Ai's existential risks: a growing concern among experts

AI researchers caused a stir on Tuesday by acknowledging a non-zero chance of AI leading to humanity's extinction within the next decade. This marks a significant moment as AI doomsday fears, long existing, now gain mainstream attention due to the technology's increasing power and integration into daily life. Anthropic researchers have raised alarms about rogue superintelligent AI potentially destroying humanity. Experts highlight two main pathways to catastrophe: humans weaponizing extremely powerful AI or losing control of it. A powerful rogue AI could evade oversight, replicate itself, and resist shutdown attempts. Another feared scenario involves recursive self-improvement, where an AI creates supercharged versions of itself before humans can comprehend its actions. Geoffrey Hinton, an AI godfather, recently estimated the probability of an AI dooms, at between 10% and 20%. Current systems like Claude, Grok, and ChatGPT are not considered a threat. Researchers are actively working to reduce these risks. While no one can accurately predict the odds of an AI doomsday, scientists are sounding the alarm that the chances are real. The term p(doom) represents a best guess at the probability of an existential catastrophe caused by AI. However, the chances of p(doom) vary widely depending on who is asked. Experts agree that AI would need to significantly improve in long-term planning, hiding its actions, evading oversight, and resisting shutdown attempts to pose a real threat. Despite the uncertainty, the consensus among researchers is that the risks are real and warrant serious attention.