4 ms·
There are an alarming number of responses seemingly completely unaware of the core thrust of the article (and NYT lawsuit). ChatGPT was able to reproduce and pu
by appplication 3y ago
There are an alarming number of responses seemingly completely unaware of the core thrust of the article (and NYT lawsuit). ChatGPT was able to reproduce and publish significant portions of NYT articles, completely verbatim for hundred-to-thousand word stretches.
It’s not derivative work. We’re way past that. NYT has an exceptionally strong case here and anyone arguing about the merits of copyright is way off the mark. This court case is not going single-handedly to undo copyright. OpenAI has very little going for them other than “this is new, how were we to know it could do this”. So knowing that, the currently trained models are in a very sticky situation.
Further, I don’t see NYT settling. The implications are too large, and if they settle with OpenAI, they will have a similar case pop up with every other model. And every other publisher of digital content with have a similarly merited case. This is an inflection point for generative AI, and it’s looking like it will be either much more expensive or much more limited than we originally thought.
A side effect of this: I am predicting that we will start to see a rise in “pirate” models. Models who eschew all legality, who are trained in a distributed fashion, and whose weights are published not by corporations but by collectives (e.g. torrent models). There is a good chance we see these surpass the official “well behaved” models in effectiveness. It will be an interesting next few years to see this play out.
- RestlessAPI 3y agoSuch a thing happened with DALLE, Midjourney, and Stable Diffusion. Stable Diffusion, when used to its fullest with thing like Control Net and LoRAs, blows the pants off of other proprietary models.
- benlivengood 3y agoMy guess is that OpenAI will be able to basically copy Google/YouYube on this and offer a system like content-ID. Specifically, ChatGPT doesn't reproduce copyrighted works by default; only by request/action of a third party user much like YouTube serving whatever videos people upload. It wasn't the intent of OpenAI to infringe copyright and in fact a lot of or most researchers believed the models were not overfitted enough to reproduce significant portions of arbitrary works.
- NemoNobody 3y agoWell I know exactly what the NYT has - a very strong case. I think this case OUGHT to upend copyright law - it's terribly broken and has been for years. Essentially, if you don't have a massive corp behind a copyright it doesn't mean anything, if a corp is behind something it can be locked forever, regardless of any limits said copyrights are supposed to have. The NYT list nothing from OpenAI using old news - they still lose nothing if openai can reproduce those articles verbatim. If the NYT wins - we lose lots. I think it's time revisit copyright, we can do that you know, it's rather dated, could use an update regardless.