this post was submitted on 01 Feb 2024
605 points (97.2% liked)

Memes

45727 readers
827 users here now

Rules:

  1. Be civil and nice.
  2. Try not to excessively repost, as a rule of thumb, wait at least 2 months to do it if you have to.

founded 5 years ago
MODERATORS
605
very upsetting (lemmy.ml)
submitted 9 months ago* (last edited 9 months ago) by cypherpunks@lemmy.ml to c/memes@lemmy.ml
 

captiona screenshot of the text:

Tech companies argued in comments on the website that the way their models ingested creative content was innovative and legal. The venture capital firm Andreessen Horowitz, which has several investments in A.I. start-ups, warned in its comments that any slowdown for A.I. companies in consuming content “would upset at least a decade’s worth of investment-backed expectations that were premised on the current understanding of the scope of copyright protection in this country.”

underneath the screenshot is the "Oh no! Anyway" meme, featuring two pictures of Jeremy Clarkson saying "Oh no!" and "Anyway"

screenshot (copied from this mastodon post) is of a paragraph of the NYT article "The Sleepy Copyright Office in the Middle of a High-Stakes Clash Over A.I."

you are viewing a single comment's thread
view the rest of the comments
[–] mindbleach@sh.itjust.works 12 points 9 months ago (1 children)

I don't give a shit about copyright for training AI.

But I don't give a shit about investors, either.

[–] webghost0101@sopuli.xyz 11 points 9 months ago (1 children)

Copyright should ceise to exist and sharing digital copies of any content should be a protected right. The best software is foss anyway.

But if i cant have that i will settle for techbross going to jail for mass theft. Either the law is equal or it is unlawful.

[–] mindbleach@sh.itjust.works 2 points 9 months ago (1 children)

Nah. Even in its current stupid state, copyright has to recognize that sifting through the entire internet to get a gigabyte of linear algebra is pretty goddamn transformative.

No kidding the machine that speaks English read every book in the library. Fuck else was it gonna do?

[–] webghost0101@sopuli.xyz 0 points 9 months ago (1 children)

For some licensing its not about how transformed it is but wether or not it was used.

Many of the books it read where not supposed to be in this library. The datasets used contains heaps of pirated content.

I repeat i am for abolishing copyright and legalizing digital piracy, including to train intelligence but if that wont be the case and piracy remains illegal then i want to see the “criminals” punisht. Nothing is worse in law then double standards that punish the small and leave alone the giants.

Remember this? He downloaded and shared a grand total of 30 songs. https://abcnews.go.com/Business/story?id=8226751&page=1

[–] mindbleach@sh.itjust.works 2 points 9 months ago (1 children)

"This law is immoral but also twist these criminals' balls off" is a double standard. Mercilessly enforcing shite laws is never a sane position.

Especially when I am telling you - this isn't illegal. The overwhelming majority of AI training is plainly transformative, and based on readily-available public materials. A torrented version of a published work is still a published work. I don't care how they got it, and obviously neither should you, But you want to act infuriated that this cutting-edge technology used dubiously-sourced... text files? Shush.

[–] webghost0101@sopuli.xyz 1 points 9 months ago* (last edited 9 months ago) (1 children)

Thats not quite what i am saying.

Personally i believe the law to be immoral and should be changed and no one should be punished for using, creating or distributing digital copies. Including tech companies.

But my personal opinions dont seem to matter. If we as a society chose to enforce these laws and the consequences for people for breaking them is harsh then logicaly those same consequences should apply to the rich and powerful.

My hope is once these laws threaten the powefull they will finnaly lobby to get fully rid of them so everyone can be free.

I am not upset they use pirated materials, i am a pirate myself because i believe piracy is morally in the right.

I am upset that admitting that in this comment could be used as evidence in police investigation, heavy fines and jail for me personally while if your rich enough you wont, and if they try you just fly off to wherever your not prosecuted.

I fail to realize how it matters how much it is transformed. CopyLEFT works are about any use where something is derived from x. Well they where “used” and the result is therefor a derived work. From that perspective talking about how transformed it is completely unrecognizable is is a strawman because copyleft doesn't care about. Its designed to destroy copyright by forcing free access on derived works.

On a sidenote, i am near certain that meta trained llama on personal profiles and messengers from facebook, instagram, whatsapp and that is why it is such a powerful model for its size. That has nothing to do with copyright and probably Fully legal using some twisted legal words. In this i see a form of exploitation that should be punished but Zuckerberg will never see jail for that so yeah any reason to show tech companies that don't rule the world i am happy to see, even if it uses a law i rather have not existing, especially when that law will keep existing for the rest of us.

[–] mindbleach@sh.itjust.works 2 points 9 months ago (1 children)

I fail to realize how it matters how much it is transformed.

And ignoring when people explain that it does.

You want copyleft to be super duper copyright, where even quoting a sentence of a Cory Doctorow novel demands an entire newspaper gets GPL'd forever. We don't care what fair use says! This anti-copyright goal demands more protection than mere copyright!

This is silly.

Transformation is where copyright does not apply, because you did something new. No matter what works you referenced - your thing is different. It's why Disney can't sue Wikipedia for articles describing their movies. It's also why Wikipedia can't sue OpenAI for models describing their articles.

[–] webghost0101@sopuli.xyz 1 points 9 months ago* (last edited 9 months ago) (1 children)

I want copyleft to destroy copyright by using the law against itself because I believe that to be the only way.

I am all about destabilizing a system that should not exists. I would go as far and say if you read a copyleft book then all your future ideas should be barred from holding patents and become public domain instantly. That is also how i treat my own ideas because my ultimum stance is that:

“The highest reward for any intelligent or creative thought is to see everyone adopt it” Copying is truly a form of flattery its tangible proof that you contributed to the world, that others perceive you as good.

But that is only half my point and its easily refutable as unrealistic and extremist. I feel like its my second stance that is getting ignored.

Which is “if a law is enforced it should be enforced equally towards all social classes

That is because I recognize my main stance is an ideal that no politician would take serious. This second one though, how can anyone disagree?

Again its about destabilizing the system. Rich people don’t like going to jail so if we enforce the strictest versions of laws we can help motivate lobbyist to get rid of copyright all together.

[–] mindbleach@sh.itjust.works 1 points 9 months ago (1 children)

Accurate enforcement of copyright would have no effect on AI because training is fair use.

Wishing it were otherwise is a misunderstanding of copyright, even if you're against copyright.

[–] webghost0101@sopuli.xyz 1 points 9 months ago (1 children)

I am willing to believe that but as far as i hear and read its not a black and white case and government all over the world are busy specifically with trying to weigh in on that.

If training copyright works is 100% proven legal then thats great news because currently i fear smaller foss ai project will be prosecuted leaving ai monopoly at big tech.

It also means i can legally train my own image Model on pictures of Hogwarts to generate cozy castle backdrops for my own art.

[–] mindbleach@sh.itjust.works 1 points 9 months ago (1 children)

It also means i can legally train my own image Model on pictures of Hogwarts to generate cozy castle backdrops for my own art.

As much as you could just look at them and draw stuff.

And with the same potential consequences if you try using a blatant knockoff as a wholly original design.

[–] webghost0101@sopuli.xyz 1 points 9 months ago (1 children)

The thing there is the architecture existed before It was used for hogwarts. Its impossible to tell for me what is custom movie set and what is real castle.

I also have no exact memory of Every movie shot of the castle. Of course i can make Sure it doesn’t have banners of Hogwarts houses but it could have a stained window in the background which is an exact match of a custom order by the studio and i would not realize.

I don’t share or sell my art so i am pretty sure its safe but i do wonder about the technicality of it all. The Eiffel tower is a copyrighted building. Am i allowed to go on a stream type “Eiffel tower” in a prompt window and press generate?

I have seen no legal statements that explains such cases and till there is i would suggest anyone who uses ai to be very careful when publishing.

[–] mindbleach@sh.itjust.works 1 points 9 months ago (1 children)

Gonna bet the floating candles aren't historically accurate.

Am i allowed to go on a stream type “Eiffel tower” in a prompt window and press generate?

Yeah. Free and clear. As much as you're allowed to take a photo of the actual extremely-visible landmark.

But you're probably not allowed to build a replica as a tourist attraction.

I have seen no legal statements that explains such cases

My impression of your understanding is not impressive.

[–] webghost0101@sopuli.xyz 1 points 9 months ago* (last edited 9 months ago) (1 children)

What you are referring to is called freedom of panorama and it is everything but free and clear. France only has a limited version of this since 2016

Wikipedia has to use a censored image for a public statue that newspapers there are still getting sued about.

I did make made an error. The Eiffel tower isn’t copyright protected but the lights on it are. So pictures by day are fair use but at night there Illegal.

I cannot fault you for not knowing any of this because how a reasonable person is supposed to know from the top of their hat wat is and is not a copyrighted work is part of the absurdity about these laws.

Your impression of my understanding does not impress you but at least I somewhat do know what i am talking about. You stating you have never read such legislation feels a bit of the joke when the topic has so much legal history.

[–] mindbleach@sh.itjust.works 1 points 9 months ago (1 children)

Glibly equating taking a picture and publishing a picture does not help your argument.

Generating an image is not about to infringe copyright. No more than drawing it and showing it to the guy beside you. Copyright is about the use of images. Mostly: copying.

Using whatever image comes out as your company logo, as if anything created by anyone through any means can ever be automatically "free and clear" of all prior art, is obviously stupid. Like come on. No shit the draw-anything machine can draw popular characters, major brands, famous scenes, etc., etc., etc. Rendering stuff we recognize is what it's for. Nobody promised you a universally bespoke image, somehow unlike anything reality had ever seen before. Why in the name of god would that expectation be a metric worth discussing?

[–] webghost0101@sopuli.xyz 1 points 9 months ago* (last edited 9 months ago)

Ai now is just a fraction of what it can do in the near future.

I did not Equate taking a picture with Publishing, my example was on a stream which is publishing. I also think people should be free to publish ai creations.

If you feel like a cool social project (inspired by music industry bankrupting kopimashin artwork made by Peter sundre):

train an ai on night pictures of the EiffelTower and build a bot to continually generate more depictions on a live stream. I predict It may take a while for the french government to notice but they will notice and take it down no matter that you don't live in france. I am only not doing this cause i dont wanna get sued.

All i am trying to explain with all of this is that copyright law is incredibly complex, often illogical, often does not get enforced but when it gets enforced it can ruin lives.

People acting like these laws are obvious when they are not is proof that they need to change drastically.