AI pioneer Andrew Ng says renewed warnings that advanced AI could pose an existential threat to humanity are “science fiction,” not evidence-based science. Speaking in multiple interviews, Ng argues that such fears are exaggerated and may distract from practical work to improve AI reliability and safety.
Ng says AI systems will never be perfectly predictable, but that safety can improve through engineering, testing, and safeguards applied over time. He also cautions that broadly slowing AI development could slow progress on safety rather than reduce risk. Some outlets report he believes the industry’s focus on extreme scenarios can be driven by publicity and attempts to influence regulation.
While Ng emphasizes engineering and testing as key approaches, his comments come amid ongoing public debate over “loss of control” and other potential engineering risks as regulators consider new rules for AI systems. Reporting references Europe’s AI Act framework, which includes both engineering-related and loss-of-control concerns, highlighting the policy context for differing viewpoints.
