UK national broadcaster BBC has announced it is drawing up plans to create its own artificial intelligence model based on its vast text archives, hoping the technology will help improve its production process.
Earlier this month, the BBC’s director of nations Rhodri Davies told a House of Lords committee that the broadcaster was considering developing an AI program, but was still deliberating as to whether to develop it unilaterally or in partnership with another company.
On Thursday, another spokesperson for the company confirmed the plans in a statement cited by Reuters, saying that the BBC was indeed looking to use its 80-years worth of audio, video and text content to train the program.
The broadcaster stressed, however, that the output of such a model would not be shared with any third parties and would only be used internally.
However, Financial Times has reported, citing sources within the organization, that the BBC was reportedly in negotiations about selling access to its large library to outside companies to help them develop their own language models.
The BBC has denied the claims, stating that it “has no agreement with any organization to use its archive to train their large language models that power generative AI tools for commercial use.”
The company’s statement comes as a number of large media organizations have already accepted lucrative deals to sell their journalistic archives to the likes of OpenAI to help train their AI programs.
In another statement issued last month, the BBC also stated it was already experimenting with the use of artificial intelligence in its newsroom, but had established certain rules to ensure that anything published would still have human oversight.
As an example, the broadcaster said that it has used AI to come up with summaries to articles and recommendations for headlines for a story. The BBC stressed, however, that the final decision as to what to publish still rests with journalists.