The rise of Artificial Intelligence has a titan in the company OpenAI, by Sam Altman. However, there is a very dark secret behind the ChatGPT company, and involves Harry Potter.
The outrage of JK Rowling it’s through the roof.
According to research published this year by David Bamman of the University of California at Berkeley. ChatGPT and other OpenAI language models have used a huge amount of text from the Internet to train their Artificial Intelligence.
So far, all normal. But the bad thing is that it was also included copyrighted books, among them those of the Harry Potter saga, written by JK Rowling.
Bamman and his colleagues revealed this year that Artificial Intelligence was capable of filling in details from a selection of almost 600 fiction books.
The team selected 100 passages from each book with a unique named character. The researchers later deleted the name and asked the AI to fill it in:
- Achievement 98% in passages from Alice in Wonderland, by Lewis Carroll.
- 76% Harry Potter and the Philosopher’s Stone by JK Rowling.
The big difference between the two is that the story written by Lewis Carroll is not copyrighted, while Rowling’s, yes.
Top 20 books that ChatGPT has memorized — similar but a little different than GPT-4
It’s interesting that Zora Neale Hurston is one of the top titles here. She and Chinua Achebe are basically the only non-white authors with a top memorized book from this study. pic.twitter.com/VWxMCHLyxq
— Melanie Walsh (@mellymeldubs) May 5, 2023
This suggests that the Artificial Intelligence of OpenAI was trained, in significant proportions, with material from both books.
But there is not only Harry Potter, but other works protected by copyright, like Ray Bradbury’s Fahrenheit 451 and George RR Martin’s Game of Thrones.
Andrés Guadamuz, specialist in Artificial Intelligence at the University of Sussex, was quoted by New Scientist about the topic.
“The legal issues are a bit complicated,” he acknowledged. “OpenAI is training GPT with online jobs that may include a large number of legitimate quotes from all over the internet, as well as possible pirated copies.”
For Bamman, it is a matter of time to determine how the rights of authors assist in the face of OpenAI’s Artificial Intelligence. “I think it is an open question that many court cases They will decide in the coming months.