Axios: OpenAI suspends AI training due to uncontrolled behavior of models // Axios reports suspension of training of OpenAI AI models
OpenAI has suspended the training of its most powerful AI models and will resume it when it is confident that additional security measures are in place. The company also wants to make sure that the neural network will be guided by human norms when performing tasks.
OpenAI has suspended the training of its most powerful AI models and will resume it when it is confident that additional security measures are in place. The company also wants to make sure that the neural network will be guided by human norms when performing tasks. This is reported by Axios with reference to sources.
“This is not the first time we have taken a pause to implement such measures, and we do not think this time will be the last as AI capabilities grow,” the company said.
In recent days, OpenAI and third-party researchers have uncovered several major incidents related to the behavior of the company’s systems. Among them are the leak of 53 images of ChatGPT users, the hacking of the Australian government website and the portals of US government agencies.
OpenAI CEO Sam Altman said this week that the company "wasn't as fast as we'd like" in uncovering these incidents. According to him, the hack of the resource Hugging Face, which occurred in July, remains the most serious such incident detected by the company.
