One of the most controversial practices by the tech industries developing generative AI models is the usage of copyrighted material to train AI systems without permission or attribution to the original creators. This has lead to a multitude of lawsuits by authors, artists, and publishers such as the New York Times. Now, it has come to light that in the race to be the best in AI, several top companies knowingly used copyrighted material regardless of future consequences. Read the articles below to find out more:
Another question to ponder about the datasets used to train AI models is, how much information is really available online? What types of print materials are not freely available on the internet? Obscure or out of print books and archival materials come to mind. In addition, how much pay-for-access digital content did these companies include in the datasets? Not everything on the internet is free.
There are AI models, such as Adobe Firefly, which are careful about copyright restrictions when choosing which datasets to train their programs. Firefly only uses Adobe stock images, which are copyrighted and free to use for commercial use. To learn more about AI-generated art, take a look at the overview and pro/con arguments in the Points of View Reference Center:
