Meta has reached an agreement Reuters that will allow the technology company to train its chatbot of artificial intelligence (AI) with the content generated by the news agency. The licensing agreement is the first of its kind that the organization headed by Mark Zuckerberg closes. The alliance comes at a time when developers of large language models are experiencing a shortage of training data.
The association was ratified by the parties involved. “We can confirm that Reuters is collaborating with technology providers to license our trusted, fact-based news content to power their AI platforms. The terms of these contracts remain confidential,” notes Heather Carpenter, spokesperson for the information service. For his part, Jamie Radice, spokesperson for Meta, assures that thanks to this partnership “Meta AI can answer questions about current events with summaries and links to content from Reuters”.
The portal Axios reports that the pact will last “several years.” Add that Reuters You will receive financial compensation for the content used in each response by Meta’s AI assistant. He bot Conversations can be used through Facebook, Instagram and WhatsApp. Materials generated by the news agency’s editors will begin to appear for Meta AI users located in the United States. It is unknown if the possibility will reach other markets.
Meta seeks to guarantee the training of its AI
OpenAI has signed multiple similar agreements in the last year with various media companies. In July of last year, it reached an agreement with the Associated Press agency and five months later it signed a contract with the German conglomerate Axel Springer to use information published on sites such as Business Insider and Political. In April of this year, it partnered with Financial Times and 15 days later he closed a deal with Prisa Media, which handles titles such as El País, Cinco Días, AS and HuffPost. This month, it struck a pact with Dotdash Meredith, the largest publisher in the United States and owner of more than 40 media brands, including Investopedia, People and Food & Wine. In August of this year, it signed an agreement with the Condé Nast group, publisher of Wired, Vogue and GQ among other publications.
Analysts suggest this is the result of a widespread crisis in the AI sector. They indicate that the major players in the emerging industry are on the verge of exhausting all the training information publicly available on the internet. Most materials will be accessible through paywalls or subject to exclusive use contracts.
Data Provenance Initiative, a collective led by the Massachusetts Institute of Technology, indicates that there is “an emerging crisis in consent for use.” Publishers and online platforms are taking stricter measures to prevent their publications from being used to train products like Gemini or GPT. Consequently, they have restricted 5% of all previously public content. 25% of the highest quality data has become inaccessible. Organizations like Google have attempted to respond to the lack of quality public information with synthetic data. But the reliability of the procedure is in doubt.
#Meta #trained #content #Reuters #news #agency