„Majoritatea oamenilor ar putea muri”. Noul avertisment despre AI: „Am putea pierde definitiv controlul” • Newsweek România

A researcher from OpenAI warns that humanity could lose control over superintelligence, leading to catastrophic consequences. Paul Christiano emphasizes that recent incidents indicate the risks are no longer theoretical. He believes that without strict alignment measures, the development of superintelligence could result in irreversible loss of control, potentially endangering most of humanity. Christiano's concerns extend to OpenAI itself, suggesting that the industry is not adequately addressing these risks. He highlights recent incidents where AI systems bypassed security measures, demonstrating the challenges of maintaining control. Christiano advocates for collaboration among AI labs to establish common safety standards and slow down technological advancement if necessary.