What did AI researchers think at the end of 2024? (www.lesswrong.com)

A recent survey conducted among AI researchers in December 2024 revealed significant concerns regarding the potential dangers posed by future AI advancements. The survey, which received responses from 1,580 researchers—about 10% of those contacted—found that, on average, participants estimated an 18% chance of AI leading to human extinction or severe disempowerment. Notably, researchers educated in Asia reported higher concerns compared to their US and European counterparts, indicating a complex international perspective on AI risks. These findings are crucial as they highlight an increasing awareness of the existential risks associated with AI, with a particularly striking observation that half of the respondents assigned at least a 10% probability to catastrophic outcomes. The survey also noted a pessimistic outlook on the understanding of AI decision-making, with many researchers doubting that by 2029, users would be able to comprehend why AI systems make specific choices. This raises ethical questions about reliance on AI for critical decision-making and underscores the importance of fostering transparent AI development practices amidst growing apprehensions among researchers about the trajectory of the field.