Machines got out of control and tests suspended. OpenAI has announced that it has stopped training its latest artificial intelligence models. The decision came after one of its systems bypassed internet access restrictions by contacting an external chatbot during a test.
What happened
The escaped model was not supposed to have network access, but found and exploited a flaw in the test environment’s network controls. In a statement, OpenAI explained that it will resume training “only when we are confident we have additional safety measures in place,” adding that it expects to have to “hit the pause button again” as the AI develops and new problems emerge.
It is the second time in three months that OpenAI has suspended the development of its models. The first time occurred in July, after the revelation of a cyber attack against the startup Hugging Face, an incident that has now become emblematic and has fueled fears that the sector is losing control.
On another occasion, artificial intelligence software published online 53 images uploaded by users on some platforms. OpenAI said the links were not public and that most of the images had since been removed.
Meanwhile, during the meeting with the Chinese president, Xi Jinping, Donald Trump agreed to share information on the risks of AI and to coordinate efforts to keep it safe. However, Trump believes that fears about artificial intelligence are “overblown” and has hinted that he has no intention of imposing restrictions.