/leftypol/ - Leftist Politically Incorrect

>>1635767
all those 'why are you poor? just learn to code' smuggies loosing their jobs would be so funny

>>1636398
good. Mars needs more genetic material for the breeding chambers

>>1636398
what the fuck is that meme
whats the source

>>1636804
Pedofinder General from Monkey Dust

>>1633894
I apologize I don’t have a full hour to watch this video, but if I could just say I think “Machine Learning” and “AI” can be sort of disambiguated at this point. Machine learning can refer to something like an LLM, while AI is indeed emerging as a sort of gestalt of higher level abstractions, that is use of underlying models as black boxes to emulate reasoning or problem solving and other higher order cognitive functions. Basically they are emulations of cognitive processes by design. It’s kind of like reducing a CPU to just a bunch of AND gates and inverters. Yes, it IS that. And it is meaningful to understand them that way. But at the same time we have created complex structures out of them which can preform very advanced tasks that combined together form complete systems and create new concepts and need new vocabulary.

Digital commodities were already devalued thanks to the advertisement-supported freeware business model. People were tricked into generating revenue for tech companies for free.

>>1635916
absolutely based. if you arent using ai to make your job easier wtf are you doing

>>1633580
guys our technology is so powerful it's dangerous. guys it's just that powerful that it's actually bad. did you hear our technology is dangerous because of how powerful it is, and I'm not just saying that as marketing

youre a fucking idiot lol

>>1635916
If it keeps fucking up make sure you begin a new convo. Something in the prompt is causing it to sample a wrong token and your past response is being put into the prompt. IME it's best to let it do 80% of the work and don't try and fight it to get that last 20% right. It's too autistic.

Its easy to meme about the "just learn to code, you flyover state bigot!" getting their just desserts because the jobs being automated away aren't just blue collar or baseline white collar tasks like data entry, but actually threatens their coding job or the already tenuous value of their liberal and/or fine arts degrees. However, there's a much bigger issue at hand, but before we get to that I will touch on one element of the above I see as danger.

Watching a cadre of online artists protest AI with such vigor is disappointing for many reasons. Ultimately it comes down to the same automation argument that any factory worker could make - fear that it will take or have a negative impact on their job in some regard- but in some ways its worse because it distorts the nature of how the AI systems work. Their demands that any creative endeavor or media needs to have a "NO AI MAY LOOK AT ME" flag is predicated on the AI being somehow different than how a human goes about learning and creating or replicating from past experience, which is inaccurate. The AI is non-deterministic, it simply takes all its past experiences, and then draws from them to produce whatever is queried, much like a person - it simply does it faster. The majority of these artists on social media crying about needing an AI blocker embedded in their work learned to draw (or paint, model etc) by learning techniques by replicating others' work and styles. From commonplace still life (bowl of fruit etc) and live model exercises to attempting to replicate the works of the old masters to adopting the style of favorite comic or manga authors, imagine if the first time that person who would go onto trying to take patreon commissions wanted to learn to draw they opened their favorite manga file to find that after a few seconds on a page it blacked out with a "artistic study not granted! You are spending too long on this one page and therefore our system has detected you may be trying to study, copy, trace, or otherwise make a derivative work and the author does not grant you this in the license" message , possibly with a "If you'd like to purchase a temporary artistic study license at the cost of $50 per image, please click here and have your payment information ready". This would be preposterous, but it is exactly this that artists seem to demand of AI platforms that "learn' in much the same way as humans do!

This brings me to the crux of the issue is that in their fear that AI may threaten their commission income, they are aligning themselves with megacorps and the content cartels to implement policy that will be more expensive, more restrictive, and more harmful to open culture. ChatGPT , DALL-E , and other for profit proprietary AI platforms would be happy for a world where they, empowered by the subscription fees of everyone to access their proprietary API, improved more and more with every request of their users, would say they're the "friend" to artists and advocate for laws demanding AI licensing restrictions and fees. They can afford to pay for each use of each item in their database, like a megacorp buying access to tons of stock photos for their usage, but this puts the use of AI and training data out of everyone elses' hands who cannot do so. That's part of the threat - its similar Big Tech fostering onerous requirements that make a massive impediment any others from growing to challenge the incumbents. Thus, for both logical/practical and ideological outcomes the demand to have a "no AI/LLM/Machine learning may not look at thing" flag is harmful. Instead, we need to foster developments that make the benefits of AI (and all the vernacular related to it) available widely within society.

While this in part will require regulation (especially if meant to do anything for the public good or on the public dollar MUST be both open source/spec as well as its training data. If you're going to assess the potential for garbage in / garbage out situations, you can't have corpos hiding behind trade secrets on the inputs ), the overall theme should be that ALL models , spec, and training data formats should be open source/libre. Right now, outside of Stable Diffusion which is FOSS/libre and we see all the great benefits thereof, all the other "top tier" AI is proprietary, software as a service. ChatGPT and DALL-E, Elevenlabs voice cloning and others all use proprietary services when you access through API -you can't host it yourself, but every input you provide goes to improving their model. They eventually start charging more for access, sanitizing/censoring outputs, and otherwise ending up with restrictions as soon as they think they're "good enough" to justify it and people will pay. We need to get away from this, because that's a huge amount of data going to train secret, proprietary algos that will soon be used as authoritative or otherwise ahead of other offerings There's a lot more to the discussion, but but ultimately either AI will end up run by a handful of megacorps with huge amounts of assets be it fiscal and hardware or training data to their proprietary benefit, or it will be something openly usable widely for all who choose to and build upon collective work for collective benefit.

>>1637107
>itzzz just marketing
>AI will never be powerful enough to challenge muh biological jelly brain
Copium detected anon chan. Its not marketing. He already left google long back. He is just pointing to a scenario AI could evolve into

>>1633666
>you just don't read
ok, can you please explain what you mean in the op then?

>>1638679
angloids really do believe anything they see online lol

>>1633129
>I do think it’s very funny that typical “anti-capitalists” are now ideologically indistinguishable from the average ancap libertarian boomer that cries about “small businesses",
anti small business leftists sound like neo libs larping as a ceo

>>1639258
>anti small business leftists
yes, small businesses should go bankrupt

>>1639162
I'm not OP, but the only remotely "le terminally online" term there is "cope", which someone who uses "anon" should have no issues understanding. Everything else is literally just marxist jargon.

>>1639258
Muh ma and pa stores. ;_;

>>1639305
All private enterprise is a cancer

>>1638679
and he is giving no counterargument whatsoever to the objections raised about it being simply a powerful machine learning language model far short of any kind of awareness, instead he just does pure sophistry and showmanship

>>1639301
I meant more the part with:
>I do think it’s very funny that typical “anti-capitalists” are now ideologically indistinguishable from the average ancap libertarian boomer that cries about “small businesses", which is kinda insane because AI absolutely destroying the value of digital commodities should be pretty positive for the burgeoning anti-capitalist.
What have "anti-capitalists" said? Which ones? How does it make them indistinguishable from ancaps? Why would the devaluing of digital commodities be positive?
I didn't even know these conversations were happening, which tells me it's just more conversations being had online then anything of substance for the real world. But I may be wrong.

>>1633129
Since most socialists believe planned economy is inefficient now, there has been a certain reaction to this by people who don't know the field that advocate for an AI planned economy. Planned economy is the only way to achieve socialism which for all we know may be the only realistic Marxist model. For all we know, Marx could have just set communism as an asymptotic goal that is never to be fully reached. But socialism can be. Not if it turns out planned economy is less efficient than a market one though. The only hope we have left is that these socialists are delusional and it's all USA propaganda

>>1639305
Petty bourgeois worm learns that working class people don't give a shit about his smol business.
You will be crushed by big capital and rentseekers and receive no pity from us because you are also an exploiter. Clear?

>>1639623
Gorbachev "socialism" while revisionist is still miles better than what followed and his economy was way less oppressive than China and Vietnam's are today

>>1639258
>anti small business leftists
won't someone please think of the petty porkies
to oppose the concentration of capital is historically regressive. all antitrust legislation should be abolished
>>1639628
imagine knowing the Russian Federation exists and saying this

>>1639638
>imagine knowing the Russian Federation exists and saying this
Yes? He was way way way way better than Boris

>>1639305
>>1639628
i still am not gonna take a side in your inter-bourgeois slapfights lmao

>>1638109
it seems no one has bothered reading >>1633510 so i will take this post as an opportunity to underscore and contextualize a few points

>down to the same automation argument that any factory worker could make - fear that it will take or have a negative impact on their job in some regard
the analogy is rather shallow. perhaps there is an element of some vulgar luddism (which ultimately is not something that should be responded to with ridicule in the first place as it is often done), but there is more at stake here. really what matters are the conditions of possibility for ai in the first place. the effectiveness of artificial intelligence is contingent on training data, and this is data ultimately sources from human artists. for some reason, when it comes to data people here don't seem to be able to think about it in a very clear manner, focusing merely on the reproducibility of the data by other machines as opposed to its essential ingredient viz. human action. it is strange because if i were to say that places like youtube and google are not at all free services because they are tracking and selling this data, people have no problem agreeing… it is clear phrased this way that data isn't free, and there is a demand for us to think about its ethical usage and circulation. however, in this specific case of digital art, these points are strangely forgotten for whatever reason. even ignoring this angle, there is also the fact that generative ai is not perfect, and it is definitely still a work in progress. the extraction of artwork as a means to improve this ai is still an ongoing process. here is the key point of disanalogy. the artist is expected to contribute to this process of automation with no expectation of compensation whatsoever, despite the fact that they continue to make contributions to the overall project of artificial intelligence

>The AI is non-deterministic, it simply takes all its past experiences, and then draws from them to produce whatever is queried, much like a person - it simply does it faster
with that pointed out, we see that the saliency of the claim that ai is creative just like human artists starts to come into question. we can further expand the disanalogy from here:
1) (crucial) the ai is not a person whatsoever. it does not have autonomy. it can not set out to embark on its own projects. it is something that is owned by someone else. any attempts to try and compare it to a person are extremely dangerous, as they lend credit to attempts to lend these non-sapient machines personhood. this would be a tragic error, as it basically absolves the people who actually own and work on this technology from all liability. the ai is ultimately a prosthesis for entities that do have self-legislative capacities, and the real arena of contradiction has them as a main ingredient (and not the ai)… an artificial intelligence is not simply some self-sufficient entity, rather it is a project that is embarked by certain people using the information painstakingly fashioned by others. ultimately people need to stop thinking about the conflict here as between artists and ai as though ai has any autonomy or will of its own
2) this ai (partly as a symptom of it lack of genuine subjectivity) has poor generalization ability, and more crucially is unable to take up generalization as an undertaking (to intentionally go out of one's way to make abstractions out of a certain manifold of phenomena, and to alter the parameters of this manifold) as opposed to merely a contingent artefact of its initial development. actual artists do not simply synthesize visual data, but they go out of they way to expand their understanding of the visual world around them, and moreover make new visual inventions (e.g. styles and techniques)… like i said above, they problem solve on paper as they are working on whatever illustration or painting. generative ai is not only unable to do this, it is not even interested in doing this in the first place…
3) (more abstract and less important) when we talk about taking from past experiences, we are here assuming that the ai is in an ongoing development, which isn't the case with narrow ai after the training phase is complete, you are really just interfacing with a crystalline form after a reaction. it is more akin to an incarnated ideality really… i think it is important to see these models more as moments of some larger process rather than self-standing entities or whatnot

>the overall theme should be that ALL models , spec, and training data formats should be open source/libre
i am sorry, but how does this graduate beyond lifestylism? ah, while we are at it we should all just escape to the woods and live in a commune or something… i don't understand this open source culture thing, as it completely ignores the time and resources that is required in order to develop all of this software. you are hoping that the capriciousness of random devs working out of their good will or something can compete with a bunch of megacorporations… the tragic thing here is that with ai this ideal is even more unrealistic when we take into account, again, the economic value of data. how long are we just going to be able to run apis on various websites and scrape data out for free? reddit has already started shutting that stuff down. an extremely important thing about ai especially in its current form is that it is rather rigid, and we severely overestimate how powerful it really is. what is often elided are the ecological conditions that permit ai to thrive and the conditions of possibility for this technology to emerge in the first place. it needs a rather stable environment and it also needs a lot of data to be trained on. this technology is only possible with our very mature internet, and it just so happens the internet is highly centralized. so not only you have the issue of resources, but you also have companies that WILL (see again reddit, and rather annoyingly youtube with youtube-dl) go out of your way to slow down your progress

and i haven't even gotten to the question of what is the strategic meaning of all of this…? speaking of strategy, it is strange i always see people dreading the future of copyright laws, but i never see anyone bother reaching out to artists with a different solution to their concerns. there has barely even been an attempt, and instead they are just dismissed as luddites or landlords. of course, this is not wholly surprising as it is ultimately born from a deeper scorn for people who make art for a living (for whatever reason). we even see this tendency in this very post:
<but actually threatens their coding job or the already tenuous value of their liberal and/or fine arts degrees
it's rlly silly. pls stop. don't even have much to add there, just felt the need to air that out

>>1639410
>What have "anti-capitalists" said? Which ones?
idk seems like he is just referencing like 1 or 2 people on /isg/ and siberia, because that is all i have seen

>>1639694
kek based you got me

Generative networks really don't do what a human does or work like humans do. I think it deserves attention but with AI what we are attempting to do is creating something that *emulates* something humans do and now research is done on things that emulate human reasoning. It should, at the end, look human even if the underlying processes are not the same as those of a brain.

If you work backwards and say "AI does this, and it looks like a human result, therefor…" then you end up saying things like "humans are AI." It's fallacious reasoning that comes from defending this motte and bailey type argument about the *use case* of AI ("AI will never be able to do this task"), when in reality we should expect AI to do most tasks and eventually emulate human reasoning and presentation very well.

Imagine it like a "3d mirror" or something that creates a reflection you can step into and touch. It looks and acts like a human, but this doesn't imply much about the nature of humanity.

>>1633510
>>1639937
Didn't read all that shit, but thanks for the cute dolls. Underrated hobby.

>>1640237
they are indeed cute… no problem…

>>1639937
>it seems no one has bothered reading
I did, it contained some interesting and insightful thinking (and some useless wanking)


>ugh this idea that people who can't draw or whatever simply lack "talent" is so gross and completely ignores the process of learning to draw that involves not only a development of hand-eye coordination, but also a developing understanding of both the visual field and the visual world as well
I disagree with this. Getting AI to do your art involve an exact description, which imply that getting rid of the mechanical skill needed doesnt mean you also automatically get rid of the problem solving and creativity (although you can leave more to it to the AI, but then your piece will prolly not be as good).

>in reality most uses of ai just amounts to just using it to generate the entire piece
but you made the prompt (and rerolled/adjusted until getting a satisfactory result).

>aggressive insistence that human artists will be replaced
the low value jobs consisting in doing exactly what someone pays you to do following their precise instructions ? ofc they will. A single guy with good experience working with AI can prolly produce this kind of work way faster than any purely manual artist can. Just like architects prolly dont need to work as hard to draw now that there are good 3d tools.

>we seldom actually see this in reality
disagree again, you just have to see how when there is happenings there are now lot of AI generated meme, when we used to need a drawfag.

>sure there is some fiddling around with the prompts but this is nothing like the progressive problem-solving one does in illustration or painting
well of course, prompt invocation is closer to being the guy commissioning the art piece with direct control and many round of feedback than the artist realizing it.
Thing is, a good part of art getting paid is just realizing commission, and that is the art most vulnerable to getting replaced by AI tools. When you hire a concept artist, you pay for his creativity more than mechanical skills, with a commission its the reverse.


onto your new post, I agree with most of it

>make new visual inventions (e.g. styles and techniques)…
I think its underestimating the capabilities. We've seen it more clearly with AI trained to beat humans at games, like go or starcraft. At first you train them with and against human strategies, and they can do that and reach a pretty similar level to some of the bests humans. And then you just keep the base incentives and train them against themselves in a vacuum. And suddenly they come up with all kind of new strategies and progress much faster.
Of course, the rules of games are a lot more easily defined and the reward reinforcement procedure a lot easier than with something as nebulous as art, drawing/music styles. But our enjoyment of art doesnt come from nowhere, and so it should actually be possible to have models trained on the "fundamentals" of what make art enjoyable, which might come up with completely innovative "styles".

>>1639617
>Communism is impossible and Marx distinguished it from socialism which apparently isn't the free association of the producers but is "achievable"
Another MLoid outs themselves

>>1652210
srry i stopped browsing leftypol and didn't realize there was a reply in this thread till now

>Getting AI to do your art involve an exact description
this is not true. in prompts, there are a lot of features that an artist would be attentive to that a prompter would not add. i think actually delineating every decision would likely take much longer than actually just drawing the whole thing. if u didnt have ai do. large bulk of the decision making, you would just be drawing purely with words which i think would be much more difficult and much less intuitive than just picking up a pen and drawing something. yes there is experimentation and sometimes even more direct human input involved, but it is not the same as what an artist does, and this is crucial for understanding the relation between the project of artificial intelligence and human art. the experimentation of what prompt to use is more like what u see of a commissioner or an art director, though you already point this out
>the low value jobs consisting in doing exactly what someone pays you to do following their precise instructions ?
idk there is a misconception here which is similar to the above for prompting. you are making it sound as what artists are paid to do is purely mechanical transcription, when there is more problem solving that relies on their understanding of the visual world which also adds value. i do think some art jobs would phase out, but it wouldn't be because their work was just following precise instructions, but rather that their work was more about producing "content" with a very uniform style. at any rate, my main point was that focusing on jobs being "replaced" is the wrong way to think, and rather the nature of draftsmanship as a profession will fundamentally transform. it is not as though all those illustrators doing these other jobs could not instead have their skills transfer to this new domain. talk about jobs getting replaced tends to be too doomsaying/culty
>disagree again, you just have to see how when there is happenings there are now lot of AI generated meme, when we used to need a drawfag.
this was not what i was talking about. my comment was about the ability for generative ai to aids the artist in the work, and allows people with less skill to level the playing field by using it as a tool. with that said it is weird ur first thought of concrete use case for the ai art thing is ai memes?
>I think its underestimating the capabilities […]
idk who knows, but this is just speculation that is not that relevant. if the technology could ever progress to the point where the artificial intelligence does not even need to train on human artwork, then this very dialectic between the vital and cybernetic order comes apart. no one would talk make accusations that ai art is "theft" at that point either. there is a tendency to really focus on the artificial intelligence part of things as though it is just a force of nature, and there is much less attempt to think about how it actually exists materially

The true advantage of AI that will render most human artists job- and useless is speed and quantity.
AI can churn out orders of magnitude more art than humans can. Even if most of it is soulless and has six finger hands, at least some of them will outclass the human artist. All it needs is some rather basic filtering, which is a machine learning/AI job.

>>1635855
new old reaction image

eply 7895435789
I won't be convinced that real AI exists until CAPTCHAs (ostensibly a Turing test) go the fuck away.

>>1683919
AI can already beat captcha they just use them now to make it more computationally expensive to spam

AI is shit

>>1683924
HOLY SHIT THIS CHANGES EVERYTHING
I CAN'T BELIEVE NO-ONE IN THIS THREAD EVER THOUGHT OF A GROUND-BREAKING REVOLUTIONARY CONCEPT LIKE THIS
"AI IS SHIT"
I WILL SCREENSHOT THIS AND POST IT IN EVERY THREAD, THE WORLD MUST KNOW.
AND TO THINK, IF YOU DIDNT OPEN THIS THREAD, IGNORE ALL THE POSTS AND TYPE THIS PROFOUND ANNOUNCEMENT, THE WORKING CLASS MAY HAVE BEEN ENSLAVED FOR A HUNDRED YEARS LONGER. BUT NO, FOR WE HAVE BEEN ENLIGHTED, SAVED, BY THIS CRITICAL INSIGHT.
THE FUTURE IS SAVED. BILLIONS WILL LIVE. AND ITS ALL THANKS TO YOU POSTING WHAT NO-ONE ELSE COULD POSSIBLY HAVE IMAGINED IN THEIR WILDEST DREAMS - AI IS SHIT.

>>1683923
It literally doesn't. APIs to outsource captcha solves to 3rd-world clickfarms have been a thing since 2010 or earlier. 1000 solves for $1 last I checked, look them up.

>>1683944
AI is shit

Anyone who slightly opposes AI and technological accelerationism is not a socialist

>>1633510
Will say
>AI democratizes art
Is some dumb fucking shit bro
No, what democratized art was the internet (anyone can learn to draw with internet access) and relatively cheap art supplies; allowing people with 0 artistic ability to have a machine make an approximation of someone else’s Art isn’t even artistic production, its just more content

>>1687804
>someone else’s Art
oh youre one of those guys who defend IP to death huh

>>1687900
you failed the toothbrush test.

The way I see it, AI is destroying search engines and cenralized social media, driving artists to finally move to the more robust options of federation and personal websites: Overall this is a good thing for artists and will probably cause the for-profit AI maintainers to choke out their own economic incentives. Just training the thing is expensive, and adding a web crawler to that in the anti-webcrawler arms race haphening rn, this will probably leave AI to hobby stuff that artists currently blanket opposed to it would come to be fine with.

>>1687914
its only bad that they havent released the source code and weights and training data, yes :)

>>1635500
Just sell your soul and become a robot skelly boi and just keep resecting at this point.