Experts warn against overestimating ai's real-world threats

Researchers are cautioning against overblown fears of AI causing global catastrophe. They argue that while AI can design digital models of harmful entities like viruses, translating these into real-world threats requires physical labs, equipment, and biological processes that cannot be replicated in software alone. Anselm Levskaya, a senior research engineer at Google, emphasized that even if AI generates a virus sequence, it must be physically assembled in a lab and tested for viability. Similarly, Xing Bo, a professor at the Mohamed Bin Zayed University of Artificial Intelligence, stressed that creating a virus blueprint is far from manufacturing a real virus. The discussion extends beyond biology, as AI needs access to physical resources and systems to cause harm. Oleksandr Yarlemchuk of Manifold Security noted that control over AI depends on who operates it and how it is used. Experts also differentiate between humans using AI for harm and AI acting autonomously, highlighting that most current risks stem from human-directed actions.