3

The provocatively titled Why AI could eat quantum computing’s lunch provides an overview about what kinds of quantum systems that AI can tackle. The most interesting part of this article is that, apparently, AI can come up with "good enough" solutions to strongly correlated systems. This is surprising because strongly correlated systems are the types of systems that we lean on quantum computers to help with. There is a silver lining. The article says:

One area where quantum computers look likely to have a clear advantage is in simulating how complex quantum systems evolve over time, says EPFL’s Carleo. This could provide invaluable insights for scientists in fields like statistical mechanics and high-energy physics, but it seems unlikely to lead to practical uses in the near term. “These are more niche applications that, in my opinion, do not justify the massive investments and the massive hype,” Carleo adds.

My question is, what are these niche applications that quantum computers have a clear advantage in simulating?

Victory Omole
  • 2,332
  • 1
  • 10
  • 24

2 Answers2

2

I would refer to @YuTong's response thread on Twitter here.

Tong makes the following points, some of which I (naively) would agree with:

  • Neural networks and quantum computing have been married together for a long time already
  • Even classical DMRG methods may probably outperform AI
  • It's probably premature to tell what quantum computing will teach us about ground states unless and until we have them.

Quoting Tong:

Compared to that quantum computing gives us the freedom to measure and probe the system in all kinds of ways.

The mere BQP-completeness of Hamiltonian simulation and, say, the Guided Local Hamiltonian problem, along with the reasonable assumption that (classical) AI$\subsetneq$BQP suggests that a quantum computer can find a better approximation to a ground-state energy than a classical AI system would.

I am not aware that classical AI has cracked RSA - I would be very surprised and suspicious of any claim. But we know a quantum computer cracks RSA. On the same vein, I would be very surprised if classical AI outperforms Guided Local Hamiltonian or any other problem thought to be in BQP\BPP.

Whether the accuracy you get from a quantum computer justifies the expense is an interesting business question that I'm not sure this forum can answer?

Mark Spinelli
  • 15,378
  • 3
  • 26
  • 83
1

Read the Fine Print's rephrased sentence applies here. The statement that quantum computers are better at quantum simulation problems than AI does not have as strong a backing that factoring with quantum does. We should expect a back and forth between classical and quantum for attacking quantum problems indefinitely like we do for optimization problems.

Victory Omole
  • 2,332
  • 1
  • 10
  • 24