Molecular dynamics (MD) is an important tool across chemistry, physics, and biology. MD connects microscopic physics to macroscopic thermodynamic observables yet is often practically limited by the sampling problem. Computing thermodynamic observables — free energies and rates --- requires the sampling of statistics from high-dimensional molecular probability distributions to form unbiased averages and correlations — without a sufficient sample the link is lost.

In this talk, I will discuss the advent of Generative Molecular Dynamics [1] as a strategy to efficiently generate independent statistics through the training of generative machine learning models. I will outline some of our recent work including implicit transfer operators [2], and our efforts to make this principle generalize [3,4].

[1] Olsson “Generative Molecular dynamics” Current Opinion in Structural Biology 96, 103213
[2] Schreiner et al “Implicit Transfer Operator Learning: Multiple Time-Resolution Models for Molecular Dynamics” Advances in Neural Information Processing Systems 36 (NeurIPS 2023)
[3] Diez et al. “Transferable Generative Models Bridge Femtosecond to Nanosecond Time-Step Molecular Dynamics” arXiv:2510.07589
[4] Antoniadis et al. "Protein Language Model Embeddings Improve Generalization of Implicit Transfer Operators” arXiv:2602.11216