Infinities, impossibilities, and the man in the white linen suit
Summary
An in-depth Exploration of Gödel and Turing’s legacies and their implications for modern AI. The piece traces the history from Hilbert to Gödel and Turing, discusses the Gödel and Darwin Gödel machines, and argues that formal guarantees for self-improving AI are mathematically impossible, emphasizing empirical validation and safety considerations. It also covers learnability limits, neural network reliability issues, and the broader sociotechnical impact of these foundational results on AI development.