Lawyer-turned-software-developer reporting in! (American, to be clear…)
My concern with AI generated code has to do with software licensing. Mainly…. What is the license of AI generated code??? What if it regurgitates entire sections of proprietary code? Unless I’ve totally missed some legal/tech news, these questions are FAR from settled.
I feel like I’m taking crazy pills sometimes, but that seems like something the Linux kernel community should be at least mildly curious about….
Look, to be fair, maybe they are! I don’t follow the kernel mailing list; I don’t know what they talk about. I’m just a humble backend developer/infrastructure guy.
But I know it worries me… personally, I mostly use LLMs in my own work via asking questions about code I’ve already written. Then I look at suggestions it’s made, or potential issues it points out, and decide what to do accordingly. But that’s just me!
It really seems like way too many people are just kinda going “eh, everyone else is doing it. It’d be impossible for the courts to force us to reverse course on this now.” Like, there’s no way the monkey is going to get the cork back in, might as well just get used to being covered in shit.
Yeah, that is a pretty shitty consequence, but it’s nothing new or different from the strandard clean room techniques the industry has been using for decades.
A model that was trained on specific code that gets put out exactly the same doesn’t seem clean room to me.
Even if you argue it was turned into weights and probability and may not have originated from that specific code, by having included it in the training set you poisoned the model with no longer being able to claim plausible deniability - being produced rather than copied.
The transformation is still a machine and algorithm, even if it goes through a ton of obfuscation, distribution, and probability.
Unironically a AI’s biggest win. Absolutely fair criticism on art and weak imitations of human creativity, but code is just math with syntactic sugar. “Owning” algorithms and hiding implementations is such an absurd concept.
I often wonder how much better and more accessible software would be if everyone wasn’t reinventing nearly every wheel all the time. What if every phone could copy-paste an Airdrop implementation? What if printers just worked? What if you could put adblockers on any device? What if every online game could be self hosted and modded? AI is (for the moment) giving an approximation of that timeline and I’m here for it.
I often wonder how much better and more accessible software would be if everyone wasn’t reinventing nearly every wheel all the time.
Well… we have open source for a while, and people are still reinventing the wheel every time. There’s always the new chat protocol that claims to be the best, but works almost the same as all others, there’s always reimplementations of some software in another language just because the developers don’t like the original language. There are forks of software due to ideological divergence. There are reimplementations of famous software just for fun. There’s a lot of other cases, all reinventing of the wheel, all the time.
It’s arguably not always a bad thing for people to reinvent the wheel, but anyway, I see it getting MUCH worse with ai. People are already going to chatbots asking to implement things before even searching for work already done on that, and we’re going to end with milions of businesses and individuals with reinvented wheels everywhere, instead of everyone using a shared codebase. If nothing happens in the near future, I see things getting less compatible with each other due to no standards in implementations
I’m incredibly against the big LLM SaaS providers and material waste besides. Anyway I don’t have a crystal ball and I’m highly sceptical that this technology will work out for average people, but it sure would be interesting if it somehow lead to less corporate control and cultural enclosure rather than how it seems to be going right now.
You believe people using systems created by big corps in giant datacenters, that depend almost exclusively on nvidia gpus to run will lead to less corporate control? I’m highly skeptical
I do write code. Code only exists as a high level abstraction that codifies logic and allows us to talk about the function of the silicone without getting bogged down in details. The utility of code depends entirely on the context of the people looking at it (eg. different code works better for a small shop of experienced devs vs a large scale enterprise vs a FOSS project).
I can show you a for loop and talk in simple English about what it’s doing or write a formal proof showing that it meets the problem parameters. I could show you the compiled ASM and we would struggle to have the same conversation even with the exact same functionality. Same math, different syntactic sugar.
The process of building software is where humans will always be needed (identify the human meat space problems, define features, manage risk and cost, set delivery expectations, etc). The specifics of how we create the code within that process isn’t as crucial.
Just playing devil’s advocate here; do you think an architect should not be able to own the blueprints they’ve produced?
To a slightly broader point, do you not agree with the notion that technical solutions can have creative implementations, and that that can be a form of artistry?
I don’t know much about architecture, but a building is generally bespoke. It’s made to a specific physical location and has to work in its surroundings and be built to local regulations. From that viewpoint I can’t really judge how much value the blueprint actually has and therefore if it matters who “owns” the IP. The architect has done real legwork to bring that specific building to life and anyone else would need to redo much of it to copy the result.
Any human effort can be a creative artistic expression. You could mass produce a jacket or a hand-stitch an equivalent jacket and both will be warm. You could code a subtle and elegant CMS for a website or use some pre-built solution, both will deliver the same content. The difference is, people will laugh you out of the room for trying to patent your home stitching technique but nod and clap when you put a license on your code.
What if every phone could copy-paste an Airdrop implementation?
That’s apple’s decision, one of many to force their users to only get their devices, similarly not being able to host your own game servers benefits the publishers/ devs. (Edit for clarity: by this I mean that this state of things isn’t caused by unavailability of open source solutions but rather by corporate greed.)
“Owning” algorithms and hiding implementations is such an absurd concept.
Not any different from protecting one’s art creations. You spend a bunch of time learning how to code and are doing it for a living so you don’t want others/ competition to take it and sell as their own. Not sure what’s hard to understand.
That’s true of course. And it’s true for a lot of people in other areas as well.
In vacuum I’m all for open source software, in fact I would like it a lot if everything was open. But reality is that as a developer in a smaller software company we’d be out of our jobs immediately as other, larger companies would take everything that’s worth taking and use their larger budgets to out advertise their copy.
Oh no! Not our beloved corporate-owned proprietary intellectual property! Won’t somebody think of the sacred copyright system, bestowed upon us by God? /s
Sorry, but it sounds like you’re saying that AI is a threat to tech Capitalism, and that this is a bad thing somehow.
Friendly reminder of how dumb people can be, thank you kind stranger.
The problem with licensing here is very obviously that this makes the GPL weaker, potentially giving more power to corporate interests. Couldn’t you take half the time it took you to write your comment, and think about it yourself? Out of tokens?
Lawyer-turned-software-developer reporting in! (American, to be clear…)
My concern with AI generated code has to do with software licensing. Mainly…. What is the license of AI generated code??? What if it regurgitates entire sections of proprietary code? Unless I’ve totally missed some legal/tech news, these questions are FAR from settled.
I feel like I’m taking crazy pills sometimes, but that seems like something the Linux kernel community should be at least mildly curious about….
Look, to be fair, maybe they are! I don’t follow the kernel mailing list; I don’t know what they talk about. I’m just a humble backend developer/infrastructure guy.
But I know it worries me… personally, I mostly use LLMs in my own work via asking questions about code I’ve already written. Then I look at suggestions it’s made, or potential issues it points out, and decide what to do accordingly. But that’s just me!
Lawyer-turned-software-developer.
Are you the inverse of lawtechie from reddit?
It really seems like way too many people are just kinda going “eh, everyone else is doing it. It’d be impossible for the courts to force us to reverse course on this now.” Like, there’s no way the monkey is going to get the cork back in, might as well just get used to being covered in shit.
I’m more miffed about GPL or copyleft code going to proprietary software and then being profited from without any attribution. It’s a poison pill.
Yes! This concern absolutely cuts both ways.
Yeah, that is a pretty shitty consequence, but it’s nothing new or different from the strandard clean room techniques the industry has been using for decades.
I see it quite different from clean room.
A model that was trained on specific code that gets put out exactly the same doesn’t seem clean room to me.
Even if you argue it was turned into weights and probability and may not have originated from that specific code, by having included it in the training set you poisoned the model with no longer being able to claim plausible deniability - being produced rather than copied.
The transformation is still a machine and algorithm, even if it goes through a ton of obfuscation, distribution, and probability.
I mean it’s clean room but without even having to pay highly skilled engineers to do all the labour. It’s pretty catastrophic to the field.
Yeah, fair enough.
If it regurgitates proprietary code the model should not have had access to for training data… well…
That street goes both ways, there’s more GPLed code in the training corpus than there is leaked proprietary code.
Unironically a AI’s biggest win. Absolutely fair criticism on art and weak imitations of human creativity, but code is just math with syntactic sugar. “Owning” algorithms and hiding implementations is such an absurd concept.
I often wonder how much better and more accessible software would be if everyone wasn’t reinventing nearly every wheel all the time. What if every phone could copy-paste an Airdrop implementation? What if printers just worked? What if you could put adblockers on any device? What if every online game could be self hosted and modded? AI is (for the moment) giving an approximation of that timeline and I’m here for it.
Well… we have open source for a while, and people are still reinventing the wheel every time. There’s always the new chat protocol that claims to be the best, but works almost the same as all others, there’s always reimplementations of some software in another language just because the developers don’t like the original language. There are forks of software due to ideological divergence. There are reimplementations of famous software just for fun. There’s a lot of other cases, all reinventing of the wheel, all the time.
It’s arguably not always a bad thing for people to reinvent the wheel, but anyway, I see it getting MUCH worse with ai. People are already going to chatbots asking to implement things before even searching for work already done on that, and we’re going to end with milions of businesses and individuals with reinvented wheels everywhere, instead of everyone using a shared codebase. If nothing happens in the near future, I see things getting less compatible with each other due to no standards in implementations
I’m incredibly against the big LLM SaaS providers and material waste besides. Anyway I don’t have a crystal ball and I’m highly sceptical that this technology will work out for average people, but it sure would be interesting if it somehow lead to less corporate control and cultural enclosure rather than how it seems to be going right now.
You believe people using systems created by big corps in giant datacenters, that depend almost exclusively on nvidia gpus to run will lead to less corporate control? I’m highly skeptical
That’s exactly what I said—that I’m highly sceptical.
It would basically have to backfire on all their plans and all the trillions of dollars they’re spending.
It would be very surprising. I would be so surprised.
I probably misread it, don’t mind me
Sorry we’re enemies now. We have to fight anytime we’re in the same thread. I wish it didn’t have to be this way.
Those are the internet rules after all, so we must follow. See you in the next thread
But what about the poor corporations making money off those games that already fired all the devs that made it after overworking them?
“code is just math with syntactic sugar”
Wow!
Tell me you don’t write code wihout telling me you don’t write code.
I write code, and I see the point they’re making. It has the same feel as “chemistry is just applied physics” and such.
I do write code. Code only exists as a high level abstraction that codifies logic and allows us to talk about the function of the silicone without getting bogged down in details. The utility of code depends entirely on the context of the people looking at it (eg. different code works better for a small shop of experienced devs vs a large scale enterprise vs a FOSS project).
I can show you a
forloop and talk in simple English about what it’s doing or write a formal proof showing that it meets the problem parameters. I could show you the compiled ASM and we would struggle to have the same conversation even with the exact same functionality. Same math, different syntactic sugar.The process of building software is where humans will always be needed (identify the human meat space problems, define features, manage risk and cost, set delivery expectations, etc). The specifics of how we create the code within that process isn’t as crucial.
Just playing devil’s advocate here; do you think an architect should not be able to own the blueprints they’ve produced?
To a slightly broader point, do you not agree with the notion that technical solutions can have creative implementations, and that that can be a form of artistry?
I don’t know much about architecture, but a building is generally bespoke. It’s made to a specific physical location and has to work in its surroundings and be built to local regulations. From that viewpoint I can’t really judge how much value the blueprint actually has and therefore if it matters who “owns” the IP. The architect has done real legwork to bring that specific building to life and anyone else would need to redo much of it to copy the result.
Any human effort can be a creative artistic expression. You could mass produce a jacket or a hand-stitch an equivalent jacket and both will be warm. You could code a subtle and elegant CMS for a website or use some pre-built solution, both will deliver the same content. The difference is, people will laugh you out of the room for trying to patent your home stitching technique but nod and clap when you put a license on your code.
You could say the same about patents for physical inventions. But there are legitimate reasons for patents to exist.
That’s apple’s decision, one of many to force their users to only get their devices, similarly not being able to host your own game servers benefits the publishers/ devs. (Edit for clarity: by this I mean that this state of things isn’t caused by unavailability of open source solutions but rather by corporate greed.)
Not any different from protecting one’s art creations. You spend a bunch of time learning how to code and are doing it for a living so you don’t want others/ competition to take it and sell as their own. Not sure what’s hard to understand.
Most programmers earn a wage, not the full value of their labor.
That’s true of course. And it’s true for a lot of people in other areas as well. In vacuum I’m all for open source software, in fact I would like it a lot if everything was open. But reality is that as a developer in a smaller software company we’d be out of our jobs immediately as other, larger companies would take everything that’s worth taking and use their larger budgets to out advertise their copy.
Oh no! Not our beloved corporate-owned proprietary intellectual property! Won’t somebody think of the sacred copyright system, bestowed upon us by God? /s
Sorry, but it sounds like you’re saying that AI is a threat to tech Capitalism, and that this is a bad thing somehow.
Friendly reminder of how dumb people can be, thank you kind stranger.
The problem with licensing here is very obviously that this makes the GPL weaker, potentially giving more power to corporate interests. Couldn’t you take half the time it took you to write your comment, and think about it yourself? Out of tokens?
Software licensing is dead, you should really catch up because what are even talking about.