Description
The world is witnessing unprecedented automation - from self-driving vehicles and sophisticated large language models to automated trading algorithms and virtual agents. As we advance such agents to operate in ever more complex environments, successful agents must make ever more context decisions. Associated memory and energy costs are escalating at potentially unsustainable rates.
In this talk, I ask: Could quantum-enhanced agents - agents equipped with the ability to process information quantum mechanically - offer a fundamental energetic advantage?
In this talk, I show how we can tackle this question using fundamental tools from information thermodynamics and quantum information, leading to the phenomenon of causal waste. When classical machines execute a target strategy, they are often forced to record information from the past that appears useless in hindsight. This 'waste' is fundamentally unavoidable, resulting in memory and energetic overheads. I then offer conditions in which quantum agents can reduce this waste - and thus achieve certain tasks with energetic dissipation below ultimate classical limits.
Note: Talk will be primarily based on work on published work in
Energetic Advantages for Quantum Agents in Online Execution of Complex Strategies, Phys. Rev. Lett. 135, 160402 (2025)
DOI: https://doi.org/10.1103/dryx-l5rt
| I am the presenting author | Yes |
|---|