A recent survey by AI Impacts involving over 1,500 leading researchers reveals that the average probability of AI causing human extinction or permanent disempowerment is estimated at 18 percent as of 2024. Many participants provided estimates exceeding 10 percent, with some suggesting a 100 percent likelihood.
Researchers have consistently accelerated their timelines for achieving human-level AI performance in surveys conducted since 2016. Furthermore, 57 percent of respondents believe it is unlikely that users will comprehend the reasoning behind AI decisions by 2029, citing increasing system complexity.
Industry experts, including former OpenAI researcher Daniel Selsam and former Google DeepMind employee Bilal Chughtai, have publicly raised concerns regarding existential risks. Selsam highlights the development of situational awareness in models, while Chughtai emphasizes the rapid pace of development since 2022.
The survey identifies AI-driven misinformation as the primary concern for the next 30 years. Other significant risks include the manipulation of public opinion and the potential for dangerous groups to access powerful tools. Consequently, researchers are calling for increased transparency and safety research.
