The Department of Justice has filed a brief arguing that training a large language model on protected texts is generally fair use, and that hindering this practice would be detrimental to national security.
The U.S. Department of Justice has filed a brief in Manhattan federal court supporting OpenAI's position in the litigation against the New York Times. This is the first time that the U.S. executive has ruled in the wave of lawsuits brought by authors, publishers, labels and newsrooms about the use of their works to train AI systems.
The argument is based on the transformative nature of the operation: a model does not use an article to inform or entertain a reader in the way its author intended, but converts the text into digital representations and learns statistical regularities of vocabulary, syntax and knowledge. What emerges from this then makes it possible to edit, translate or produce new text. The thesis adds that restricting training would penalize scientific progress, economic competitiveness and national security.
Such a filing has no binding value for the judge. Instead, it provides a point of support for technology companies that contest these claims, and it comes at a time when the administration is also negotiating its own position in OpenAI's capital — a point that the brief does not mention.
Read the full article on the original source
Go to the source →