The Aaronson Oracle
Summary
The article explains how a simple n-gram frequency model can predict user presses better than random by counting five-press runs and backing off when data is sparse. It traces the lineage from Hagelbarger’s SEER to Shannon’s mind-reading machine to Scott Aaronson’s demonstration and discusses why humans misperceive randomness and how the method highlights the difference between free will and algorithmic prediction. It emphasizes backoff and context length as core ideas in statistical modeling.