Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Can the model use the text without making a copy (in memory) to process it?


technical copies are different, there are exemptions


Can a human?


Not relevant. What humans do isn't necessarily viewed the same way as what machines do.


Machines need to be prompted with the exact prefix to be able to retrieve any copyrighted fragment, and that doesn't work most of the times. So the intent for copyright circumvention is in the prompt. Temperature settings also matter.


That could be besides the point as for the training the machines will create a copy in their memory - or is that no longer the case?


I don't believe transient copies are the real issue here. Having a temporary copy stored locally in memory is more of a technical nuance than anything. That copy remains isolated and is not being distributed or shared. In contrast, using something like BitTorrent actively spreads copies across the internet.

There are also cases where loading copyrighted content is unavoidable in order to even view the license terms. For example, a website could deny the right to copy its content before you even see the licensing details. Temporary copies enabled through normal usage like this should be permissible.


I think this really comes down to the specifics of the situation.


For displaying/viewing any digital device needs to create an in memory 'copy'. If you would extend copyright to that extreme, it would become completely nonsensical imho.


Not a lawyer, but if you show a streamed movie to an audience without having any rights, how would that be covered? Only by considering the stream provider?




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: