Published October 17, 2026
| Version v1
Preprint
Open
Computing Machinery and Understanding: From the Imitation Game to the Turing
Description
In 1950, Alan Turing replaced the question "Can machines think?" with an operational experiment, the imitation game. This paper proposes the same move for a contemporary question: does a machine understand? I propose that one measurable component of understanding is the discovery of compact structure that continues to predict previously unseen observations, combining Shannon information, predictive coding, Minimum Description Length, and physical resource accounting. I define Turing gain. One Turing (1 T) is one bit of net held-out explanatory compression: the reduction in held-out total description length produced by a model, program, theory, or abstraction relative to a declared baseline, after paying the description cost of the explanation itself. For baseline B, candidate M, sealed held-out data D, declared shared background G, and measurement profile P: T(M;B,D,G,P) = [L(B|G)+L(D|B,G)] - [L(M|G)+L(D|M,G)]. A theory can score negative Turings: an explanation that costs more complexity than the structure it captures fails, in public, by construction. The claim is deliberately narrow. The Turing measures one operational component of understanding, its predictive and explanatory-efficiency dimension, not understanding itself; it does not measure causal direction, semantic grounding, agency, or value alignment. Paired with physical measurement it gives derived quantities such as Turing rate (T/s) and Turing yield (T/J), addressing a blind spot: the field has no standardized measure of net held-out explanatory compression per joule. Preregistered criteria are given at two levels, measurement validity and scientific falsification, stating what would fail the claim. Turing taught us to replace an unanswerable question with an operational one. I ask whether machines can discover explanations that make unseen reality less surprising.
Files
computing-machinery-and-understanding.pdf
Additional details
Related works
- Is identical to
- https://github.com/aien-dev/aienos.com/blob/main/public/research/computing-machinery-and-understanding/paper.pdf (URL)
- Is supplemented by
- 10.5281/zenodo.23437568 (DOI)