Yann LeCun originated none of his ideas
3 days ago
- #Yann-LeCun
- #AI-criticism
- #media-misrepresentation
- Yann LeCun has been falsely portrayed as the originator of key AI ideas like convolutional neural networks (CNNs), critiques of large language models (LLMs), and world models, despite not originating any of them.
- LeCun's PR campaign, supported by Meta and the media, has successfully painted him as a lone genius, ignoring predecessors like Kunihiko Fukushima (CNNs) and critics like Gary Marcus (LLMs).
- LeCun rarely credits prior work, engaging in 'plagiarism of ideas' by appropriating others' theories without acknowledgment, as documented by researchers like Schmidhuber.
- His critiques of LLMs and scaling hypotheses were preceded by others, including Gary Marcus and Emily Bender, yet LeCun presents these as his original insights.
- LeCun's advocacy for commonsense reasoning and world models ignores decades of prior research by figures like John McCarthy, Ernest Davis, and Herb Simon.
- The Wall Street Journal's portrayal of LeCun as a lone skeptic of LLMs is misleading, as surveys show many researchers share doubts about LLMs achieving AGI.
- LeCun's new startup focuses on world models, an idea with roots in the 1950s, yet he rarely cites pioneers like Schmidhuber or Fei-Fei Li, who are also working on similar concepts.