LogFAQs > #974908851

adjl posted...

I'm inclined to say they shouldn't be, given that the concept of fair use hinges on creating something new and LLM's just rearrange whatever they've collected instead of actually creating anything, but it's largely unprecedented legal territory and it's likely going to come down to whose lawyer can be the most convincing.

I mean, "just rearranging whatever they've collected" is basically language in a nutshell. How do you separate "creation" from what LLMs are doing?

The idea that LLMs - most AIs, honestly - are taking bits and pieces of things used to seed the model is a fundamental misunderstanding of how AI works. When an AI bot like ChatGPT is fed a bunch of text, it's not going to clip a sentence from work X and staple it to another sentence from work Y; it's going to break down all that text to try and understand how language works, largely by guessing which words go together in sequence to result in a specific meaning.

I honestly don't think that this case will go anywhere. Right out the gate, there's the question of standing. In order for someone to sue for damages, they have to prove that they've suffered actual harm. Saying, "This technology might threaten my livelihood in the future!" isn't actual harm. The angle they appear to be going for is claiming copyright infringement based on the usage of their writing, but that seems like a pretty weak claim, at least against these particular defendants, especially as the copyrighted work isn't being reproduced in their products.