On: OpenAI says it has evidence DeepSeek used its model to train competitor

[cscurmudgeon]

Scraping data is different from scraping outputs from a model.

[Rebelgecko → cscurmudgeon]

Copyright is weird and often legal ≠ moral, but I'm having a hard time constructing a mental model where it's ok to scrape a novel written by a person but it's not ok to scrape a story written by chatgpt

[jacobgorm → cscurmudgeon]

No it is not, data is data, whether it gets loaded from a file on disk or generated by multiplying lots of matrices.

[openrisk → cscurmudgeon]

Because the data is mine and the model is yours?

[idle_zealot → cscurmudgeon]

Like, in a strict literal sense, sure? Do you mean to make a claim about moral or legal differences?

[wkz → cscurmudgeon]

Technically, sure. What is the moral distinction though?