All about Artificial intelligence
One Judicial decision in Californiain the United States, rejected an action against the Goal for the use of works protected in Artificial Intelligence Training (IA). Meanwhile, a new process was opened against Microsoftwith similar allegations. Both cases reinforce the clash between the creative sector and technology companies about the legal limits of using Copyright material to develop general AI systems.
In the case of the goal, the court considered legitimate to use books and other content to train the Llama model. Already Microsoft faces an accusation of using, without permission, About 200,000 irregularly scanned books to develop your megatron system. Both episodes rekindle the debate on the application of the doctrine of “Fair use” in the United States.
District Judge Vince Chhabria, from San Francisco, decided that the use made by the goal in training its Llama model was sufficiently transformative to fit as a legitimate use under American legislation. The case was moved by authors who claim that their books were used without consent. Among the cited works are The Bedwetterby Sarah Silverman, and Oscar Wao’s fantastic lifeby Junot Diaz.
Despite the favorable decision to the company, the judge acknowledged that there is an open question about the economic impacts of the use of these works. “No matter how transformative the training isit is hard to imagine that it can be fair to use protected books to develop a tool that allows the creation of a potentially infinite flow of competing works, ”said Chhabria, according to the AFP Agency.
The goal celebrated the decision. A company spokesman has stated that open source AI models “are boosting transformative innovations” and that fair use is an important legal basis for technology development.
Microsoft is the target of new action for supposed use of pirated books
While the goal obtained judicial victory, Microsoft faces a process opened by a group of authors at the Federal Court of New York. The action claims that the company used digitized books illegally to train the Model Model Megatronwhich responds to commands in natural language.
According to the complaint, the Microsoft system was built based on a database composed of about 200 thousand books piratesused to feed the algorithm with structures of language, style and themes of protected works.
The document states that AI is able to generate content that “Imitate the syntax, the voice and the themes” of the authors used in training. The information is from Reuters Agency.
If Anthropic reinforces pro-technology trend
Actions against Microsoft and Meta add up to a series of judicial disputes in the US involving AI training with protected works. This week, Anthropic has also benefited from a favorable decision by the American Justice, who understood that the practice of training went with books without permission fits the doctrine of fair use.
However, the court maintained the understanding that the company can still be held responsible for the use of pirated copies of works during the process. The case was detailed In the digital look report.
Read more:
The dispute between technological innovation and copyright
The two cases exemplify the tension between the development of General AI Modelswhich require large volumes of data to function, and the interests of authors and creators who seek to preserve their rights and sources of income.
Technology companies insist that they make legitimate use of the material, creating tools “Transformators” that do not replace the original works, but generate new forms of content. Already authors, musicians and visual artists argue that they are being explored without compensation.
With the advancement of AI and the increase in the use of natural language based models, the discussion about what it configures or not fair use must continue in the courts and public debate.
