Artificial intelligence versus journalism What do internal documents reveal about the future?

Artificial intelligence versus journalism What do internal documents reveal about the future?

A set of documents related to.  against the American artificial intelligence company OpenAI revealed concerns and doubts within the company's management and its partner Microsoft about using the newspaper's news articles to train models.

The Federal Court for the Southern District of New York had made available a set of documents that were previously confidential as part of the lawsuit filed in December 2023, in preparation for the start of the trial and its ruling.
According to the new documents, Microsoft employees expressed their fear that OpenAI is carrying out what they described as the biggest labor theft in history, indicating the possibility that the company is entering a vicious cycle of destruction that will reduce the quality of content and materials used in training models.

The report also noted concerns within OpenAI about the negative impact of its ChatGBT tool, as management feared that people would turn away from reading newspapers and websites, according to an internal memo in June 2023.

As Microsoft’s head of applied science, Brent Hecht, stated in an internal presentation in January 2024, the AI ​​tool “Copilot” caused a 93% drop in traffic to the New York Times website, according to an independent.

For its part, Microsoft, through its spokesman Alex Horick, clarified in a statement that this presentation and the internal memos written by Hecht do not reflect the company's viewpoint, stressing that the company's official position is on the court records.

Microsoft CEO Satya Nadella takes a position directly opposed to his partner OpenAI, explaining in testimony at a hearing earlier this year that training models on exclusive, subscription-based content requires a special license.

He added that if he had known that OpenAI had used content of this type in training its models, he would have requested that the models be retrained without this content based on the partnership agreements between the two companies.

A destructive cycle that harms models
Hecht argues that artificial intelligence tools are harmful to the economic model adopted by the majority of newspapers and media sites in the world, since their use may allow users faster access to content without the need to visit newspapers and sites.

This harms media outlets that rely on paid subscriptions to access exclusive content, and even outlets that rely on advertisements that appear on their websites.

Ultimately, this damage causes newspapers and media outlets to disappear and online content to weaken, preventing models from accessing high-quality content for future training.
 

Post a Comment

Previous Post Next Post

Secret Island Game