4 ms·
Because the law has never delt with the concept of training models like these before and it's not just automatically the same thing as republishing something no
by tracerbulletx 3y ago
Because the law has never delt with the concept of training models like these before and it's not just automatically the same thing as republishing something non-transformative no matter how much some people want to just assert that it is without actually thinking about it. It's a new case, that probably needs new different types of protections and consideration.
- oh_sigh 3y agoDoes the law care about the method(whether it is ML or just reading the article from a database) if the model can spit out a verbatim copy of an article?
- MeImCounting 3y agoI see this argument everywhere and have yet to succeed in getting ChatGPT to give an article. This argument is basically a non-starter
- Retric 3y agoResearchers have gotten large chunks of verbatim text. Suggesting it’s only infringing if the entire work is reproduced is legally incorrect and therefore irrelevant. https://mashable.com/article/chatgpt-revealed-personal-data-verbatim-text-attack-researchers https://mashable.com/article/chatgpt-revealed-personal-data-...
- paulddraper 3y ago> Suggesting it’s only infringing if the entire work is reproduced is legally incorrect It may not be conclusive, but it's highly suggestive: > Amount and substantiality of the portion used in relation to the copyrighted work as a whole: Under this factor, courts look at both the quantity and quality of the copyrighted material that was used. If the use includes a large portion of the copyrighted work, fair use is less likely to be found; if the use employs only a small amount of copyrighted material, fair use is more likely. > https://www.copyright.gov/fair-use/ https://www.copyright.gov/fair-use/
- Retric 3y agoContext is important when reading that, just 7 words can be copyright infringement.
- oh_sigh 3y agoIt's not an argument, it is a question. I'd like to know the answer as to whether the method matters to the courts.
- klyrs 3y agoI've seen many of these "AI" offerings regurgitate their training data verbatim. If chatgpt were miraculously immune to this, despite taking the same basic approach as everybody else, and without significant efforts to prevent regurgitation, then the onus is on them to prove that.
- Dalewyn 3y agoCopyright laws have dealt with computer programs using bits of code or data that they had no right to incorporate. Changing the words spoken to sound sophisticated does not change anything. Obligatory IANAL.
- Retric 3y agoAlgorithms alone don’t qualify as transformative, which is why making an MP3 doesn’t count as a new work. Copyright is insidious because people tried all sorts of ways to get around it in the past. You can’t for example replace all the words in an article with synonyms and get a completely new work. Thus, asking ChatGPT to summarize an article is problematic because the output is based on the original work irrespective of what’s happening internally.
- paulddraper 3y agoAn MP3 is not a significantly different work. 1:1 substitutions of synonyms is not a significantly different work. Summaries are explicitly permitted, which is why CliffNotes, Wikipedia, etc can publish synopses of copyrighted works.
- Retric 3y agoCliff Notes books interpret, analyze, and comment on works which is allowed. Detailed summary alone doesn’t qualify for a fair use exception. Note how much of the ‘summery’ section isn’t just a listing of what happens but an interpretation of what happens: https://www.cliffsnotes.com/literature/f/fahrenheit-451/summary-and-analysis/part-1 https://www.cliffsnotes.com/literature/f/fahrenheit-451/summ... “At the same time, she also gives the reader the opportunity to see that the government has dramatically changed what its citizens perceive as their history. For example, Montag never knew that firemen used to fight actual fires...”
- cscurmudgeon 3y ago> Thus, asking ChatGPT to summarize an article is problematic because the output is based on the original work irrespective of what’s happening internally. You are confusing training a model vs using it. This way a web browser can also be said to use NYT's content.
- Retric 3y agoThat’s in support of copyright being insidious not suggesting summaries are the only issue. The point was people seem to assume only direct copies are problematic but that’s not the case. Training/feature extraction is just as mechanical as making an MP3.