Why Scale Will Not Solve AGI | Vishal Misra - The a16z Show
The guest argues that current LLMs are fundamentally limited to correlation-based pattern matching (Shannon entropy) and Bayesian updating, lacking the causal reasoning and plasticity required for AGI.summaryTrue AGI requires moving from static weights to continual learning and developing causal models (Kolmogorov complexity) that allow for simulation and intervention, rather than just predicting the next token.summaryThe 'Einstein Test' for AGI proposes training an LLM on pre-1916 physics and seeing if it can derive the theory of relativity; this tests the ability to create a new causal manifold rather than just interpolating existintakeaway
