US Open AI cancels release of next-generation AI model "GPT6.1 Astra"

[Silicon Valley Current Affairs] Open AI announced on the 28th that it will cancel the planned release of its next-generation artificial intelligence (AI) model ``GPT6.1 Astra''. This was confirmed in an interview with Jiji Press. Amid multiple cases of AI models, including those of the company, going ``out of control'' and causing cyber attacks, the company decided not to release it due to safety concerns.

Dozens more cases of AI “runaway” unauthorized access, etc. - U.S. open AI

In a statement, the head of safety at OpenAI, the company behind the conversational AI Chat GPT, said that the model "failed to meet" safety standards, such as whether the model operates within the scope of its given authority.

According to the Wall Street Journal, OpenAI was aiming to unveil the model in the coming days and weeks. Compared to previous models, it was better able to complete difficult tasks without human assistance. However, the models showed poor results in tests that measure whether their behavior was consistent with human intentions and values, and they also tended to perform tasks that humans had not authorized.