Since the proliferation of large language models and generative AIs, a serious battle has emerged between publishers and AI developers. OpenAI CEO Sam Altman is one of the parties to these discussions ChatGPT He stated that such tools cannot be developed without copyrighted content.

According to a recent study, content on the Internet below 57 percent It has the signature of artificial intelligence. According to a study by researchers from Cambridge and Oxford, artificial intelligence is overloaded and the quality of content is deteriorating.

Copy of copy…

In their study, researchers revealed that the answers given by artificial intelligence get worse every time and cannot provide accurate information. from the University of Oxford Dr. Illa Shumailov, “model collapse It is surprising how quickly it starts and how hard it is to notice. First, it affects minority data, such as underrepresented data. Then, it affects the diversity of the outputs and reduces variance. Sometimes, a small improvement is observed in the majority data, masking the performance decline in the minority data. The collapse of the model can have serious consequences.” he said.

Researchers are now trying to determine the cause of the problem. Per team first cause of the problemsharing AI-generated content without control. Of course, the AI’s “manufacturing” of non-existent information, constantly churning out new copies of lower-quality copies, only seems to make the problem worse.