In Tests, GPT-4 Strangely Itchy to Launch Nuclear War

A team of Stanford researchers tasked OpenAI's latest large language model to make high-stakes, society-level decisions in a series of wargame simulations — and it didn't bat an eye, recommending the use of nuclear weapons. It's a worrying indication that, without any major human oversight, AI models could eventually have the ability to sway policymakers […]

ai models artificial intelligence bat decisions eventually gpt gpt-4 human language language model large language large language model launch major nuclear nuclear war nuclear-weapons openai oversight researchers series simulations society stanford team tests war weapons