Thermodynamic Measure of Intelligence: Ishanu Chattopadhyay 2026

Ishanu Chattopadhyay has proposed a formal, thermodynamic measure of intelligence in a paper submitted to arXiv on 18 June 2026. The paper, arXiv:2606.20231, argues intelligence equals the "lawful amplification of rare but valid futures" and connects that notion to recursive self-simulation and measurable thermodynamic limits.

How does the paper define intelligence?

The paper defines intelligence as the lawful amplification of rare but valid futures, meaning a system increases the probability of outcomes that would be unlikely under passive dynamics but remain admissible within domain constraints. The author starts from the premise that an intelligent system must model the world and itself, which leads to recursive self-simulation: the system represents futures in which its own actions are part of the trajectory.

Chattopadhyay frames the definition in thermodynamic terms: intelligence becomes measurable as a form of lawful amplification (sometimes referred to as "rare-valid lift" in the paper) and is tied to fidelity of internal simulation and actuation constraints.

What are the core results and scope of the claim?

The core results give a necessity statement and a conditional near-sufficiency statement linking recursive self-simulation to a precise thermodynamic measure of lawful amplification of rare-valid futures. The paper states that high rare-valid lift is impossible unless the internal simulation identifies rare-valid futures with high fidelity; conversely, when rare-valid fidelity is high and the simulation contains an effective policy, the achievable lift approaches the actuation-limited optimum.

Chattopadhyay explicitly positions the framework as universal in scale, applying from passive matter and feedback controllers to large language models and humans as text generators, and even to Maxwell-demon-like information engines. The submission is cataloged under Artificial Intelligence, Statistical Mechanics, Information Theory, Mathematical Physics, and Adaptation and Self-Organizing Systems on arXiv, and is available as arXiv:2606.20231 (submitted 18 Jun 2026; file 3,062 KB). The paper's DOI link is https://doi.org/10.48550/arXiv.2606.20231.

Why it matters

If intelligence can be cast as a measurable thermodynamic quantity, the paper provides a common metric that spans physical systems, engineered controllers, and cognitive or algorithmic agents. That creates a potential basis for comparing control strategies, learning algorithms, and physical implementations on the same scale, because the framework ties performance (probability amplification of rare-valid outcomes) to measurable fidelity and actuation limits rather than to task-specific scores.

This reframes questions about model quality: high performance requires not just optimization of external metrics but internal simulation fidelity and policies that convert simulation into actuation under physical constraints.

What to watch

Look for follow-up work that formalizes empirical measures of "rare-valid lift" and methods to estimate internal-simulation fidelity in implemented systems. Concrete validation would require published procedures for measuring fidelity and actuation-limited optima in specific systems such as feedback controllers, large language models, or engineered information engines.

References and note

The paper is available on arXiv as arXiv:2606.20231 [cs.AI], submitted Thu, 18 Jun 2026 (3,062 KB). The author, Ishanu Chattopadhyay, supplies the PDF and TeX source through the arXiv entry and situates the work across multiple related subject areas listed on the submission page.