To Innovation - Breaking Politics, Economics, Crypto & IT News

learn more
Reporting from Watford, UK and LA, US since 1996
learn more




TOP STORIES
John McAfee death was a suicide, autopsy confifms

John McAfee's relatives will finally bury him now



Two US authors try to sue Meta and ChatGPT over 'copyright issues'



They argue that AI has definitely read their books



12.Jul.23 3:50 PM
By They argue that AI has definitely read their books
Photo Photobank

   650

Two US authors try to sue Meta and ChatGPT over 'copyright issues'
The American comedian Sarah Silverman and two authors accuse Facebook's parent company Meta and ChatGPT developer OpenAI of copyright infringement. According to the writers, the tech companies trained their artificial intelligence without permission using their books.

In addition to Silverman, the case involves authors Richard Kadrey and Christopher Golden. The evidence presented in the case is that ChatGPT is capable of generating book summaries, which would only be possible if ChatGPT was trained on the texts from the books.

As an example, Silverman's book "The Bedwetter" is mentioned. The documents also include summaries of Kadrey's book "Sandman Slim" and Golden's book "Ararat" as examples.

Although the mentioned summaries have some inaccuracies, the plaintiffs argue that it demonstrates "that ChatGPT retains knowledge of certain works in the training dataset." The lawsuits seek unspecified compensation on behalf of the writers.

OpenAI and Meta have gathered millions of texts from the internet to train their language models. The companies are vague about the specific sources used. However, by asking ChatGPT questions about a book, for example, it can be determined if that book was used to train the language model.

What is permissible and what is not?

The lawsuit is very interesting because there is a lot of uncertainty regarding AI-generated texts. For example, the question is whether copyright protection applies to texts created based on copyrighted material. Additionally, it is unclear whether language models can be trained using copyrighted works.

Similar questions arise in industries such as the music industry. Artificial intelligence can generate new music based on the music of two artists, for example, making it appear as if both stars collaborated.



Back to the list


Related Information: