It does make sense. When podgietaru wrote their doo-dad programs, I guess they made their work available for humans to use and they didn't mind giving their work away for free. Some other person might say "I'm giving my work away free, I don't care if humans use it or AI, for any purpose". Both are legitimate choices, each according to their own.

As for "everyone benefiting from training data" - sure, but the AI companies are not investing Billions of dollars out of the goodness of their hearts, it is towards one and only singular goal of making profits (at some point). People might be sympathetic to these AI companies if they at least behave decently - they take everyone's work (text, software, fiction, music, images, videos...) without paying a penny to anyone. If they take everyone's work for free, they should give away anything that is built on that work also for free. This is before we even get to environment, privacy, hammering sites by not respecting robots.txt etc issues.

If I stole all veggies from your garden but spent time/money/effort making a meal, I should at minimum share it with you for free, no? Even if I spent my own money making the meal, it was made from stolen raw material...

▲

CamperBob2 2 hours ago | parent [-]

If I stole all veggies from your garden but spent time/money/effort making a meal, I should at minimum share it with you for free, no?

Yes, and that's exactly my point. We are in violent agreement. It's shared with me, with you, with OP, and with everybody else. We can get a delicious meal for free or we can pay somebody else to serve us a slightly-tastier version.

Here, disregarding copyright law has fulfilled the very purpose of copyright law: to advance the useful arts and sciences. Copyright law was the best tool we had to accomplish that before, and now we have something better.

▲

podgietaru an hour ago | parent | next [-]

I wouldn’t be happy if those companies used my vegetables to open a Michelin star restaurant, even if I got a free McDouble out of the deal.

Plant based I guess I don’t know metaphors are hard.

▲

cycomanic an hour ago | parent | prev [-]

But the AI companies are not publishing the models? Moreover they are charging for access to the models.

	▲	CamperBob2 an hour ago \| parent [-]
		Sure they are. Surf around on HuggingFace and you'll see dozens of open-weight models up to 120B parameters in size from for-profit US companies, freely downloadable (if not freely runnable, alas). Probably hundreds of them, at this point. And a 3T parameter model is scheduled to be dropped by the Chinese on Monday. They should certainly publish more, and if somebody were to argue that model weights trained by scraping copyrighted data should inherently be accessible to everyone, I'd be 100% in favor of that.