Hacker News Reader

Comment by hfhdjdks | Hacker News Reader

So Chinese LLMs are bad actors, but USA LLMs are the good guys?

I don't see it that way, but I'm sure from an American perspective that how it seems.

I don’t really see it as good guys or bad guys - just that China (and Russia) don’t really care too much about American copyright.

And there seems to be an an obvious advantage from my perspective to having an information vacuum that is not bound by any kind of copyright law.

If that’s good or bad is more of a matter of opinion.

You've missed the point he was making -- that Chinese and Russian companies don't care about American copyright and will do whatever is in their interest.

And although you were being flippant, yes, Chinese LLMs are bad actors.

What? This is about whether one country wants to cede a massive economic advantage to another country.

On the other hand, you could also argue that if AI takes all financial incentives from professionals to produce original works, then the AI will lose out on quality material to train on and become worse. Unless your argument is there’s no need for anything else created by humanity, everything worth reading has already been written, and humanity has peaked and everyone should stop?

Like all things, it’s about finding a balance. American, or any other, AI isn’t free from the global system which exists around us— capitalism.

>financial incentives from professionals to produce original works

People produce countless volumes of unpaid works of art and fiction purely for the joy of doing so; that's not going to change in future.

Anecdotal but I know lots of creatives (and by creatives I also include some devs) who've stopped publishing anything publicly because of various AI companies just stealing everything they can get their hands on.

They don't mind sharing their work for free to individuals or hell, to a large group of individuals and even companies, but AIs really take it to a whole different level in their eyes.

Whether this is a trend that will accelerate or even make a dent in the grand scheme of things, who knows, but at least in my circle of friends a lot of people are against AI companies (which is basically == M$) being able to get away with their shenanigans.

Why should OpenAI be the one making money off their hard work, even if they do it for free?

The whole "AI training blackhole" thing is a myth. As long as humans are curating the content generated by ML, the content generated is still valid training data. Remember, for every ML generated image you see online, someone had to go through countless attempts to get it to create exactly what they wanted.

So the US should stop enforcing copyright or child labor laws because some other countries may not, giving them an economic advantage?

yes

Well yeah... If they want to keep the lead on AI (which everything indicates they want).

In contrast to child labor laws, which are intended and written to protect vulnerable people from exploitation, current copyright laws are tailored to the interests of Disney et al.

If they were watered down, I wouldn't see any moral or ethical loss in that.

Copyright law is far from perfect, but the concept is not morally bankrupt. It is certainly abused by large entities but it also, in principle, protects small content creators from exploitation as well. In addition to journalists, writers, musicians, and proprietary software vendors, this also includes things like copyleft software being used in unintended ways. When I write copyleft software, it is my intention that it is not used in proprietary software, even if laundered through some linear algebra.

I'm also far more amenable to dismissing copyright laws when there is no profit involved on the part of the violator. Copying a song from a friend's computer is whatever, but selling that song to others certainly feels a lot more wrong. It's not just that OpenAI is violating copyright, they are also making money off of it.

With the exception of source code availability, copyleft is mostly about using copyright to destroy itself. Without copyright (which I feel is unethical), and with additional laws to enforce open sourcing all binaries, copyleft need not exist.

So it is not good when people use copyleft as a justification for copyright, given that its whole purpose was to destroy it.

Source code availability (and the ability to modify the code on a device) is the most important part, IMO , regardless of RMS's original intention. Do you feel that it's ethical that OpenAI is keeping their model closed?

No, because I think such restrictions are unethical in the first place. However, in regards to training, I think it might be a necessary evil to allow companies to ignore copyleft, so smaller entities can ignore copyright to train open models.

Reply on news.ycombinator.com