What is it about?
With the rise of both deep learning (AI) and quantum computing, there has been a lot of excitement about simply combining them for incredible results. Our 2019 paper took a critical look at this assumption, arguing that the fundamental philosophies of the two fields are actually at odds. Deep learning works by using massive numbers of simple calculations, while quantum computing gets its power from a small number of extremely complex qubits. We outlined several major roadblocks to naively "porting" AI to a quantum computer. For example, the way AI models "learn" doesn't map well to quantum search algorithms, the sheer size of modern neural networks is too large for today's scarce qubits, and essential AI techniques like "transfer learning" are fundamentally incompatible with the laws of quantum mechanics (specifically, the "no-cloning theorem"). Our conclusion was that you can't just put a classical AI architecture on quantum hardware; the real path forward requires inventing entirely new quantum algorithms designed from the ground up for learning.
Featured Image
Photo by BoliviaInteligente on Unsplash
Why is it important?
This work provided a much-needed dose of realism during a period of intense hype. By clearly articulating the deep, physical barriers to combining AI and quantum computing, it helped ground the conversation and steer it in a more productive direction. Instead of trying to force classical methods onto quantum computers, the paper argued that the true bottleneck was the lack of new, quantum-native learning algorithms. It also proposed a clear set of criteria to help researchers evaluate whether a machine learning problem is even a good candidate for a quantum computer in the first place, helping to focus efforts on problems where quantum mechanics can offer a genuine advantage.
Perspectives
This paper came from my tendency to always see connections between different fields. By 2019, deep learning and quantum computing were two of the most exciting topics in science, and I wanted to seriously explore how they could—or couldn't—fit together. This was a direct follow-up to my first quantum paper; while that one was about building a quantum computer, this one was about what we could realistically do with one. My analysis led me to be a bit of a skeptic. I realized the hype was outpacing the physics, and that the core principles of deep learning were simply incompatible with the rules of quantum information. For me, this work solidified a key idea: you can't treat a quantum computer as just a faster version of a classical one. It requires entirely new ways of thinking. This paper was my attempt to bridge these two incredible fields, not by saying it was easy, but by clearly defining the deep and interesting challenges that needed to be solved first.
Rohit Goswami
University of Iceland
Read the Original
This page is a summary of: Qubit Network Barriers to Deep Learning, December 2019, Institute of Electrical & Electronics Engineers (IEEE),
DOI: 10.1109/wrap47485.2019.9013687.
You can read the full text:
Contributors
The following have contributed to this page
