And it's forbidden to do that in certain contexts. Selling a service that regurgitate licensed content is neither legal for humans or machines. German court just reminded OpenAI:
The minute it becomes feasible for the RIAA to charge you a fee every time you have a song playing in your head you can bet they'll be sending you a bill or a legal threat. They'll even come after you for singing when it's profitable enough.
Copyright and performance rights are two separate things. It's completely fine for me to go and perform (not record - that does need a license) the latest hit song until my heart is content.
> It's completely fine for me to go and perform (not record - that does need a license) the latest hit song until my heart is content.
Only in private. Copyright law can give the owner exclusive rights to perform a song publicly. If the lawyers can convince a judge that your singing counts as a public performance you can end up on the hook for not getting or being covered under a performance license.
I think your copyright argument is focused on media, like music. This appears be a specific exception that applies to text. Music sampling for example is a direct copy of the recording but quoting text, even though it's a copy, is a new work because although the words are the same it's not the original copied (as in the quote is written or typed by OpenAI).
Do you mean they simply paste information obtained without citation?
Because if you ask for an opinion on a subject it generates new information itself based on the data gathered.
It does sometimes quote sources, which are properly noted and attributed, but how is that wrong? People write books (for money) all the time and reference sources.
I'm not understanding why you think the LLM is different from a person in how it uses information to produce new work.
