The Optimal Choice of Hypothesis Is the Weakest, Not the Shortest

If A and B are sets such that $$A \subset B$$ , generalisation may be understood as the inference from of a hypothesis sufficient to construct B. One might infer any number hypotheses A, yet only some those generalise How can one know which likely generalise? strategy is choose shortest, equating ability compress information with (a “proxy for intelligence”). We examine this in context mathematical formalism enactive cognition. show compression neither necessary nor maximise performance (measured terms probability generalising). formulate proxy unrelated length or simplicity, called weakness. if tasks uniformly distributed, then there no choice performs at least well weakness maximisation all while performing strictly better one. In experiments comparing maximum minimum description binary arithmetic, former generalised between 1.1 5 times rate latter. argue demonstrates far proxy, explains why Deepmind’s Apperception Engine able effectively.