A False Equivalence: AI Copying is not the same as Human Copying
An LLM is not a creative mind, it is a regurgitator of human creativity.
A false equivalence has been in the air when it comes to describing LLM ingestion and AI training on human data sets
Jim Amos wrote the thoughtful critique below and linked to this article:
Jim wrote:
"An AI being trained on the work of artists then creating new content based on it is just the same as a human artist being influenced by other artists".
"Wait until you find out how a photocopier works"
"If I borrow a book from a library then write my own book that was influenced by what I just read am I plagiarizing?"
An LLM is not a creative mind, it is a regurgitator of human creativity. It understands nothing. It invents nothing. It only sees in statistical patterns of numbered encodings that programmatically map to words.
An LLM is trained on work, in some instances, that was never granted permission to be used in a training model, including a notorious trove of material known as "books3" which is a known quantity of pirated books. The law may not have acknowledged this yet, but that doesn't mean it never happened or that it wasn't wrong - we can make an ethical judgement here, even if the law hasn't caught up yet.
Similarly, companies like Midjourney and Stable Diffusion have already been caught red-handed, through leaked documents that show they know exactly which artists they trained their model on without permission and then encouraging end users to specifically prompt for the art styles of those artists.
An LLM is not an individual, it is a machine existing simultaneously as innumerable copies of itself. It can copy and regurgitate on a scale that could never ever be achieved by a single human with unlawful intent.
Unfortunately, it's getting harder to Google good articles that have covered this false equivalence popping up in comment feeds everywhere.
If any of my followers have found similar articles please share them in the comments below. I do think this one below is pretty good and at least frames the arguments that AI evangelists and other types of apologist keep trying to make.
As Jim wrote in a different post:
“I think we should set higher standards, avoid buying into the hype, and try not to forget that life is a gift and much more wonderous than machinations dreamed up by corporate salesmen.”
Our Human Intelligence™ team heartily agrees.