When we talk to researchers in the quantum community, we see a similar split on the question of whether NISQ-level machines will be usable for commercial or scientific gain or whether we need to wait for error-corrected machines to appear before we see this. IBM is firmly in the latter camp and has consistently indicated that they hope to see customers use their quantum machines for quantum gain in the near future before full quantum error correction becomes available.
There is a clear business advantage for IBM if this can be achieved. This will expand their market from only users accessing their machines for training and demo purposes to adding additional users who will use the machines for regular production. IBM has definitely made progress in increasing the quantity and quality of hardware qubits. But to accelerate the achievement of quantum superiority as early as possible, IBM has also been researching how innovative software algorithms and hybrid classical/quantum processing can help achieve quantum superiority with near-term NISQ processors. Some of the innovations they have introduced in recent years include things like dynamic circuits, circuit knitting, serverless executionand error mitigation techniques (which are different from error correction) such as Probabilistic Error Cancellation (PEC) and Zero Noise Extrapolation (ZNE). In general, we believe that IBM has pursued this technique more aggressively than most in the industry.
In a recent demonstration that IBM did with the University of California, Berkeley, they worked on the time evolution problem of a 2D cross-plane Ising model. Like Google’s 2019 Beyond Classical experiment and updated this year, this demonstration pits the quantum versus classical approaches. The IBM quantum machine used is their Eagle 127 qubit device with a circuit spanning up to 60 layers of up to two-qubit gates and a total of 2,880 CNOT gates while the classic machine they are benchmarked on is Supercomputer heart at the Lawrence Berkeley Lab National Energy Research Scientific Computing Center (NERSC) and Basic supercomputer at Purdue University.
An important point in this simulation is that a quantum circuit 60 layers deep will typically make so many errors with the gate current fidelity level, that the end result is usually garbage. But in this simulation, the main technology used by IBM is Zero Noise Extrapolation (ZNE) as described in this video to reduce errors so that the end result is still valid.
The main challenge when comparing the performance of quantum computers is how do we know when the output is valid. In this regard, IBM and UC Berkeley took a similar approach to that of Google. They started running simulations at lower levels of complexity and seeing how the quantum results matched up with the classical results. At lower levels of entanglement, classical machines can simulate the results exactly, and when the classical and quantum results are compared, the results are exactly the same. This continues as the level of entanglement continues to increase and provides confidence that the quantum machine is doing the right thing. At higher levels of engagement, the classical machine no longer has enough performance to simulate the output precisely and must start using approximations. But the quantum machine keeps going.
IBM and UC Berkeley have published papers on Nature Magazine along with several press releases, blogs, and videos explaining this research in more detail. Here’s the list:
- An IBM press release announcing this “Beyond Classical” demonstration is available Here.
- IBM blog post with additional information available Here.
- A UC Berkeley press release describes the research Here.
- Technical paper published in Nature Magazine title Proof of the usefulness of quantum computing before fault tolerance can be seen H
- There are videos from IBM and UC Berkeley explaining their approach and some of the challenges of conducting the experiment Here.
- Another video from IBM explaining how mitigating errors can make quantum computers useful can be viewed Here.
- And finally, a video from IBM providing additional technical background on the experiment is available Here.
IBM believes that the techniques demonstrated in this experiment will soon be applicable to real-world problems and create what they call the Quantum Utility. This development may have a significant impact on the quantum computing market. Global Quantum Intelligence (GQI) has written an article analyzing the potential impact of this and a description of the next 100×100 challenges IBM hopes to tackle next year. The article was posted on the QCR website here and also included as part of our technical collection (along with many other things) on the GQI website.
June 14, 2023