The British public broadcaster BBC plans to use its extensive archive of audio, video and articles, built up over decades, to train its own AI model. This can then be used to control other internally developed AI applications. This is reported by the Financial Times business newspaper. Last modified: March 22, 2024, 1:44 PM
Top manager Rhodri Davies told a government committee this month that the BBC may bring in an external partner but may carry out the entire process independently, the FT said.
At the same time, talks are said to be underway to license BBC content to companies such as Amazon, which are working on their own AI models. Davies: “We of course have to look at the costs. But it is clear: the BBC has almost 80 years of audio and video archives, as well as a huge text archive online.”
The BBC would not be the only publisher that can monetize its own archive. Last year, Axel Springer signed an agreement with OpenAI, the company behind ChatGPT, for the reuse of pieces. OpenAI is in discussions with other companies with large archives that can serve as training data.
At this time, the BBC has no agreement with any AI companies to allow the archive to be used as training data, a spokesperson said. More at Financial Times
