New Thermodynamic Framework Proposes Universal Measure of Intelligence

The concept of intelligence has long been debated, but new research proposes a precise, universal definition: intelligence is the ability of a system to lawfully amplify rare but valid future outcomes. This means increasing the probability of desirable futures that would otherwise be unlikely under passive dynamics, while still adhering to the constraints of the environment. The framework posits that an intelligent system must model both the world and its own place within it. This leads to the idea of recursive self-simulation, where the system represents future trajectories that include its own actions. The central findings connect this architecture to a thermodynamic measure of 'lawful amplification of rare-valid futures.' Crucially, the research demonstrates that high 'rare-valid lift' (the amplification of these desirable futures) is impossible without the internal simulation accurately identifying them. Conversely, if the simulation is highly accurate and contains an effective policy, the achievable lift approaches the optimal level. This suggests that recursive self-simulation is not just a plausible feature of intelligence, but a necessary and nearly sufficient condition for achieving high thermodynamic intelligence, offering a measurable scale applicable across various systems from simple feedback controllers to advanced AI and humans.