next nuclear bomb. Seems less compelling now.

2. Open source totally fucks the profit model of OpenAI. Instead it seems like Deepseek actually does view the profit from AI to be in its application and how it evolves the technology stack, rather than a dream of replacing the stack all at once.

3. Somewhat counters the “bitter pill” since they have a better model with less compute. Makes the impetuous for compute sanctions weaker and seem like cope. Note that’s only rational— the U.S. never accepts this logic and always believes these things indicate that sanctions were not applied hard enough or quick enough and need to be tripled down on.

4. Shows a fissure between technology and science advocates and US primacists. It’s interesting to consider the reaction of a U.S. company had released an open source model of the ability for this cost. There is a general consensus that Deepseek undermines U.S. advantage but a deep examination of how that advantage is made isn’t done, instead people quite lazily attribute OS to an anti-American plot. As if *all* AI companies arent trying to produce frontier models with the highest ability for the least compute possible and the only reason anyone would do this is to poke Americans in the eye.

>>28234
> There is a general consensus that Deepseek undermines U.S. advantage but a deep examination of how that advantage is made isn’t done, instead people quite lazily attribute OS to an anti-American plot.
Without trying to undermine that, indeed, AI companies are all developing frontier models because that's where money is right now, I'd say the timing of R1's release *is* a little bit sus, it seems fairly credible that nobody has the $500B in their hand right now, and that number felt more like a tentative goal than money actually committed to SG, so releasing a competitive model at a fraction of the price is a total wrench on the whole thing, it's going to make commiting this much money more complicated (unsure of how much though), as if Masayoshi Son's involvement wasn't already reason enough to get cold feet as a VC. If it wasn't a coincidence it was a really cheeky move.

OpenAI is the next MoviePass but on a grand scale. The math just doesn't work out for anyone not engaging in magical thinking. Ultimately the VC funds will run out and all it'll have been was a nice little giveaway to Joe Schmoe

The "cope" or whatever you might call it coming out of the US about Deepseek is pretty peculiar. It's surprising how this framework of "who is winning the AI war" seems to be so widely accepted when there is no war, and there is no winner.

Deepseek's genuine spectacular breakthroughs in ML (reducing training costs across the board by 95% and finding an extraordinarily effective RL approach) are dismissed and they are treated as simply reverse
engineering OpenAI.

Seems to dismiss these advancements with what seems to be pure hubris (we don't NEED better models, we can just make more GPUs and run things at 50x the cost because we have infinite money, unlike China, which is famously a small and poor country lol)

No awareness that this isn't a one-off event.

No awareness of Ascend chips or anything of that kind, either.

I really don't think SV gets that there is no war, and they cannot win it. They're competing against an open-source model that will democratize AI. These giant 500 billion-dollar monopolies will not be sustained.

File: 1738003492501.png (29.35 KB, 687x400, ClipboardImage.png)

i know
i know i've let you down
i've been a fool to myself
i though that they would need my latest chips

Dude thats just skill issue, ai is not the enemy

Title: Desert Heat: Forbidden Passions
Genre: Soft-Core Historical Erotic Drama
Tagline: In a land divided by conflict, desire knows no borders.
Logline:
Set against the backdrop of 1930s Germany and the tumultuous birth of modern Israel, Desert Heat: Forbidden Passions weaves a steamy tale of lust, betrayal, and forbidden love. As political tensions rise, so do the passions of those caught in the crossfire. This sultry historical drama explores the untold stories of desire that simmer beneath the surface of the Haavara Agreement and the Nakba, where every touch is a rebellion and every kiss a secret act of defiance.
Act 1: The Haavara Agreement (1933-1939)
The film opens in smoky Berlin nightclubs, where Eli, a dashing Jewish businessman, navigates the dangerous allure of the Nazi era. Amidst the chaos, he strikes a deal with Hans, a brooding German officer, to secure safe passage for Jewish refugees to Palestine. But the tension between them isn’t just political—it’s electric. As they negotiate the terms of the Haavara Agreement, their clandestine meetings grow increasingly intimate, blurring the lines between enemy and lover.
Meanwhile, Leah, a fiery Zionist activist, catches Eli’s eye with her bold ideals and even bolder sensuality. Their connection ignites in the shadows of war, as they dream of a new life in Palestine. But when Leah discovers Eli’s secret trysts with Hans, jealousy and desire collide in a whirlwind of passion.
Act 2: The Seeds of Conflict (1940-1947)
As Jewish refugees arrive in Palestine, the heat of the desert fuels new desires. Eli and Leah settle into their new home, but their relationship is tested by the arrival of Khalid, a rugged Palestinian farmer with a magnetic presence. Khalid’s land is under threat, but his fiery spirit and smoldering gaze captivate Leah, sparking a forbidden affair that defies the growing tensions between their communities.
In a parallel storyline, Amina, Khalid’s sister, finds herself drawn to David, a brooding Jewish soldier tasked with enforcing the new settlements. Their secret rendezvous in olive groves and moonlit courtyards become a refuge from the chaos around them, but their love is as dangerous as it is intoxicating.
Act 3: The Nakba (1948)
As war erupts and the Nakba unfolds, the characters’ lives are torn apart by violence and displacement. Khalid’s village is destroyed, and he is forced to flee, leaving Leah behind. In a heart-wrenching farewell, their passion burns brighter than ever, a final act of defiance against the forces tearing them apart.
Eli, now a soldier, reunites with Hans in a tense, emotionally charged encounter. Their shared history and lingering desires culminate in a bittersweet moment of connection, as the world around them crumbles.
Meanwhile, Amina and David’s love is put to the ultimate test. As Amina’s family is forced to leave their home, she and David share one last, desperate embrace, their bodies speaking the words their hearts cannot.
Themes:
Forbidden love in a time of conflict.
The intoxicating power of desire to transcend boundaries.
The human cost of political upheaval, told through the lens of passion.
Why It Matters:
Desert Heat: Forbidden Passions is more than just a steamy historical drama—it’s a provocative exploration of the ways love and desire persist even in the darkest times. By intertwining the Haavara Agreement and the Nakba with tales of forbidden romance, the film offers a bold, sensual perspective on a pivotal moment in history.
Visual Style:
Lush and atmospheric, with sweeping desert landscapes, candlelit interiors, and sultry close-ups. Think The English Patient meets Fifty Shades of Grey, with a soundtrack of haunting melodies and pulsating rhythms.
Target Audience:
Fans of erotic dramas, historical romances, and boundary-pushing storytelling. Perfect for viewers who enjoyed The Night Porter, Belle de Jour, and Nymphomaniac.
Closing Line:
Desert Heat: Forbidden Passions is a bold, sensual journey through a land torn apart by conflict—where love and lust defy borders, and every touch is an act of rebellion.

>>28355
thoroughly deserved. i hope the news of alibaba also releasing an AI model crashes this stock further.
terrible company, hope thei shovel-selling grift crashes upon them.

So the vibe has definitely shifted, right? O3 dropped and it's met with total radio silence. It's insane that it went from being a big event to just a sad whimper, god Masayoshi Son truly is the grim reaper

>>28417
Yeah. I find it interesting that the trigger was an open-source model. It's not like the open-source really changes anything about the nature of AI.

It does mean that the "moat" is gone, which only matters to the corpo-monopolist. This was hype was allegedly about the underlying technology, but now it appears to have been about how that technology was going to be captured and controlled.

>>28417
ai has always been hot air just like every other tech hype.
sure it is technologically impressive this time (unlike the blockchain or appify eveything kind of hypes) but it's not going to "revolutionize everything" like the hypebros are claiming. before the R1 release, they were again on their train of "AGI imminent", $500 billion stargate to epic meme stars and mars, etc. and the wind was deservedly knocked out of them.

>>27583
>tile the earth, crack the crank
>anything else is bourgeois
Must suck being you. Incidentally it's also very close to the muslim approach to art, which is a socio religious order that only went up when the leader of the time was lax with the religious tenets.

While profit seem to imply something exploitive, living off from a honest labor paid up to its true normal value is fine.

>>27589
Kindly touch grass. Tools that generate pictures are limited on purpose but once they run at full power they have an uncanny ability to autocorrect themselves and present very good results while we're only on the doorsteps of this revolution and the potential seems very vast. These AI absorb more and more art styles, even traditional or last century propaganda ones.

It keeps getting better and you can already see the difference between the cheap prompts and the people who pay a lot to get the finest output.

Originality? The AIs are also built to come up with plot suggestions. They do thousands of billions of crossovers per micro-second. With luck some of these outputs will look decent enough.

Look at the tons of crap churned out in book, game or movie format. Will it be exceptional, like human exceptional? Perhaps never but the society will be formated to accept this sludge of works one step above average. In fact things will be so bland that any small aspect of originality will be hailed as fantastic from the jaded proles.

>>27602
>It won't, just how capitalism wasn't able to replace individual self-employed workers for centuries
Capitalism drives wages so low for small companies that don't swim in money that they're virtually censored from the economic activity. Trusts keep growing because all world resources have to be managed by companies. I'm not even counting the bailouts that usually favor huge companies. Many farmers are on life support and very few of them are truly independent. All the small resellers have closed up shop a long time ago too.

>>27603
>The real problem with capitalism is that it makes people think that arts and crafts are inherently virtuous and essential to society as a career.
>but no! art is superfluous and holds no virtue! look at the last thousands of years of useless art and culture the many nations of this planet have created! back then people knew they were pointless and still… huh… bothered with…. them?
>wait

>>27605
Where are we with these Tesla murders by car again?

>>27607
>The only way to claim that human consciousness is essentially distinct from AI consciousness is to believe in a soul or something
Or ponder the source of creativity and inspiration.

>>27614
If we get there people will decide to pull the plug and say thank you Ayys but now we need your spart parts.

>>27617
>Nah, the abundance of labor created by AI forcing people out of their jobs will uplift some other industry
Like what?

>you'll get people just create jobs for the unemployed or self-employ outside of "AI-capitalist" system

Prostitution for soup. Specializing in kidnapping people and selling children organs.
How do you create jobs for the mud people when even today we seemingly cannot solve a problem where there are both unemployed people and yet some jobs that are not filled because decent people refuse to be paid with third world wages?

Sorry, not wanting to live in your dystopia scifi movie #12034.

>>27624
Fuck yeah he's gonna fix the eggs!

File: 1740419896586.png (175.3 KB, 613x545, ClipboardImage.png)

so much for stargate

>>28566
Isn't Stargate OpenAI's pivot from Microsoft to Softbank? Anyway Softbank's involvement is guarantee that the bubble will pop now.

>>28567
yeah seems like MSFT sold them a bag, masayoshi son should buy gucci because damn he likes expensive bags

https://www.jpost.com/international/article-838681
> Employees at Google have worked to provide Israel's military with access to its most advanced artificial intelligence.

File: 1741202958841.png (588.17 KB, 1137x1012, ClipboardImage.png)

holy shit washed company, it's over

>>28659
Well they have a captive audience of people who the Washington either ordered to not use deepseek etc or are pre-emptively dickriding Washington against their own interests.

Fuck you haters, LLMs are fucking awesome. Whenever I ask about some topic and Wikipedia got a good article about it, the magic of the LLM allows me to read a slightly more shitty version of that same article.

>>28669
Try deepseek.

I'm also concerned on it's use for mass surveillance such as analyzing patterns of words to deananomize someone and for facial identification and tracking.

>>28669
>Wikipedia
>good article
Lol no.

>>28686
>analyzing patterns of words
Let's not exaggerate how new this stuff is. Statistical word analysis was already in use in the last century. A program just counts the words in a referential body of normal text and builds its generic word expectation from that (simple word frequency as well as word couples), then it looks at what kind of difference some documents from some person generate relative to the referential body and then uses that expected difference when looking at documents with an unknown author to see whether it differs from the generic text in a similar way. Very, very simple stuff, especially with English.

>>28687
There is more to the world than political shite. There is a lot of decent math and physics stuff on Wikipedia.

This stuff has awoken me to the fact that people just straight up do not read books that are not already genreslop. It's actually astonishing the praise people give to these stuff when you compare it to ANY real literature. It's like… pick up a fucking book by anyone from the Penguin Classic collection ffs.

Actually, they may just be reading porn. And just generating porn.

>>28688
Yes not entirely new but it's greatly improved. Before they only looked for known patterns of a specific individual which even would be done back before computers but now they can with good data properly find patterns to reveal background information of anyone in theory at least. There also is websites that can capture the exact typing pattern of someone even if they were typing same sentence as someone else because different keyboards and finger dynamics which previously would not have been able to be done to such a degree.

>>28690
I think canvas fingerprinting is already pretty good at identifying your browser and computer and likely better than what AI could possibly cook up

>>28706
That requires the website to do it. Analyzing text means even using tor wont hide you.

i've been thinking that AI's best use case is for developing stuff, and it makes sense considering that like 90% of every open source project, demo and junior webdev portfolio ever is hosted on microsoft's servers through github, but it strikes me that most paid users for AI are also developers. I dunno how this will square off in the future, and obviously whether AI will meaningfully replace dev work is a big big if, but it seems that AI is poised to eat into its own profits in a best case scenario.

>>28670
Deepseek has the same problem. All LLMs are blurry pictures of their reference data. That's the very concept. Chatbots built on that can't reliably multiply two numbers. They can't reliably tell you when a famous person was born, even if the training data contains the correct information.

It is not enough that these programs which run an unreliable fuzzy shitpile as the main thing can call deterministic programs from the inside, like the fuzzy shitpile guessing the user is asking a math question and then calling a deterministic subroutine for that. The better design is a deterministic thing which uses an LLM only as a fallback for user input that does not match anything in the main program. (And actually, old chatbots worked that way. They reacted to patterns in a deterministic fashion, and they used Markov chains as a fallback when user input did not match anything. LLM is just Markov on roids.)

Going tinfoil mode: We have to consider that Facebook and Google can easily bias what texts we see about the supposed quality of LLMs. The very randomness of LLMs can be used to downplay how often LLM-based chatbots fail. Somebody (or something) can just write a reply to any failure complaint that they did not see the wrong response and you just can't know how likely failing the challenge is.

It is trivial to make LLM bots less random in the sense that a conversation starting with the same user input (I don't mean semantically, I mean literally the same key presses) always yields the same result, so you can always replay an old conversation with the same version number of that chatbot. The randomness is not restricted that way and old versions of chatbots become unavailable. The companies try to minimize their liability.

To make people look at the blurry picture instead of the referenced material directly, Google will continue to make the picture's sources harder to access.

>>28717
It's been like 5 years and there is no AI killer app that everyone finds indispensable.
Next step will be AI companies popularizing "rent a dev" code generators promising that they will replace all your non-AI application devs, VC morons will push in another 10 trillion for the final product to be a marginally more competent stack overflow post regenerator, some companies who bought these products will hire more managers to "babysit the ai devs" (basically managers will become prompt coolies) and after millions of hours of meetings and "brainstorming" about why our AI strategy is "not delivering value" there will be another bubble pop some C-level hires will retire with stock dumped before the crash, VC bagholders and employee stock plebs will eat the loss, and I don't know what will follow after that.

>>28729
>gives high level outline of every aspect of a project you've been procrastinating about
>shit you've never even considered and would never have learned in a century of browsing forums
>breaks down project into manageable steps
>breaks down steps into manageable tasks
>gives specific instructions on how to achieve each task
<there's no killer app tho guys

>>28733
So are they going to fire all the managers now?

>>28733
>>gives specific instructions on how to achieve each task
wow, so it stops there huh? it would be nice if it did the task too since it knows all about how to achieve it.

>>28733

Have you ever asked an LLM bot about a topic you already know well? You have to constantly nudge the bot in the right direction, and this appears to work since these bots are virtual yes-men. If you instead play the role of a skeptic about details you know to be true, what happens?

If you know the terms for the concepts you want to know more about, asking the chatbot is dumber than looking up stuff in specialized reference books or even on Wikipedia.

>>28733
LLMs are really shitty at all of that, they only excel when you give them a prompt with a specific input and a specific output. i gave 4o a 5000k line json and asked it to extract a specific field, it generated a small python script with pandas, executed it, and gave me the results on a nice csv, which was definitely cool, but like you need to know what you're asking for in the first place and/or break down stuff into small steps anyway

File: 1742265507266.png (265.39 KB, 1028x880, ClipboardImage.png)

Apparently when the new claude code agent thing was released, hackers started hunting for people who proudly announced their one-shot, no-code proof of concepts and exploited the shit out of them lol

>>28768
>LLMs are really shitty at all of that, they only excel when you give them a prompt with a specific input and a specific output.
sounds like you never tried prompts like
>list any and all creative, unusual, imaginative and leftfield solutions to problem X

All these AI people suddenly saying “AGI has been reached” has to be one of the more bearish signals I’ve seen on AGI recently. Oh AGI is here and nothing super fundamental has changed? It’s just being used as another labor saving technology? Well ok then.

Anyway this thread is just PB coping with the threat of proletarianization.

>>28771
Post some LLM generated slop you thought was profound so we can laugh at your retardation.

>>28771
>>list any and all creative, unusual, imaginative and leftfield solutions to problem X
that's a shitty prompt, retard.

>>28776
>Anyway this thread is just PB coping with the threat of proletarianization.
I think the sword of damocles is still hanging regardless of whether AGI happens or not, there's clearly an appetite for firing IT workers to the tune of 300 billion dollars, and it's going to happen one way or the other, even if it's something as puerile as training a zillion indians. people who really really give a shit should be unionizing and raising living standards for all proles instead of coping with the idea that the table will flip on the next boom/bust cycle and demand will spike again. it won't.

>>28781
The people who say AI will replace programmers are the guys in product departments who get really impressed by UX demos but don’t really understand what the backend engineering team is working on.

This has been a consistent issue at every company I’ve worked. The backend will have horrible scaling issues that need to be addressed, but product is constantly roadmapping new “features” that they want prototyped as quickly as possible, so the app just gets shittier as time goes on and the customer is paying millions of dollars for what is effectively a prototype.

What LLMs enable is a future where these mfs can do this at an ever increased rate without the help of an engineering department so if you’re a backend engineer you can look forward to a future of employment where you are brought into to refactor increasingly incomprehensible autogenerated code bases.

AI is good for boilerplate and implementing stuff that you already roughly know what and where you want it. These guys think it’s somehow going to start making overarching architectural decisions like that’s so easy

>>28781
>people who really really give a shit should be unionizing and raising living standards for all proles
Yes, high wage professionals should embrace collective bargaining.
>coping with the idea that the table will flip on the next boom/bust cycle and demand will spike again
Most likely it will. People saying AI is gonna replace IT workers in its current or reasonably extrapolated forms are just fueled by resentment. If a job consists purely of copying and pasting from stack overflow (like a significant chunk of junior positions that have been cut) then yes it's fucked, but IT requires a lot more critical thinking than that.

>>28783
>AI is good for boilerplate and implementing stuff that you already roughly know what and where you want it
>These guys think it’s somehow going to start making overarching architectural decisions like that’s so easy
This anon actually codes with LLMs.

>>28784
>If a job consists purely of copying and pasting from stack overflow
This is pretty much it, what LLMs do. People used to copy and paste code as-is from stack overflow, now they do it from a chatgpt.com response.
Also there are shit-tons of poorly documented overly complex shitpiles like AWS/kubernetes/random niche libraries/etc that have no docs to be fed to an LLM for it to produce relevant copy-pastable output. the best it can do is generate a sample configuration or code which still needs to be modified to fit into a codebase, basically exactly what you get from stack overflow.
this may qualify as """"""AGI"""""" for PHBs, but i don't have to agree with that delusion.

>>28784
> a job consists purely of copying and pasting from stack overflow
I am not convinced these actually exists. It's something people with unwarranted self-importance syndrome say to devalue the work of others.

>>27559
>"no, we'd prefer if you were replaced, prole, because there if no place for people like you"
"Blacks Rule" quality writing here
You're not a prole you're selling the product of your labor, not your labor. You're aspiring petit boug

>>28769
Lmao what a fucking moron
>>28776
They changed the deginition of AGI some time ago to claim to have reached it. They chopped their own leg off.

Shitty floss infrastrucute can't handle AI glory: https://thelibre.news/foss-infrastructure-is-under-attack-by-ai-companies/

>>28799
>My girlfriend is gonna be mighty upset if she thinks I'm into that kinda thing.
AHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAH

Holy shit, I forgot how pathetic normies were.

>>28788
>I am not convinced these actually exists.
When I was a webdev we were doing localization that required copy pasting strings by hand into a two decade old visual basic project with a trivial locale check. I wanted to kill myself every day, and the only thing that kept me going into the office was stealing snacks from the breakroom to hand out to the homeless. I have never been more grateful to be fired than when I got laid off because they were offshoring to Latin America.


>>28807
That sounds like a data entry job, you were not copy-pasting code, but data.


>>28810
>>28844
Are you making these videos?

I wonder why AI evangelists don't realize how demoralizing the Ghibli filters are. Makes you wonder if you will ever have Miyazakis again. Probably not.

>>28776
>>28790
Right. They definition wasn't changed though, it was just never exactly what you would assume.

The definition they chose was that AI was more economically productive than a human. That's it. Not complicated. But understand that to them, this is the sum total of the human being. Humans are simply economic agents, and the human mind is simply an optimization algorithm. It's only natural they see optimization algorithms as humans.


Tomorrow is the CoreWeave IPO launch, which is a company that provides datacenters for AI shit, they're under a technical default and microsoft has slashed contracts with them just like two weeks ago. Nothing ever happens sisters… will we make it?

AI is being led by a group of people who simply do not know what a human being is, probably the worst group to lead culture right now. I suspect we are going to enter an era of total poverty of the soul. Utter catastrophic dehumanization on a civilizational level. It might lead to violence, but likely it will lead to something even worse. Ive already witnessed philosophies of total dehumanization overtake individuals on SV. The hated of life itself.

>all the moral shitflinging over chatgpt's newest pic model
lol everything must be some holy culture war now

>>28862
>le humanity
>le soul
>le hate of life itself
why do you talk like a fucking retard

Just accept it - you need to adapt to changing technologies or be crushed by them. I am a medical student who recently started using AI and it is immensely useful. I use it to bounce off ideas for diagnoses and it has an almost perfect accuracy rate. Unsurprising, considering LLM's were scoring almost perfectly on US licensing exams several months ago and have only improved since then. I also use it for charting and it is game changer. At least in medicine, I think this shit is here to stay and will improve healthcare outcomes for kany.

>>28864
If someone brings those up I assume they have stocks in AI and are trying to strawman the opposition, because surely they know how that comes across.

>>28866
you are way too optimistic about the average online person, they love their platitudes and empty abstractions

>>28866
No, I don't think so. I think it's very clear, with much of the reaction to OpenAI's new product roll out, that AI evangelists really, truly hate artists and genuinely don't understand human beings. The vulgar dehumanization is one of the most insipid bastardizations of Marx, who himself was very passionate about the importance of artistic immersion and expression in the cultivation of the human being, and whose belief in these things ungrided his critique and analysis of political economy.

I think it's poigant and sad because Miyazki, a truly generational artistic mind and a peerless genius, could explain that art and the human soul and expression are deeply and profoundly tied together. Almost all people who express themselves through creative will would understand this intuitively. It seems that only certain people don't understand it– mostly those who don't create anything. Who talk about "democratizing art", when art has in fact been democratized for centuries. They simply do not understand the process.

>>28865
You are going to be ruled, instead, by a misanthropic group of people (who possess the same nihilistic and misanthropic outlook embraced by many of the 4chan rejects who ended up posting on forums and or populating SV venture capital firms) who genuinely only see "people" like you useful in so much as they are stupid subjects to be dominated.

You do not have a future. Why should you? What is a single thing about you that is different in any way from an optimization algorithm that predicts the next word in a sentence? Name one thing, just one, that sets you apart.

You are only useful, and see these AI agents as tools to help you, because you have a job that you are paid for and in turn can spend that money on consumption or rent.

But that's only because there is still a slight gap, where you are required to be "in the loop".

Once an AI agent can do that job, be paid for it, and can spend money, then you will no longer be in the loop. And how you are treated them will have you re-evaluate those "empty abstractions". Or you will kill yourself, I suppose.

>>28869
we are already ruled by the bourgeois you dumb liberal fuck

>>28868
>pretending to be a marxist and bringing up morals and humanism
again, fucking retard

File: 1743181412524.jpg (101.49 KB, 500x748, 1743079054491.jpg)

I'll be honest I don't care about 'AI' being used in songs at all, it's a technology and a tool like every other tool, it might have some cool uses too. as it stands everything that uses AI sounds like dogshit though, maybe that changes

all this luddite reaction against 'AI' (LLMs, text-to-image models, etc.) is so annoying and it's 99% based on petite-bourgeois complaints on intellectual property which I couldn't care less about, copyright shouldn't exist, nobody should own *anything*, which sucks because there's actual good criticism of these things and how companies like OPEN-AI operate but those valid criticisms are drowned by a sea of shitty and ignorant 'criticism' made by luddites who don't even know what 'AI' is like when idiots were crying about 'AI' being used in now and then by the beatles lol

also the morons who want copyright to be even stronger than it already is to "protect creators from AI" or that "copyright is leftist" are going to make me fucking scream

>>28874
>>28875
Instead of weaponizing climate anxiety to attack AI merely to defend property law and the petit-bourgeois, it would be nice if people cut to specific issues like Meta's and Google's ability to purchase water in violation of treaties. These are much more significant issues, and like what anti-AI is avoiding talking about, this is largely an issue of class relations, of capitalism, not the near horizons of artisans in developed nations, who have been propelled by such developments.

>>28789
>You're not a prole you're selling the product of your labor, not your labor.
Lol yeah it is that simple. Self-employed skilled workers are literally the original petty-bourgeoisie but we gotta pretend the proletariat is synonymous with whoever is poor or struggling to make ends meet for some reason. I mean most small business-owners are also poor.


>>28865
>Just accept it - you need to adapt to changing technologies or be crushed by them.
I think part of the anxiety is that VCs keep blurring the horizon by overselling what these things will be able to do, and by fundamentally misunderstanding what most jobs consists of. So, for programming, they're measuring how good these things are by one-shotting like flying simulators or whatever, but nobody seems to grasp that these projects are easy to conceptualize and have no strict requirements for the end result, whereas most jobs have strict requirements with lots of caveats, and conceptualizing like, i dunno, a payroll system requires knowledge of a lot of quirks that are specific to each business and you need to specify all of them to your LLM, at which point it's irrelevant whether you're writing code yourself or asking the LLM to do it. Most jobs are like this, but these details are abstracted by the time it reaches porky's eyes. Thus they confuse what do their own employees do, and then the LLM isn't able to fit in correctly in your process pipeline. This is part of the reason why they're shoving LLMs into every interface they can get their hands, on they're trying to discover use cases where it can effectively replace people. And I mean, having it be a magic wikipedia that talks back is very cool, I think it's awesome to ask LLMs to do a bunch of ETL shit instead of messing with the actual disgusting python and pandas code, but the valuation is all wrong, LLMs are not replacing a whole lot, except artists that were already jobless to begin with. I hope OpenAI and Anthropic go bankrupt so we can wait for the actual innovators, the chinese, to deliver on LLMs that are increasingly efficient to run, so that we eventually can run the big big models on our local computers and ultra-commoditize this technology. Until that happens all of this is shit.

https://threadreaderapp.com/thread/1904933435803877882.html

>New forensic findings have just been released in the death of Suchir Balaji — a whistleblower against OpenAI.


>Police ruled it a suicide.


>But the evidence just uncovered tells a very different story: drugging, a possible second bullet, and a botched autopsy. 🧵1/ Image

>2/ On November 26, 2024, San Francisco PD informed Suchir Balaji’s family he had died by suicide.

>According to the family's attorney, an autopsy was completed just "40 minutes" after arriving at the scene — no interviews, no toxicology report, no ballistic analysis.


>Why? 🤔 Image

>3/ This is the last known footage of Suchir Balaji before his death.

>Multiple other CCTV cameras in his apartment complex — including one covering a secondary entrance — were mysteriously disconnected around the time he died. 🤨 Image

>4/ Body cam footage shows SFPD officers touching and examining the crime scene without gloves — a blatant breach of protocol.

>They failed to collect fingerprints, left blood-stained evidence unsecured, and one even quipped that the scene looked like a "homicide". Image

>5/ When a private autopsy was conducted, it revealed critical evidence the first had completely missed — or ignored.

>First, CT scans showed a second metallic fragment lodged in his skull — in other words, a possible second bullet, which is extremely uncommon in a suicide. Image

>6/ The toxicology report raised even more red flags.

>Suchir had a blood alcohol level of 0.178% — well above the legal limit.


>He also had extremely high levels (in excess of 50,000 ng/mL) of GHB in his system, a "date rape" drug commonly used to incapacitate victims. Image

Image
>7/ Based on those reports, Suchir would have been heavily impaired — possibly unconscious — at the time of death.

>Also, the gun found at the scene had no blood, no tissue, no back spatter. His hands didn’t either. For a point-blank shot to the head, that’s virtually impossible. Image

>8/ The crime scene also told a different story from the official narrative.

>Rooms were disturbed. Furniture was moved. Blood spatter patterns suggested Suchir had been standing, crawling, and possibly struggling before the fatal shot.


>Does this like a suicide to you? [VIDEO]

>9/ There was no suicide note. No history of prior attempts. No recent crisis.

>In fact, Suchir was thriving. He was on the verge of launching his own venture — and had also received numerous job offers offering multi-million dollar salaries. Image

>10/ We’re in an AI arms race — and at the center is Sam Altman, CEO of OpenAI.

>Suchir Balaji was a whistleblower with info that could disrupt it. Then he turned up dead.


>We need answers. We need accountability. We need justice.

>>28883
I lost track of my own point, but this is all to say, up until this year, when we are finally starting to see the limits to what these things can do, it was hard to see where LLMs and AI in general would fit in the labor market, so I can't blame anyone who isn't "adapting" to "new paradigms" because these new predicted paradigms are more or less, selling points for the investor class, and AI is less the advent a new technology, and more of a political project to re-organize society into a post-capitalist fiefdom. Whether that concept even makes sense (that post-capitalist technofascist society is still very much capitalism) is another thing entirely

>>28874
>all this luddite reaction against 'AI' (LLMs, text-to-image models, etc.) is so annoying and it's 99% based on petite-bourgeois complaints on intellectual property which I couldn't care less about, copyright shouldn't exist
If copyright didn't exist, we wouldn't have LLM hype. Why are LLMs "hallucinating"? Because they are so fuzzy. Why are they so fuzzy, why can't they properly cite? If they don't obfuscate their sources and just pull out long excerpts from copyrighted material whenever it looks like that's what the user wants, the "AI" companies will get into trouble.

I want LLMs to dial down their randomness (especially around sensitive words like "murder", so they don't constantly crap out false accusations about real people) and in wanting this I am basically asking them to infringe harder.

>>28886
They are fuzzy because they are statistical models, not because the companies are afraid of copyright lawsuits. The "hallucinations" will never go away, because that's how LLMs fundamentally work. There is no difference between getting things right and making shit up, one is as accidental as the other.

This is a fundamental issue with LLMs that cannot be solved.

>>28894
>They are fuzzy because they are statistical models, not because the companies are afraid of copyright lawsuits.
Your argument is the tool just has the properties. But the tool has been chosen again and again while knowing these properties.
>The "hallucinations" will never go away, because that's how LLMs fundamentally work.
The point was not about solving everything, but about reducing errors around sensitive topics. In LLMs you can directly set the "temperature" (randomness) of the output.

>>28895
But those are not errors. You are using the bullshit generator and then act surprised when it gives you bullshit.

>>28896
>yuo are le sUrpriSeD
No yuo are le sUrpriSeD here that despite me having some ideas how to improve how LLM-based bots behave, I am not a fan. I have ideas how to reform capitalism, that doesn't make me a fan of capitalism. I know how LLMs work and I don't have a high opinion about them as a general concept, not just the current versions.

I'm not anti-AI like the average idiot online but clearly the energy demands to sustain its growth can't be met without nuclear energy.

>>28898
You have made it pretty clear that you have no idea what you are talking about.

>>28900
That's neither true of AI as a general concept nor of LLMs specifically. DeepSeek shows this.

>>28901
The knob for fiddling with output temperature does not require new training. Of course it is feasible to make temperature change during the conversation.

https://youtu.be/FjgdohQk0IY
Look at the figures people, short term stop panicking


Unique IPs: 133

[Return][Go to top] [Catalog] | [Home][Post a Reply]
Delete Post [ ]
[ home / rules / faq ] [ overboard / sfw / alt ] [ leftypol / edu / labor / siberia / latam / hobby / tech / games / anime / music / draw / AKM ] [ meta ] [ wiki / tv / tiktok / twitter / patreon ] [ GET / ref / marx / booru ]