The False Glorification of Yann LeCun (garymarcus.substack.com)

AI critic Gary Marcus published a broadside accusing Yann LeCun of running a decade-long PR campaign—supported by Meta and amplified by the press—that overstates his originality and sidelines predecessors. Marcus contends a recent Wall Street Journal puff piece tied to a LeCun startup exemplifies this mythmaking. He argues LeCun is widely credited for five main threads—convolutional neural networks, critiques of LLMs and pure scaling, advocacy for commonsense/physical reasoning, and world models—but in each case predecessors and parallel contributors (e.g., Kunihiko Fukushima and Wei Zhang on early CNNs; Emily Bender and Marcus himself on LLM limitations; Herbert Simon, Jürgen Schmidhuber, Fei‑Fei Li, Ernest Davis on world models and common sense) were often omitted or unacknowledged. For the AI/ML community this matters beyond personality: attribution shapes scientific credit, research priorities, funding, and public narratives about progress and risk. Marcus’s critique underscores enduring technical debates—whether scaling alone yields AGI, the need for integrated world models and commonsense reasoning, and the importance of neurosymbolic approaches—and warns that corporate PR can distort those debates. Whether LeCun’s new venture will produce genuinely novel advances remains open, but Marcus’s piece is a prompt to correct the historical record, better cite antecedent work, and guard against singular narratives steering research and policy.