Active AI for Scalable Control of Quantum Systems

Abstract:

Simultaneous progress in quantum computing and artificial intelligence offers unprecedented opportunities for bidirectional breakthroughs. Arguably, the complexity of these systems has grown beyond manual control, necessitating a paradigm shift in how we approach them. Taking inspiration from human cognition and the scientific method, we propose a new research program centered around the concept of Active AI, where agents continuously hypothesize, control, and learn from live interactions with physical systems.

In this seminar, I will explain how quantum computing provides the ideal testbed for this vision, as evidenced by early results on:

Real-time Bayesian control of quantum hardware [1-3],
Learned efficient representations of quantum systems and hardware [4],
A quantum machine learning advantage for processing quantum data directly on-chip [5],
RL algorithms, and human-in-the-loop protocols [6] for quantum error correction,
Autonomous tuning and control based on developed digital twins [7-9].

As a bonus, I will discuss how this research program can also benefit classical AI, including a statistical physics-based understanding of deep learning architectures~[10] and the development of scalable active inference algorithms.

[1] F. Berritta, T. Rasmussen, J. A. Krzywda et al. Nature Communications 15, 1676 (2024).
[2] F. Berritta, J. Benestad, L. Pahl, M. Mathews, J. A. Krzywda et al. PRX Quantum 6, 030335 (2025).
[3] F. Berritta, J. Benestad, J. A. Krzywda et al. Phys. Rev. X 16, 011025 (2026).
[4] S. Samadi ... J. A. Krzywda. arXiv:2510.13578 (2025). (Accepted in Phys Rev. App).
[5] O. Danaci ... J. A. Krzywda, arXiv:2605.21346 (2026).
[6] https://erratiq.xyz/,
[7] R. Koch, J. A. Krzywda, https://github.com/OpenSpin/ReadSpyn.
[8] J. A. Krzywda et al., SciPost Physics Codebases 043 (2025).
[9] B. Kreft, J. A. Krzywda, https://github.com/KREFT-QS-13/learning-QDs-parameters.
[10] C. Chalkias, E. Van Nieuwenburg, J. A. Krzywda (To appear on arXiv in June/July 2026).

July 2nd at 11:00 ONLINE.
Zoom link: https://us06web.zoom.us/j/84248911743?pwd=ZsiAOQbRgYCm5IFsArOnb18Fj5IsZh.1
Meeting ID: 842 4891 1743
Passcode: 394021