Inflation as an Information Bottleneck

<!--HTML-->We discuss how tools from information theory combined with machine learning a direct map from the microscopic parameters of a string inflation mechanism to the CMB observables allows us to systematically estimate the observable footprint of the theory uncertainty. We demonstrate this using the example mechanism of single-field axion monodromy inflation. We find that inflation acts as an information bottleneck, with a surprising number of microscopic parameters decoupling from the CMB observables.