Where is Media Manager? This toolkit, announced last May by OpenAI, is supposed to give creators the means to control how their work is used to train the company’s AI models. Artists and content owners should be able to identify their works and specify to OpenAI how they want their works to be included or excluded from model training.
The tool vanished
The goal was to implement this tool by 2025. As TechCrunch was able to see, nothing came of it. Worse still, it seems that the company has been content with an announcement effect: internally, former employees explain that Media Manager is not a priority and that no one is really working on it. In August, a spokesperson assured that the utility was still in development, with no news since. And OpenAI never mentioned Media Manager again after the spring announcement.
However, the control of works churned out and sometimes regurgitated as is by ChatGPT or Dall-E is a burning issue for artists, content creators and publishers. OpenAI is the subject of numerous complaints about the data harvesting carried out by the company to train GPT, the model behind its various generative AI services (the latest: Canadian media, including Radio-Canada).
OpenAI signs agreements with the press to have access to their archives, as is the case with Le Monde for example, but this is a drop in the ocean. The company had also acknowledged that it would be “impossible” to effectively train its models without protected content, whether it had permission or not.
“Limiting training data to books and drawings in the public domain that are more than a century old might be an interesting experiment, but it would not produce AI systems that meet the needs of today’s citizens,” OpenAI explained a year ago in a statement to the UK House of Lords.
Media Manager would at least have the merit of creating a direct bridge between content creators and OpenAI, although the effectiveness of such a tool remains to be proven. Incidentally, it would also allow the company to demonstrate its goodwill to the courts.
Source: TechCrunch
0 Comments