Shannon Got AI This Far. Kolmogorov Shows Where It Stops
The article discusses the differences between Shannon entropy and Kolmogorov complexity in the context of artificial intelligence. It highlights how modern deep learning models optimize for statistical regularities rather than understanding the underlying processes that generate data. This distinction is illustrated through a historical experiment by Michelson and Morley that failed to support the prevailing theory of light propagation.
- ▪Shannon entropy measures the statistics of outputs, while Kolmogorov complexity measures the structure of the generating process.
- ▪Deep learning models are trained to minimize cross-entropy loss, focusing on statistical regularities in data.
- ▪The article uses the failed Michelson-Morley experiment to illustrate the difference between expected and observed results in scientific inquiry.
Opening excerpt (first ~120 words) tap to expand
Shannon Got AI This Far. Kolmogorov Shows Where It Stops.Vishal Misra14 min read·Mar 7, 2026--30ListenShareThis post previewed a conversation I recorded with Martin Casado for the a16z podcast. The ideas here came up in that discussion — consider this the written version.Press enter or click to view image in full sizeA map that knows everything it has seen — and nothing beyond its edge. The equation above it was not found by fitting curves. It was found by asking what kind of universe would make the anomalies disappear.Here is a question that sounds like a trick but isn’t: is the number pi simple or complex?Your intuition probably says complex. The digits go on forever without repeating. No pattern jumps out.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at Medium.
Discussion
0 commentsMore from Medium
