Lol - asking ChatGPT about IP rights... Now *THAT* is ironic!
Here's a link to a cool presentation from last year's DEFCON conference talking about how easy it is to extract private data from LLMs (even when they're not supposed to leak such information).
https://www.youtube.com/watch?v=O7BI4jfEFwA
While the entire presentation is really good, the first few minutes in particular offer 3 clear examples of ChatGPT producing output that clearly infringes on copywrite. (The NY Times example is especially egregious.)
Adam


buffo
Reply With Quote
