OpenAI engineers knew it: they secretly admit that they are sinking the media with their AI
The New York Times lawsuit against OpenAI and Microsoft is raising eyebrows in the AI sector. And it’s no wonder. The judge has begun to declassify internal documents from the trial. Therefore, we now know that the engineers of these companies themselves had been aware of their lack of ethics for years. Basically, they debated among themselves whether taking millions of press articles without asking permission was legal or if they were directly committing the greatest job theft in the history of humanity.
All these messages that are coming to light are from between 2020 and 2024, just the time when the language models were about to hatch. The curious thing about this whole topic is the hypocrisy. On the one hand, managers went out to say in public that using copyrighted texts was totally legal, hiding behind “fair use” (fair usein English). But inside, behind closed doors, the technical investigators warned that all this could be a disaster. He even recognized that they were attacking the professionals who created the information they needed to train the models.
Worried about destroying their own supply chain
Brent Hecht, who was director of applied science at Microsoft, wrote in 2023 that they were creating a product designed to destroy their own supply chain. That is, they knew that LLMs could destroy the media business model and that, in turn, would harm them because they would not have new training data. On the other hand, Jack Clark, former policy director of OpenAI, told Sam Altman and Greg Brockman in a message back in 2020 that all this was going to make them the symbol of the typical invasive Silicon Valley company that doesn’t care about anything.
On the other hand, Nick Turley, who was the head of the ChatGPT team, blurted out in a June 2023 report that his technology was a death threat to the media. Because? Well because people you no longer need to go to the original source of the news. The engineers knew and assumed perfectly well that no one clicks on the external links that the AI uses as a source. They assumed that they are making a product that replaces the journalist and that steals the clicks of those who live. By the way, something very similar to what Google does with AI Mode.
As if that were not enough, the trial paperwork has also revealed that they knew how to bypass newspaper protections to read paid articles. There is a message where an OpenAI employee tells the president, Greg Brockman, that they have managed to avoid the paywall of the New York newspaper. His reaction was to give the go-ahead.
Now Microsoft is trying to throw things out as best it can. Satya Nadella has had to testify before lawyers that, of course, licenses must be paid for subscription content. And their spokespersons excuse themselves by saying that what their scientists think in an internal chat is not the company’s official position.
