Walter G. Wasser

Managing Extreme AI Risks Amid Rapid Progress

Introduction: As artificial intelligence (AI) advances rapidly, it’s imperative to address the extreme risks associated with these developments. A recent article published in Science by leading experts in the field outlines the urgent need for technical research, development, and adaptive governance to manage these risks effectively.

RAPID PROGRESS, HIGH STAKES: AI technology is evolving at an unprecedented rate, with companies racing to develop generalist AI systems capable of matching or surpassing human abilities in various cognitive tasks. This progress promises immense benefits, such as curing diseases and elevating living standards, but it also poses significant risks, including large-scale social harm and the potential loss of human control over autonomous AI systems. The stakes are incredibly high as these systems could amplify social injustices, enable criminal activities, and facilitate automated warfare and surveillance.

REORIENT TECHNICAL R&D: To manage these risks, we need to reorient our technical research and development efforts. Current deep-learning systems lack crucial capabilities, and more resources are required to ensure AI safety and ethical use. Key research areas include improving AI oversight, robustness, interpretability, and inclusivity. Addressing these challenges will require dedicated R&D and significant breakthroughs to ensure AI systems do not pursue unintended or harmful goals.

GOVERNANCE MEASURES: Effective governance is crucial for managing AI risks. National and international institutions must enforce standards to prevent misuse and recklessness. Current governance frameworks are inadequate and must evolve to keep pace with rapid AI advancements. Policies should automatically trigger stricter requirements when AI hits certain capability milestones, ensuring proactive risk management. Governments need comprehensive insight into AI development, mandatory risk assessments, and robust safety measures to mitigate potential harms.

Conclusion: Managing extreme AI risks requires a combination of technical research, proactive governance, and adaptive measures. With the right approach, AI can be steered towards positive outcomes, avoiding catastrophic risks and ensuring a safer future for all.


  • Bengio, Y., Hinton, G., Yao, A., Song, D., Abbeel, P., Darrell, T., Harari, Y. N., Zhang, Y.-Q., Xue, L., & Mindermann, S. (2024). Managing extreme AI risks amid rapid progress. Science, First Release. DOI: 10.1126/science.adn0117.
About the Author
The author is a specialist in nephrology and internal medicine and lives with his wife and family in Jerusalem.
Related Topics
Related Posts