Preprint warns of catastrophic AI risks if no action is taken within five years (news.uq.edu.au)

A recent preprint study conducted by researchers from The University of Queensland and MIT FutureTech highlights alarming risks associated with artificial intelligence (AI), suggesting a significant likelihood of catastrophic outcomes within the next five years. After surveying 272 international experts in AI, the research indicates that there is at least a 10% chance of dire incidents, including over one million deaths and over $100 billion in financial damages. Key risk areas identified include AI-driven cyberattacks, advancements in autonomous weaponry, and failures in governance in the rapidly evolving AI landscape. The findings are particularly significant for the AI/ML community, as they reveal the urgent need for proactive risk mitigation strategies. Experts emphasized that, even with some safety measures, many risks remain troublingly high, jeopardizing national security and democratic values. They warned that without collective action—including laws and treaties—there may be inadequate incentive for organizations to prioritize safety over competitive advancement. The study calls upon stakeholders in AI development to engage earnestly with these risks, reinforcing the notion that AI's integration into critical systems necessitates a robust framework for ethical governance and safety.