No consent was given by the content creators, no consultation was attempted, and certainly no share of compensation offered, for what is now a commercial enterprise.
This is not to endorse the European position - which is foolish, regressive and short sighted - but only to outline the case overly bureaucratised societies make when established protocols encounters innovation.
Potentially, but I’m glad somebody is taking a minute to question whether or not a handful of tech companies have the right to consume everyone’s information, build their entire business with it and then sell that back to the people whose data they wouldn’t have existed without in the first place.
That’s quite literally what google has done for 25 years, just with advertising for monetization.
I think we're finally coming around on that (I think a lot of regulators were slow on antitrust in the internet space because the harm to the consumer/environment is not immediately obvious - i.e. it's mostly non-monetary externalities)
Same as America as a whole and offshoring to reduce labor cost. The externalities are slow but they're clearly there in the increasing polarization and the growth of the precariat.
https://barac.at/essays/request-for-funding-a-business-exper...
If you limit scraping and fair use, it further entrenches the power of the tech and content giants.
If you're upset about big tech's power and influence, you should be cheering on the free use of publicly available information, as it has the best shot at unseating that power.
What has been closed? Every day there's new foundational models being released. It's exhausting trying to keep up with the pace of change - Alpaca, Vicuna, LLaMA, etc. I'd be shocked if there wasn't a truly open source foundational model with performance equivalent to OpenAI's GPT-4 by the end of the year.
Any move to limit fair-use and scraping of publicly available information through copyright laws, no matter how good of a reason it is, gives more power to the biggest companies.
Explain?
If you block fair-use and scraping, those that hold the most data and copyrights (the largest companies) will be the only ones able to create quality foundational models, thus holding the keys and further entrenching their power.
Too many people seem to think that just because something’s public it means they’re allowed to do whatever they want with it. That’s incorrect. The only barrier is whether someone you copy from is willing to bring legal action.
The relevant quote, originally written in 1990 and cited in 94, is thus:
" [If] the secondary use adds value to the original--if the quoted matter is used as raw material, transformed in the creation of new information, new aesthetics, new insights and understandings--this is the very type of activity that the fair use doctrine intends to protect for the enrichment of society. "
We've wandered into uncharted legal territory with only a few light posts guiding the way. Nothing about this is settled or obvious.
https://www.thefashionlaw.com/ai-trained-on-copyrighted-work...
Doesn't need to be private in order to be copyrighted.
That being said, this is not what the current ban is about. The ban was ordered by GDPR regulators who have suddenly decided that using publicly available information on the web to train machine learning models is not allowed due to "privacy concerns". It has nothing to do with copyright.
We can’t just live in a world where something is by default up for grabs by big companies just because it’s out there.
Scraping seems to create a few dozen winners and 99% losers from the data we have.
They built their businesses on scraping, then when they had their lock-in and monopolies, they turned around and fought against scraping to keep up-starts from eroding their business.
This blog is a great overview of the current landscape around scraping laws.
https://blog.ericgoldman.org/archives/2022/12/hello-youve-be...
Samsung got bit by it https://twitter.com/GergelyOrosz/status/1643903974536347649
They also tried to do the same with the book industry, fortunately publishers had money to pay for lawsuits against Google.
> handful of tech companies
What about the Open Source models? What prevents the rest of the companies to innovate here?
Does that mean I have to pay you royalties for thinking about it or mentioning it? I don't think so.
GPT is like a primitive intelligence that read a large part of the web and got influenced by it, nobody deserves compensation or consultation for the public data it was trained on.
If you put something on a public internet page, you should expect that other beings, artificial or not, are going to get influenced by it.
If that it to be our future, we have to figure out how to rearrange incentives so that humans continue to be motivated to create the content that feeds the machine intelligence or the whole system falls apart.
Yes, if HN were to post copyrighted work that only they could spread, then you would be liable. In the same trend, any patents that HN comments share should not be redistributed/sold by you.
Chatgpt circumvents copyright law in a way that is hard to detect. How many sentences of moby dick does it take to prove chatgpt is plagiarising other peoples protected works?
It's also totally legal to print out a bunch of patents and redistribute those in a book or whatever (except on certain secret patents which aren't relevant here). You can't copyright a patent filing. The only legal restrictions are on building and selling things that infringe on patent claims.
If this is "backward" and "foolish" then I'm a caveman
> If you put something on a public internet page, you should expect that other beings, artificial or not, are going to get influenced by it.
Well yes, and that's exactly why we have laws and regulation
Example: you can't just use "open source" code however you please, you have to abide by the license
How is it different than a person reading several books on a subject and giving a speech based on what they’ve learned? Obviously if the speech was repeating verbatim what was read, this would not be acceptable. But that is not the case here.
Likewise, should the authors of books a child learns to read with be able to claim copyright over all works and language the child produces afterwards simply because their material formed the basis for that child’s neural network related to reading and writing?
But these bans are not based on potential copyright violation of content that was used to train the original model. The ban is primarily based on GDPR violations, so privacy. The ban in Italy seems to be based on three point:
- A data breach occurring on March 20.
- Mass collection of private data for training the model (not sure whether this refers to ChatGPT user queries or the original data, which could contain privacy-sensitive data, also see 'right to forget').
- No age verification.
It's about privacy concerns, not copyright.
All these governments have a duty to their aging population, which also tends to make up the majority of the active voter base. Every action they take tends to be multipronged but at the end of the day it's meant to protect those social programs. Consider mass migration to increase labor force and birth rates. Sure they can claim it was purely out of good will, but you'd have to be pretty myopic not to see the reality there.
I doubt tax income on copyright dividends even registers.
Foolish and regressive because we don't want a foreign corporation to raise almost endless profit off of our data? Keep your products in the US then, nothing of value will be lost if we prevent access to Meta, Twitter and all this bullshit.
Yet americans constantly complain about china not honouring ip protection laws. Why should a company such as openai be allowed to steal content and then reproduce it without royalties?
I assumed consent was hidden deep in some ToS agreement.
I see the argument. (I'm not sure it's the one Italy made.) Still, it's weird to see GDPR turning into a generic IP bludgeon being cheered on.
OpenAI is the worst of the SV/SF "if we get big enough before we get caught it won't matter" mentality. Unsurprising given who's behind the company.
The model weights?
They're integers. There's no personal information stored in the model.
The fact that the model can generate someone's personal information is like saying that a person can type someone's personal information into a Word document so we should ban Office 365.