To improve how their AI models learn, Anthropic and Microsoft-backed (MSFT) OpenAI have focused on two main strategies. First, they train models using fake versions of apps, called “gyms” or ...

In a new report, OpenAI said it found that AI models lie, a behavior it calls “scheming.” The study performed with AI safety company Apollo Research tested frontier AI models. It found “problematic ...

Introduction The COVID-19 pandemic led to major disruptions in society across many spheres, including healthcare, the economy and social behaviours. While early predictions warned of an increased risk ...