# Gary Marcus American cognitive scientist, professor emeritus at NYU, and the most persistent public critic of the claim that scaling [[Deep Learning]] alone leads to general intelligence. He has argued since the 2010s that neural networks lack reliable compositional and causal reasoning and that symbolic structure must be reintroduced — the position now organized under [[Neurosymbolic AI]]. Marcus founded the machine-learning startup Geometric Intelligence, acquired by Uber in 2016. He testified before the Senate Judiciary Committee alongside [[Sam Altman]] in May 2023 on AI oversight, and writes the *Marcus on AI* newsletter, which functions as a running audit of frontier-lab claims against delivered capability. His argument that tool-augmented reasoning models represent an unacknowledged neurosymbolic turn — and that the [[Scaling Laws]] narrative persists because it is what investors want to hear — is documented on the [[Neurosymbolic AI]] page. Distinct from the [[AI Safety]] and [[AI Doomerism]] critiques: Marcus's objection is that current systems are less capable and less reliable than advertised, not that they are on the verge of superintelligence.