Certainty–Scope Conjecture
For every artificial-intelligence mechanism , its certainty and scope satisfy a universal tradeoff of the form , where is a constant independent of . Equivalently, no mechanism can simultaneously attain maximal reliability on every input and unrestricted breadth over unstructured data.
References
- A Limit of AI, I: A Discrete Information-Theoretic Proof of the Certainty–Scope Conjecture — Minds and Machines — Yiyang Jia, Alberto Messina, Luciano Floridi
Source: Minds and Machines
Progress summary
A 2026 paper claims to prove a sharp finite version of the conjecture, but the unrestricted claim remains open and another paper claims to disprove its original universal formula.
Floridi’s 2025 conjecture proposes a tradeoff between certainty and scope , expressed as . The original formulation was explicitly presented as unproved.
Known results
- Floridi, 2025: formulated the certainty–scope tradeoff and its two principal corollaries, but supplied no proof. [Source: https://arxiv.org/abs/2506.10130]
October 2026 finite-case proof claim
Jia, Messina, and Floridi claim an exact finite-classification theorem: a Fano-type converse bounds certainty by scope and mechanism capacity, with the -ary symmetric channel attaining equality; they also claim extensions covering composition, side information, post-processing, tools, routing, and context-wise feasibility. The claim establishes only a finite model, not the unrestricted conjecture, and has no independent assessment in the retrieved sources. Separately, an arXiv paper claims that the original universal product bound fails under the published definitions, including with Kolmogorov-complexity and Shannon-entropy variants; this disproof is likewise unverified.
Current status (as of October 2026): the finite-classification result and the claimed disproof of the original universal formulation are both unverified; the unrestricted Certainty–Scope Conjecture remains open.
Solutions 0
No solutions have been posted yet.