RT Rohan Paul: Yann LeCun's (@ylecun ) latest talk at ETH Zürich Scaling LLMs to reach AGI is "impossible" A large language model is trained on about...