Why most Quantum Machine Learning won't scale (and what will)

Join our quantum perspectives seminar series (online event)

Quantum machine learning (QML) is often motivated by asymptotic computational advantages, yet these analyses rarely account for the physical cost of executing quantum algorithms. The prevailing assumption is that early fault-tolerant quantum computers will make such costs negligible. In this talk, I challenge this assumption by introducing a simple wall-clock cost framework that captures the overhead associated with repeated state preparation, wavefunction collapse, and the no-cloning theorem. This perspective reveals that many popular QML architectures remain prohibitively expensive despite fault-tolerance. A shift toward hardware-aware design is crucial to reduce practical training times from millennia to milliseconds, offering a real roadmap toward empirically useful QML.

Add event to calendar