alt Hacker NewsI think one of the more prominent issues folks take with mass training on OSS is that the companies doing it are now profiting for having done it.
In his follow-up post he talks about him open sourcing old games as a gift, and he doesn't much care how people receive that gift, just that they do.
He doesn't acknowledge that Anthropic, OpenAI, etc, are profiting while the original authors are not.
The original authors most of the time didn't write the software to profit. But that doesn't mean they don't care if other people profit from their work.
It's odd to me that he doesn't acknowledge this.
Replies
>I think one of the more prominent issues folks take with mass training on OSS is that the companies doing it are now profiting for having done it.
What makes this more objectionable than profiting off open source projects by using it directly? eg. tech giants using linux as a server OS, rather than having to pay microsoft thousands per server for a windows server license? With the original GPL, they don't even have to contribute back any patches.
There is a major difference between open-sourcing a completed product versus being an open source maintainer, and I'm disappointed that Carmack is drawing a false equivalence here.
Isn't that the case, and even the point, of all open source, even before AI?
What's the point of a gift if the receiver isn't allowed to benefit/profit from it?
For instance, do you think Linus is upset that ~90% of all internet servers are running his os, for profit, without paying him?
Of course he isn't, that was the point of the whole thing!
Are you upset Netflix, Google, and heck, even Microsoft are raking in millions from services running on Linux? No? Of course you aren't. The original author never expected to be paid. He gave the gift of open source, and what a gift it is!
Are you suggesting that authors didn't know or understand that commercial exploitation of their OSS contributions was possible? If so, that is a complete misrepresentation of history. There has always been open-source licenses that disallowed commercial use. Authors have chosen not to use them, and instead chose licenses, such as MIT/GPL, that allowed commercial use. And there has always been commercial use of OSS. Big companies, small companies, tech companies, oil and gas companies, weapons manufacturers, banks, hardware companies, etc. They all use OSS and they all make a profit from it, without giving anything back to the people who originally wrote it. That's not an edge case or an unexpected consequence, it a fundamental tenet of free (as in freedom) software: You do not get to choose who uses it, or how they use it.
[delayed]
What seems stranger to me is not acknowledging, that most popular OSS explicitly permitted for profit use. It's essentially what made them popular.
Obviously LLMs are new and nobody knew that they would happen. But the part where most popular OSS willfully committed to broad for profit use is not.
> But that doesn't mean they don't care if other people profit from their work
This doesn't make sense. You make something and put out there, for free, of your own will. Why do you care if someone takes it and makes a profit? Shouldn't you have taken that profit route yourself before if that's what you wanted?
> I think one of the more prominent issues folks take with mass training on OSS is that the companies doing it are now profiting for having done it.
He says it's a gift, and if people do whatever, he doesn't care; he already gave it away.
I think it's interesting that nobody would cry that Fabien should shovel cash from his book sales towards Carmack, nor should those who learned how to code by reading source owe something to the authors beyond gratitude and maybe a note here and there.
Even things like Apple's new implementation of SMB, which is "code clean" from GPLv3 Samba, but likely still leans on the years and years of experience and documentation about the SMB protocol.
It has never been the case that publishing a work entitles you to a share of all profits that are downstream of your work. Copyright law protects your ability to receive profits that result from the distribution of the work itself, but that's quite limited.
If you publish a cookbook, you should get a portion of the sales of the cookbook itself, and no one should be allowed to distribute copies of it for free to undermine your sales.
What you don't get is a portion of the revenues of restaurants that use your recipes!
It's also odd to release software under a license allowing commercial use if the authors didn't want that.
> He doesn't acknowledge that Anthropic, OpenAI, etc, are profiting while the original authors are not.
How is this different than any company that uses the open source software?
I find this argument hard to swallow. If open source contributors want to profit from their code being used and prevent big companies from using it or learning from it, open sourcing it would be an irrational choice.
A lot of the use of open source code has directly breached the terms under which that code is shared and they are now monetising the sale of this code.
> But that doesn't mean they don't care if other people profit from their work.
He clearly states his opinions. He doesn't care if other people profit from his code.
>> GPL would prevent outright exploitation by our competitors, but those were to allay fears of my partners to allow me to make the gift
He believes other members in OSS community should have this mindset. Of course it might not be fair, especially for members who are as financially fortunate as him. His point is clear nevertheless.
That's the point? I agree and roughly it's one of two.
A: you made this as a free gift to anyone including openai B: you made this to profit yourself in some way
The argument he makes is if you did the second one don't do opensource?
It does kill a ton of opensource companies though and truth is that model of operating now is not going to work in this new age.
Also is sad because it means the whole system will collapse. The processes that made him famous can no longer be followed. Your open source code will be used by countless people and they will never know your name.
It's not called a distruptive tech for nothing. Can't un opensource all that code without lobotomizing every AI model.
Its a lot less odd when you remember that he's running an AI company himself.
It's not even the profit, but that there is often no new code being contributed.
AI provides an offramp for people to disengage from social coding. People don't see the point because they still don't understand the difference between barely getting something to work and meaningfully improving that thing with new ideas.
If folks don't want LLMs scanning their codebases we should just make some new OSS licenses. Basically, "GPL/BSD/MIT + You pinky promise not to scan this for machine learning".
Either it works and the AI makers stop stop slurping up OSS or it doesn't hold up in court and shrinkwrap licenses are deemed bullshit. A win/win scenario if you ask me.
I haven't given a cent to openai or anthropic but they have given me many thousands of tokens for free.
>I think one of the more prominent issues folks take with mass training on OSS is that the companies doing it are now profiting for having done it.
I've noticed this thing where people who have decided they are strongly "anti-AI" will just parrot talking points without really thinking them through, and this is a common one.
Someone made this argument to me recently, but when probed, they were also against open weights models training on OSS as well, because they simply don't want LLMs to exist as a going concern. It seems like the profit "reason" is just a convenient bullet point that resonates with people that dislike corporations or the current capitalist structure.
Similarly, plenty of folks driving big gas guzzling vehicles and generally not terribly climate-focused will spread misinformation about AI water usage. It's frankly kind of maddening. I wish people would just give their actual reasons, which are largely (actually) motivated by perceived economic vulnerability.
Carmack is the same person comfortable with delaying talks of ethical treatment of a digital being, or what even constitutes one until in his eyes "they demonstrate the capabilities of a two year old" by which point, with the scale we distribute these models at, and the dependence we're pushing the world to adopt on them, we'll be well into the "implicit atrocity zone", and so far down the sunk cost trail, everyone will just decide to skip the ethics talk altogether if we wait that long. This is in spite of being a family man, which raises serious questions to me about how he must treat them. It does not surprise me at all the man has blindspots I could fit a semi-truck in.
I'm no Carmack, but everything I've released as open source is a gift with no strings (unless it was to a project with a restrictive license). A gift with strings isn't exactly a gift.
If you take my gift and profit, it doesn't hurt me, there were no strings. Your users presumably benefit from the software I wrote, unless you're using it for evil, but I don't have enough clout to use an only IBM may use it for evil license. You benefit from the software I wrote. I've made the world a better place and I didn't have to market or support my software; win-win.
I've done plenty of software for hire too. I've used plenty of open source software for work. Ocassionally, I've been able to contribute to open source while working for hire, which is always awesome. It's great to be paid to find and fix problems my employer is having and be able to contribute upstream to fix them for lots more people.