The neural network is not scraping content for its own purposes, it is for the purpose of the people who are running/training it.

And yes, one person reading a piece of content without paying money for it is far, far better than one person/corporation scraping all of the world's content in order to profit off of it.

> Is a human choosing to do this theft really better than a neural network scraping content

Probably so. I think the differentiation is in the scale at which scraping is done for those kinds of systems.

It’s probably about the same. The difference with sites like e.g. Perplexity is that they have a business model which requires “acquiring” said content for free whereas a single person is just a single person.

> Somewhat related comparison, Is a human choosing to do this theft really better than a neural network scraping content for its own purposes?

Here’s a similar comparison: “Is a human recording a movie at the theatre to rewatch at home really better than the one who shares the recording online?”

Seeing as you’re calling it “theft”, presumably what you mean by “better” is “causes less harm / loss of revenue to the work’s author / publisher”.

I’d say the answer is pretty clear. How many people go through the trouble of bypassing paywalls VS how many use LLMs?

Saying “a neural network scraping content for its own purposes” doesn’t even begin to tell the whole story. Setting aside the neural network is unlikely to be doing the scraping (but being trained on it), it’s not “for its own purpose”, it didn’t choose to willy nilly scrape the internet, it was ordered to by a human (typically) intending to profit from it.