- The ChatGPT developer company has asked for a partial dismissal of the NYT’s copyright infringement lawsuit.
- OpenAI representatives believe that the publication hacked GhatGPT to generate fake proofs.
- According to the NYT, this is promt engineering (prompt engineering).
OpenAI has accused The New York Times (NYT) of hacking the ChatGPT chatbot. According to the developers, this was done for the sake of generating fake evidence in a copyright infringement case.
The company assures that the publication paid a third party to hack the language model. After that, journalists together with unnamed specialists used “deceptive prompts” that violated the terms of use of ChatGPT. They used them to get the chatbot to reproduce the publication’s copyright-infringing material.
“The allegations in the NYT’s complaint fall short of its famously rigorous journalistic standards. The truth that will be revealed in this case is that the Times paid someone to hack OpenAI products,” the company said in a statement.
The publication’s lawyers do not deny hiring experts to handle ChatGPT. According to them, it is about promt engineering (prompt engineering). This is the process of creating and optimizing text queries (prompts) for generative models in order to get the desired answers.
A team of experts allegedly examined the chatbot for evidence of the publication’s alleged illegal use of NYT material. Advocates see no problem with this approach and compare it to the so-called “red attack” – a simulation to assess the security of systems.
In November 2023, a copyright infringement lawsuit was filed against OpenAI and Microsoft. The documents allege that the chatbot developers illegally used tens of thousands of articles and scientific materials to teach the language model.
Meanwhile, OpenAI has unveiled a new service for converting text to video. The product, called Sora, is available for artists, designers and filmmakers for now. They should give feedback on how to improve the model to make it as useful as possible, the company said.
Recall, we wrote that OpenAI taught ChatGPT to memorize facts and preferences of users.
