The author brings up several examples along the lines of: 'why write a blog post about how something was built' or 'why bother making a technical note public', or 'why bother writing a new language', etc, since the LLMs will just ingest it and never credit you and no one will know.
What this really points to is a new ecosystem which is now built on theft.
I agree, the depression is real, and it has destroyed the community of code as well as the individual spirit of wanting to do, well, anything. What's the point of building anything? Anyone could prompt an AI to get some result, there is no real intelligence necessary. But what has led to this situation? Theft. The plundering of all our collective work and all the intellectual effort that was put into it.
This became really clear to me today while I was playing around with the new Illustrator/Photoshop AI-coded ripoffs of the Adobe suite. (Excuse me, VectorCraft is an "open-source, clean-room reimplementation of Adobe Illustrator, rebuilt in pure Rust" per their Github).
I may loathe Adobe and wish I didn't have to pay them another dime. But there is a fundamental difference between a group of people sitting down to build a replacement for Illustrator (like e.g. Affinity), and someone prompting an AI to just rewrite Illustrator from scratch. The AI is laundering other people's work. The AI contains a compressed version of billions of lines of code, and from that it is unpacking essentially a pirated Frankenstein binary, washed of attribution. It may have no lines of code in common with Illustrator, but everything it has was taken from code that someone wrote.
Imagine someone writes a book and you want to distribute it for free. If you reconstruct the book by taking random words clipped out of newspapers and pasted together, that's still a copy of the book.
The question is, what's anyone's motivation to write anything anymore? Who's going to write a new, better language? But we would not be in this situation except for the grand theft which allowed the creation of LLMs in the first place.
What this really points to is a new ecosystem which is now built on theft.
Only if you redefine "theft" to mean learning from copyrighted material, which is more expansive than anything the RIAA or other copyright cartels have ever argued.
it has destroyed the community of code
Speak for yourself. I'm in a Discord where a bunch of people are working with AIs and each other to get Linux running on Apple Silicon Macs, and it's great to see the rapid progress.
It may have no lines of code in common with Illustrator, but everything it has was taken from code that someone wrote.
I have shocking news about where the code that you write comes from.
If you reconstruct the book by taking random words clipped out of newspapers and pasted together
Sure, which is irrelevant because that's not what LLMs do. ("What about this study that found that you can get them to recite paragraphs from Harry Potter?" Well, I can recite large portions of the Ghostbusters screenplay and my brain is not illegal).
The question is, what's anyone's motivation to write anything anymore?
I wanted a thing, so I built it, and made it available in case other people want it too. AI doesn't affect this other than making it possible to build things faster.
I didn't buy into the "copying unauthorized zeros and ones is theft" narrative (because it's not even an argument) when old music industry executives were screaming about Napster, and I'm certainly not gonna start now.
There's absolutely a moral debate to be had in this topic, but if your starting position is redefining common language and twist the meaning of words beyond reason, then you've already shut it down before it's even started.
I don't think it's twisting common language, but let me clarify: Tensors are a form of compression. If I take a song that a musician put out for free, convert it to a different file format and sell it (that's the key here, the commercial use), that's theft. It was theft when it was Napster, too, but even then there's a big difference between freely distributing pirated stuff and selling it.
The LLMs hoovered up everything, compressed it, and put it up for sale as bootleg IP. That's especially offensive since so much of what they took was already freely available with attribution.
I think your issue is believing that Imaginary Property rights exist at all (especially since you’re basically advocating for software patents with this: “ It may have no lines of code in common with Illustrator, but everything it has was taken from code that someone wrote.”). Heck, how is a human supposed to learn in your view then?
If they get no attribution and have no financial incentive, why should a person create anything? You can wish for a world without property rights, but we live in a world where there are property rights. The question therefore isn't whether the vacuuming of license-based IP violates your personal belief system or not, it's whether laws were broken or not.
This leaves us in a situation where some people can break laws and become enormously, almost infinitely wealthy, whereas other people cannot. And that is intolerable.
For example: I was a taxi driver in 2001 and I had a great idea. What if I started a company that gave people rides for money, with an SMS-based "app" that worked on a flip phone? I started coding it. The only trouble was, that would have been completely illegal. But when someone showed up with a bunch of VC and started doing it, regardless of the law, suddenly we entered a two-tiered world where I (as a cab driver) had to pay to be licensed and undergo inspections and be patrolled by a special taxi police force, whereas they didn't. And I couldn't change my prices to compete.
Now we've gone another level. Imagine I tell Claude to code me a backend and app that's a clone of Uber, and I distribute it to all the Uber drivers. Do you think, for one second, that I would be allowed to operate that business in the wild-west fashion that Uber operated when they started up? Or Airbnb?
Whether you like property rights or not, laws were broken in the training of LLMs, and the net result is a much less level playing field in which the property rights have now essentially all been coalesced into closed models which you have to pay to query. What was once created freely is now owned by Anthropic in the form of Claude, and charged for.
My contention is that this is a much more unequal situation than one in which every author owns, for a time, the right to profit from their own work or to gain recognition by attribution.
There are licenses which allow you to use open-source work, with attribution.
There’s a reason why so many concepts in the sciences are named after somebody. Usually it’s the discoverer.
I’ve got no idea what this implies about learning for humanity as whole, though.
Why is the argument "well it's fine when a human does it". It's not a human, or a living being, or even concious. It's software operating in our social plane (language) owned by a bunch of yahoos in the bay. They are lining their pockets reselling ideas from books they burned..not sure they should be given the same grace a lone individual reading Adobe's source code line by line.
It’s technology built, funded, and used by humans, so of course it’s fine to me. The rest just reads as a the usual political rant against the rich, which I simply couldn’t care less about if I tried.
> Heck, how is a human supposed to learn in your view then?
In your view then, it would be entirely legal for me to torrent all the movies in the world as long as I say it's "to learn"?
If you agree that LLM-produced code in this instance may have no LoCs in common with the source program, in what sense is this different to the replacement being written by humans? Humans also learn to write code by reading it and synthesising that reading to produce new code.
I agree that the motivation for such a consortium being formed is lower when LLMs can vomit out any program of which one can conceive, and I agree that the value in publishing is lower for no-one reading it/attributing it. But I think the functional replacement of a tool which is run in a way that is extortionate or exploitative is an objectively good outcome regardless of whether the replacing tool was small-batch bespoke artisanal Rust or synthetic Rust extruded from the bilgepipe.
Isn't the fact that "anybody" can sit down and write a clone of Illustrator without years of work incredibly liberating?
If the tradeoff is giving up attribution, then is that worth giving this power to other people? This has and will made things that were previously impossible possible.
I think theft is a pretty strong term since what's been "stolen" is now freely available to everyone. People will have to find newer forms of motivation.
None of this work was previously found or accessible by people unless it found the right venues. For example, many people have written languages, only a few have gotten lucky enough to get on HN and get people even paying attention to them. In this new world, it's really not any different. If you find the right venues, people will know what you've created, if you don't as before, nobody will care or know.
Some call is theft, others call it a final liberation of information
Indeed in a weird way, one can also argue that this state is the ultimate copyleft.
If anybody can create any software anytime, everybody gets a lot of the Four Freedoms of Free Software (change it to make it do what you wish, distribute your modified version, etc), without even "having the source".
And all the programmers constantly feeding their new code into training is the contorted version of GPL'ing it.
This is my belief. As I’ve said before, RMS would never need to write a printer driver if an agent could just RE it and write a replacement. Hell, I’ve now done that myself for a device that only had an old binary-only Linux driver.