Unredacted filings: Microsoft exec privately called AI scraping 'the largest theft of labor in human history'
- Newly unsealed court filings show a Microsoft executive privately called AI scraping "the largest theft of labor in human history", according to TechCrunch.
- The same filings describe Microsoft and OpenAI scraping paywalled New York Times content, building datasets from it, and running the operation together.
- Internal messages in the filings warned that the data practices would gut publishers, a position the companies did not state publicly.
Hacker News opinions
Honestly it's more like the absolute maximum possible degree of theft. There can't be a larger one, it's everything current and past at once.
It's the robbery of all our culture so they can sell it back to us at a markup. Every great fortune has a great crime at its heart, and the most valuable companies on the planet will probably come out of this one. And since justice can be bought, nothing will come of it.
At least the Chinese AI companies are doing some good service open-sourcing their models back to the public.
Would regulation even help here? Right now you can download free models trained on that stolen data. Regulate it and compensate people, and only rich companies can do it, and they definitely won't give it back for free. I keep putting "stolen" in quotes because nobody calls a human reading a book and learning from it stealing.
Every time you write software or build a machine that automates something, you're committing a crime. Every time you learn from a colleague, get better than them, get promoted while they get fired, same thing. Go provide justice there first.
What if it was free, like Wikipedia? At this point all human knowledge needs to be collated in one place and queryable, otherwise progress gets too damn hard because there's too much to learn manually. It helps regular people who wanted to do something but didn't know where to start, while search engines fill up with spam.
And here we are, just watching and doing nothing.
Marx would say theft of labor is fundamental for the capitalist system to operate. The sin is only when you steal from the capitalists. That's why you can't distill from a commercial model.
I get the sentiment and partly agree, but the original hasn't gone anywhere. You can still accumulate knowledge the old way. Maybe it's not theft of knowledge we should be angry about.
Lots of the original content isn't available anymore though. Bots kill sites, AI kills monetization, and both end with the source material disappearing.
If corporations already owned the consumer, with AI they do it by orders of magnitude more. If nothing stops AI from being used to farm the masses for data, we're getting the sci-fi dystopia.
If someone asked me what the largest theft of labor in human history was, I'd have said slavery.
Never ended, it just changed form.
Gulags and forced work camps come to mind too, but I guess this one is larger in man hours.
Funny, TechCrunch argued IP infringement is not theft back in the music piracy era. How quickly the tide turns.
They said theft of labor, not theft of the things being trained on.
For code specifically, how much do LLMs invent versus just pull existing solutions off GitHub? Permissive licenses are fine, but anything derived from copyleft code should stay copyleft. A study on the total development cost of a Linux distribution put the kernel alone at about $1.4 billion.
They do synthesize solutions for problems that exist in your codebase, so it's not copy-paste or an extract from another repo. They work a lot like humans, who build knowledge by pulling from books, blogs, repos, and their own experience.
The most shocking part is that they have a Microsoft exec who actually knows what he's talking about.