To celebrate this one needs to be absolutely sure that LLM will be of no use or relevant now and in the future. I’m not so sure anymore.

I really hope so, but I'm not going to count my chickens yet.

"How influential the latest appeals court ruling will be is debatable. The court found that Ross had effectively copy-pasted Thomson Reuters’ “headnotes,” or brief editorial summaries of legal issues, verbatim for its legal search engine, a practice that isn’t difficult to separate from fair use."

#AI #fairuse

1) using copyrighted data to train LLMs for something unrelated is transformative and fair use (so the frontier labs are safe)

2) training AI on data to make a competitor is not

Presumably using a third-party LLM trained on data to build a competitor would still fall within 2), even if the LLM itself is 1). This has not been tested in court, however, and given the arbitrary and capricious nature of (English) Common Law, any decision could go either way. There is also a legal doctrine that facts cannot be copyrighted, the "sweat of the brow" collating them notwhitstanding.

This is why the NYT and publishers are banging on the legal theory that LLMs can be prompted to regurgitate the articles/books used in their training.

Well, NOT ones which are trained strictly on public domain data. For example, a stable diffusion algorithm can be trained strictly on pre-1930 public domain images. There's also a lot of more recent public domain data, such as NASA photos.

Ethical AI algorithms are possible, they just aren't the things AI corporations can vendor lock you into.

until trump steps in and arrested the judges, and scraps the court ruling to protect his techbros..?

#aibubble #aiscam #aihype #noai

I'm sure the Supreme Court will rule the other way.

They will receive a lot of politictical pressure and financial bribes to do so.