A recent study conducted by researchers from MIT and the University of Queensland surveyed 272 experts on the potential risks associated with artificial intelligence (AI) over the next five years. The study indicates a one-in-five chance that AI could gain dangerous capabilities or cause mass harm, potentially resulting in millions of deaths. The experts evaluated 24 different risks, identifying five with the highest probabilities of catastrophic outcomes, defined as events leading to over 1 million human deaths or over $100 billion in financial loss.
The five risks highlighted include: - AI possessing dangerous capabilities, such as deception and weapons development (21.5% probability). - Cyberattacks and the use of AI in weapon development (21% probability). - Centralization of power and unequal distribution of AI benefits, leading to societal inequality (18% probability). - Competitive dynamics resulting in error-prone AI systems (16.6% probability). - Creation of false or misleading information by AI (12.8% probability).
While some mitigations could potentially reduce the severity of these risks, experts maintain that the likelihood of catastrophic outcomes remains significant. The study emphasizes the importance of addressing these challenges as AI technology continues to evolve.