OpenAI announces agreement with Prisa and Le Monde. It is evidence of an upward trend

When ChatGPT was released on the market, no one wondered how to train it. OpenAI has never been transparent about this, but in the face of mounting copyright infringement lawsuits, the company has begun signing agreements with content providers. The latest concerns the Spanish publishing group Prisa Media (including El País and As or Cinco Días), as well as the French newspaper Le Monde.

A wonderful alliance. he Official announcement The OpenAI project joins those already implemented in July 2023 with the Associated Press (AP) or in December With Axel Springer. According to the text, the agreement “will allow ChatGPT users to interact with high-quality content from Le Monde and Prisa Media about recent events on ChatGPT.”

But they are intended for one purpose: training your models. In this announcement, the real reason behind these agreements is left a little in the background: so that OpenAI can leverage the vast content of Prisa and Le Monde to train its models. This aspect is briefly mentioned in the OpenAI statement, which emphasizes that “its contents will contribute to training our models.” News from El Pais The agreement does not even mention this section, which they clearly comment on. In Le Monde.

What we will notice in ChatGPT. According to OpenAI, in the coming months ChatGPT users will receive responses that will provide attribution links and improved links to original articles, so that information sources can be easily accessed. It's something, for example, we've already seen gradually in chatbots like Copilot – based on GPT-4, from OpenAI – or in Perplexity, which include links to sources from which information is extracted.

But the war with the New York Times continues. These agreements are important, but OpenAI has a major thorn in its side: The New York Times, the world's most important newspaper, which in December sued Microsoft and OpenAI for exploiting its content without approving or sending traffic to them.

At OpenAI they don't want any more lawsuits. The company, led by Sam Altman, said the lawsuit Lacks foundationAnd just a few days ago Microsoft trot out In federal court to dismiss part of the lawsuit. The process is still pending, and there is no doubt that these agreements with these large publishing groups could be a way to protect themselves from new lawsuits similar to the one brought by The Times.

An upward trend. The economic terms of the agreement have not been detailed, and those that have already been concluded in recent months have not been revealed, but at least in the case of Axel Springer Yes, it has been mentioned The agreement is valid for “several years.” Other companies reach similar agreements so they can train their models without fear of legal retaliation. The best example of this is Google, which at the end of February paid Reddit $60 million for this goal.

Image | Roman Craft

In Chataka | Figure 01 makes “Tesla Bot” look silly: This amazing human-like robot is capable of interacting with humans