2017, the boom of quantum machine learning: Although the first quantum algorithms for machine learning have been proposed as far back as in the late 90s, this research subfield is now attracting an ever increasing interest from the quantum tech research community. As several articles on the topic are uploaded each week, there is an effort to grasp the reaches of the new discipline, with multiple reviews appearing recently such as a recent one that even provides a discussion on the meaning of quantum-enhanced artificial intelligence. Link

The central idea underpinning quantum machine learning is simple: since quantum machines promise the advantage of carrying out some computations in a completely unconventional way and potentially exponentially faster, it would be desirable to exploit them to broaden the tools of standard machine learning. The proposition comes with promising indications, such as that quantum machine learning subroutines can solve a special set of linear algebra systems with far less steps, as well as fascinating possibilities, like exploring the effect of artificially injected quantum noise in data analysis, as this might play a crucial role similarly to noise in machine learning in the classical domain.

Still, there are a number of serious theoretical challenges to tackle as we are just starting to scratch the surface, including two fundamental bottlenecks known as the *input* and the *output* problem. The “classical” data needs to be “quantized” in order to be fed to the quantum computer. Loading a quantum random access memory (qRAM) seems a somewhat expensive task, and moreover the quantum output needs to be retrieved. Once such interconversions are taken into account, the promised advantages might be nullified by the overhead. A succinct but technical overview of the current state of the art is published in *Nature*. Link

For these reasons, the most promising machine learning applications for the first generation of quantum machines might be the study and optimisation of problems that are intrinsically quantum, such as in quantum chemistry and dynamics, where one plays with quantum data from the starting point.

Much remains to be seen, due to the critical difference between quantum and classical machine learning state of affairs: while engineers are obtaining impressive results just by tinkering with neural networks, even if the machines’ rationale is hard to interpret, we are still waiting for the deployment of quantum machines harnessing the power of quantum algorithms. Meanwhile, standard machine learning techniques are applied to optimise quantum experiments and its paradigms are starting to flow into quantum information theory.

This is a collection of my articles on quantum technology, part of my Quantum Tech Newsletter. You can read the original posts also on Medium:

- Gravitational Quantum Sensors
- Quantum Advantage
- Analog Computing
- Quantum Internet
- Quantum Games
- Open-Source Quantum Tech
- Quantum Machine Learning
- Space Quantum Communication

© Nathan Shammah — 2018 and beyond.