zlacker

>>ssgodd+(OP)
Solidly rooting for NYT on this - it’s felt like many creative organizations have been asleep at the wheel while their lunch gets eaten for a second time (the first being at the birth of modern search engines.)

I don’t necessarily fault OpenAI’s decision to initially train their models without entering into licensing agreements - they probably wouldn’t exist and the generative AI revolution may never have happened if they put the horse before the cart. I do think they should quickly course correct at this point and accept the fact that they clearly owe something to the creators of content they are consuming. If they don’t, they are setting themselves up for a bigger loss down the road and leaving the door open for a more established competitor (Google) to do it the right way.

>>kbos87+Na
Doesn't this harm open source ML by adding yet another costly barrier to training models?

>>hacker+1k
It doesn't matter what's good for open source ML.

It matters what is legal and what makes sense.

>>onlyre+zl
The law on this does not currently exist. It is in the process of being created by the courts and legistatures.

I personally think that giving copyright holders control over who is legally allowed to view a work that has been made publicly available is a huge step in the wrong direction. One of those reasons is open source, but really that argument applies just as well to making sure that smaller companies have a chance of competing.

I think it makes much more sense to go after the infringing uses of models rather than putting in another barrier that will further advantage the big players in this space.

>>shkkmo+co
It’s disingenuous to frame using data to train a model as a “view,” of that data. The simple cases are the easy ones, if ChatGPT completely rips a NYT article then that’s obviously infringement; however, there’s an argument to be made that every part of the LLM training dataset is, in part, used in every output of that LLM.

I don’t know the solution, but I don’t like the idea that anything I post online that is openly viewable is automatically opted into being part of ML/AI training data, and I imagine that opinion would be amplified if my writing was a product which was being directly threatened by the very same models.

>>rndmwl+Mu
All I can ever think about with how ML models work is that they sound an awful lot like Data Laundering schemes.

You can get basically-but-not-quite-exactly the copyrighted material that it was trained on.

Saw this a lot with some earlier image models where you could type in an artists name and get their work back.

The fact that AI models are having to put up guardrails to prevent that sort of use is a good sign that they weren't trained ethically and they should be paying a ton of licensing fees to the people whose content they used without permission.

>>bluefi+Ez
>You can get basically-but-not-quite-exactly the copyrighted material that it was trained on.

You can do exactly the same with a human author or artist if you prompt them to. And if you decide to publish this material, you're the one liable for breach of copyright, not the person you instructed to create the material.