Fascinating Guardian profile of Song-Chun Zhu, a leading AI scientist who left UCLA for China after nearly three decades in the US. He argues that today’s AI, dominated by large neural nets and benchmark-driven research, has drifted away from the deeper pursuit of intelligence — reasoning, causality, social and physical understanding. Zhu says China offered him resources and freedom to chase these harder, less fashionable questions, while US academia felt increasingly constrained by political pressure, funding structures, and a preference for “safe” incremental work.

Raises two questions: Are we losing diversity of scientific thought by letting scale-driven AI monopolize the field? And what does it mean for global research leadership if the next big paradigm shift in AI comes from outside the US?

Fiddle, I was ninja'd by seven hours. But yes, great article, although it's light on what he thinks should replace and/or complement neural nets (including, but not limited to LLMs). But given The Guardian's audience, that's natural.

Kind of a thinking man's Gary Marcus. Skeptical of LLMs but actually expert in AI.

A long and seriously complex story.