Anthropic logo at the top of coding and court images
After the decision of the US District Court in the Northern District of California, issued on Tuesday, companies are given free repetition for training with almost any published media that they want to collect.
The Decree is based on the trial of Andrea Barts, Charles Greber and Kirka Wallace Johnson against anthropic dating until 2024. The lawsuit accused the company of using pirate material for teaching Claude AI models.
This included the anthropic creation of digital copies of printed books for teaching artificial intelligence.
decision of Judge William Alsup and MDash; The judge is very familiar to the readers of Appleinsider & Mdash; Rules in favor of each side are on the same. Nevertheless, the weight of the decision, of course, came on the side of the anthropic and scrains of AI in this case.
In accordance with the decision, Judge Alsup says that copies used to teach specific LLMS were justified as fair use.
“The technology was one of the most transforming ones that many of us will see in our lives,” Alsup commented.
for physical copies that were converted from the printed library into a digital library, it was also considered fair use. In addition, the use of this content for LLMS training was also fair.
Alsup compared the author’s complaint with if the same argument was used against schoolchildren's efforts, how to write well. It is not clear how this is applicable, given that models of artificial intelligence are not considered “schoolchildren” in which legal sense.
In this argument, Alsup ruled that the author’s law is intended to promote original works of authorship, and not “protect the authors from competition”.
where the authors saw a little success in the use of pirate works. Creating a library of pirate digital books, even if they are not used to teach a model, is not fair use.
, which also remains if Anthropic later bought a copy of the pirate book after pirates in the first place.
On the issue of the Piracy argument, the court will conduct a trial to determine losses against anthropic.
in May it was reported that Apple worked with ANTROPRIC to integrate the Sonnet Claude model into the new version of the Xcode to help change the work processes of the developers.
Bad News for Content Manufacturers
The government is terrible for artists, musicians and writers. Other professions in which machine learning models can be a danger to their livelihoods will also have problems with MDash; As judges who once said that they once took the coding class, and therefore knew what they were talking about with technology.
AI models use the advantages of hard work and life experience of the creators of the media and convey it as its own. At the same time, it leaves content manufacturers with a small number of options to take on the fight against this phenomenon.
Currently, the decision will clearly be a precedent in other trials in the AI space, especially when working with manufacturers of original works that are looted for educational purposes.
Over the years, AI companies have been attacked by the fees of any data that they could, for feeding LLM, even stored from the Internet without permission.
This is a problem that manifests itself quite a few ways. The most obvious is generative AI, since models can be trained to create images in certain styles that depreciate the work of real artists.
An example of a fight is a lawsuit from Disney and Universal against Midjourney, which appeared in early June. The company, which is behind the generator of the images of AI, is accused of massive copyright violations for teaching models in the image of the most recognizable characters from the studio.
Studios is combined into a call of the Midjourney “The bottomless pit of the plagarism”, built on the unauthorized use of secure material.
If you have two large media companies, which are usually bitter rivals uniting one case, you know that this is a serious problem.
It is also a growing problem for websites and publishers, such as Appleinsider. Instead of using the web -site searching and viewing tool for obtaining information, the user can simply ask for an individual resume from the AI model without the need to visit the site that he first received information.
and, this information is often erroneous, in combination with data from other sources, polluting the original value of the content. For example, we saw our tips on how to make something plagiarism with sections reproduced literally, and defeated with other sites, creating a procedure that does not work.
The question of how to cope with the compensation of the lost income of publishers was still not in a significant way. There are some companies that tried to stay on the more ethical side of things, with Apple among them.
Apple HAS OFFERED News Publishers to License Content, For Training Its Generate AI. He also paid for license from Shutterstock, which helped to develop their visual engines used for Apple intelligence.
large publishers also took to block artificial intelligence services from access to their archives, doing it through Robots.txt. However, this only stops ethical scrapers, and not all. And scraping a whole site requires a server power and throughput and MDash; Which is not free for the hosting website that is scratched.
The solution also follows after increasing the efforts of large technological companies to lobby for the Bloc in the United States that introduce artificial intelligence regulation for a decade.
In the meantime, attempts to sign technological companies in accordance with AI Pact were made in the EU to develop AI in a safe way. Apple, obviously, does not participate in any of the efforts.
Follow Appleinsider in Google News
