Quantum computers are fast, but they produce a lot of wrong answers. IBM says its error mitigation technique could fix that.
Today’s most powerful supercomputers can simulate complex weather patterns and the birth of stars, but even a modest quantum computer could outperform those machines. The mysterious nature of quantum mechanics has thus far made quantum computing little more than a curiosity, but IBM is claiming a significant breakthrough. According to IBM’s Jay Gambetta, we’ve reached “the era of utility.”
Quantum computers are a hot topic for research, but they haven’t been useful for doing calculations yet. We keep trying because of the incredible potential of quantum computing, which takes advantage of weird quantum properties like entanglement, interference and superposition to accelerate calculations. For example, the digital computers we’ve used for decades use transistors to signify a 1 or a 0. Using superposition, a quantum bit (qubit) can be a 1, 0, or both. Holding multiple values allows a qubit to perform multiple calculations, whereas a classical computer must do them individually.
These systems are incredibly fast, true, but they don’t produce consistent results—at least until now, claims the study published in Nature. Google claimed quantum supremacy in 2019, producing calculations it says would have taken thousands of years on a digital computer. However, a follow-up analysis has shown that a conventional computer can do the same tabulations if given a little more time. IBM’s claim is not about speed as much as it’s about reliability. For a quantum computer to be useful, it needs to give the same answer every time, and IBM took a big step in that direction with a demonstration of error mitigation.
The team created a simulation of 127-atom bar magnets with a 127-qubit computer, a system known as an Ising model that is often used to study magnetism. At that scale, the magnets are affected by quantum factors, making it impossible to simulate on a classical computer accurately. The researchers leveraged quantum interference to nudge the results, pushing them farther from the solution. By introducing noise into the calculations, the IBM Quantum team could understand the effects of noise in the simulation, working backward to reach the ideal, low-noise result.
They ran the quantum Ising simulation 600,000 times to find the right configuration but verifying that answer was a different problem. Since Ising models are so complex, digital simulations only estimate some parameters. Comparing the quantum results to classical estimates, some configurations showed higher accuracy in the quantum system. In others, the values diverge, but there’s no established solution. In those cases, we don’t know if the quantum solutions are right, but the team suspects they are.
Modeling magnets is just one way to test quantum error mitigation, but the idea is to make quantum computers usable for general problems. We may still be years away from that, but IBM is continuing the quest.
Article brought to our attention by the one and only Mike Worley