Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Maybe AI content is just a way to launder copyright.


You can only say that if you also say that the creativity of human beings themselves is just a way to launder copyright.


This false equivalency just keeps being made.

Yeah absolutely, acceptable in case of humans and not acceptable in case of GPT.

GPT is not human. It is employed by a human to get around having to benefit other humans with attention and money.

Meanwhile human world is structured around fate, happiness of humans -- such things guide our considerations as to what to consider legal and what not.

There are other differences that make it OK for humans and not GPT (like we can attribute and credit what we learned from, whose ideas we used in work and what inspired us) but they are less significant than above principle


Machines -- at least today -- are extensions of human beings, and pretty democratic, really, compared to history. You can't really talk about limiting machines without also limiting humans. GPT is a tool. Almost anyone can use tools much like it. So what are we trying to protect against?


Yes. Machines including GPT are not the ones doing things, they can't be compared to humans. And humans may desire to do things with machines that should not be considered acceptable because of how they affect human society, and humans may or should be limited (as they already are in all aspects of life that involve other humans).


There's this magic phrase that makes all the difference: at scale.


That seems like saying a steam shovel is bad because it moves dirt at scale. Machines do increase the reach of people, that is exactly what they are for.


People scale. There's billions of us. And we compete.


AIs scale faster. Just one instance can do the work of billions of people, and have its owner profit off it.


How can the owner profit off something that is "free" to make for anyone? Mass spamming is hard to achieve with current day social networks.


Mass spamming is the very raison d'etre for social network companies. It goes by different names though - "sponsored content", "fanpages", "influencers", etc.

The spam did not disappear with people moving away from e-mail. On the contrary, it's just been given legitimacy by the platform owners - who all happen to make money from spam.


"Good artists copy; great artists steal"

-Pablo Picasso


He writeth best who stealeth best, all things both great and small.

For greater minds that wrote them first, from nature stole them all.

-- Totally me.


I like that phrase! I'm probably going to steal it :-)


Now that joke time is over, I'll disclose that it seems to be attributed to one Robert E. D. Cattley, in here: http://www.nytimes.com/1983/12/20/science/homer-s-sea-wine-d...


Well, the uncomfortable truth is that our copyright regime doesn't really make sense. An AI is sort of the equivalent of a junior coder or an art student - it can look at existing work and mash it up and produce something in the same style that's novel (inspired by a prompt - which is like your teacher giving you an assignment). That's copyrightable when it's a human being doing it, and there is no inherent protection for art style, I can make all the Picasso-style artwork I want even if I'm blatantly ripping off his style, and I would own the copyright on the resulting works. Like, if I do a painting in the style of Picasso, should Picasso get the resulting copyright? Why?

And programmers are not without sin either - how many people here run to stackoverflow and copy some answer, or mash it up and produce something similar-but-cleaned-up? That's going to get a lot more powerful in not too many years too... and the thing is, is your little DTO file with blocks of fields and setters and getters really novel and worthy of a government license to cultural exclusivity, or is it a relatively obvious implementation of an API? Which (per Oracle v. Google) are also not really copyrightable...

Humanity hasn't squared the reality of most content being low-value generative trash already (despite being human-generated!) with the reality of a system that automatically awards up to ~150 years of exclusivity (lifetime of the author +75 years). Like, that's the long and short of it. A lot of what we do as humans is just building up a squishy water-filled neural net ("how to make a painting in the style of Picasso") and running random prompts and goal-seeking through it and cleaning up the generative output. Now AIs have gotten to the level of "average college student" and it turns out that's not actually all that difficult or interesting anymore, but we're still working in this model where that's supposed to be worth ~150 years of cultural exclusivity and protection automatically.

If you only get copyright on the things that are truly novel and have no previous creative basis or cultural/stylistic heritage... that's a pretty big change to the copyright system for humans too! Especially in light of the fact that there's only a few ways to implement most configurations/functions... does Spring own all the copyright to your config classes because they're basically doing the same thing as the spring demo projects? hard to see much novelty in that sort of thing pretty much ever, so those probably should not be copyrightable at all under a "totally novel" requirement, but if they do surely Spring owns the copyright.

As mentioned in a sibling comment, the "at scale" bit is surely relevant in a practical sense, but, it doesn't change the underlying-principles basis. This is a situation where "at scale" has just thrown the underlying problem into stark relief. A 2-lifetime (life of author +75 years) governmental grant of cultural exclusivity doesn't really make sense in a world where humans are constantly and trivially mashing up each others' ideas even within a single lifetime, it just doesn't fit the way we exchange ideas in general. And now you have the college-student-bot that can generate a million shitty DTO classes or artworks-in-the-style per day that makes that plain. But ultimately it's just doing the same thing as an art student but faster... if you put an art student in a Mechanical Turk box and made them draw whatever you demanded, would they not own the copyright? And would it be a problem if India had an art school that set up an "industrial art" process and employed a million students in boxes drawing whatever you said to draw? It would probably be problematic for DeviantArt commissioners, but would it violate the principle of copyright?

We just haven't managed to square that reality where most human works are fairly trivial derivatives of other works (your deviantart commissions really don't mash up any similar works or styles that came before? really?) and yet worthy of 150 years of license to cultural exclusivity. And the college-student-bot is blowing that model up. People still need to eat, of course, but that's an orthogonal problem to the societal framework we've set up not really making sense. We can solve the copyright problem and not the eating problem (as we have many times before - buggy whip manufacturers adapted too), or we can solve the eating problem and not the copyright problem, or solve both or neither. People not being able to eat and shelter themselves is a capitalism problem, not a GPT problem, and we will probably have to address the idea that everybody needs to work to survive as machines continue to displace the last strongholds of human labor (as I'd argue creative work has become). We should not be restricting our ideas on collaboration and cultural generation just because it produces inconvenient outcomes for the current copyright regime and capitalistic model, the model needs to bend to fit humans and not the other way around.

This isn't the only place the inherent contradictions and problems of copyright come up either. The idea that a nighttime photograph of the Paris skyline could be copyrighted by someone else other than the photographer who created it is pretty janky too. Is a photographer not adding significant creative value with composition/etc? Literally just having that one element makes my photo derivative? Maybe that's the analogy to GPT/stable-diffusion I guess - but the thing is, it still works the same with a human as with a robot, an AI-generated Paris skyline still would be copyrighted by the architect too. And if the argument is that the code owner or original artist owns the style, then, Pablo Picasso owns the copyright to all my paintings too, just like the architect of the Eiffel Tower owns the copyright to my photographs of the Paris nighttime skyline. And Spring (or Oracle!) owns the copyright to a great deal of your Java code. So that probably is not going to work out great for you either.


Are answers to Stack Overflow not supposed to be copy pasted, just right away? It seems hard to believe, i thought that was the whole purpose of the site!

Well, copyright suffers from the same problem communism suffers. For communism to work, everyone should be on board. For copyright and patents to work as expected, every country on earth, every town and every city's laws should be almost the same regarding to that. As soon as even just one municipality anywhere on the planet, is free of patent laws, the exact same moment a factory will pop up and produce an expensive patented pharmaceutical product which it will sell for half the price, or 1/10 of the price.

Capitalism on the other hand, works even almost everyone else doesn't like it, and 2 people agree on a peaceful exchange of commodities, following a price the market sets, or it is discovered right away.

As a rule of thumb, anything that requires everyone to implement rules and follow them, for it to work, will fail sooner than later. When a product in some part of the world is selling 100x of the normal price because of some patent, counter incentives are detonated in some other part to produce it and sell it for 50x the price. This huge profit margin creates a chain reaction of other economic actors producing it as well, and selling it for 25x the price.

Additionaly the eating problem can be solved by capitalism, not communism. Exchange of art for electronic money, a marketplace of information is what we need. Bitcoin (not BTC) will play an instrumental role in that.

A.I generated images of picasso, are awesome. I created many of them, one series of them are "Darth Vader riding a red tricycle, in the style of picasso". They are so good you're gonna fall of the chair.


That's certainly what my impoverished artist friends have taken away from it.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: