IBM’s Eagle quantum computer completed a calculation that stumped the conventional supercomputer it was pitted against.
Quantum computers have sparked excitement for a few decades, but researchers haven’t built one that universally outperforms all conventional computers. This is partly because all existing quantum computers are “noisy” – their results are corrupted by errors, in the same way that sound can get lost in a crackly recording.
Now, Abhinav Kandala at IBM and his colleagues have shown that even a noisy quantum computer can be more accurate in its calculations than a conventional machine.
They compared the performance of IBM’s Eagle quantum computer, which contains 127 quantum bits, or qubits, with that of a supercomputer at the Lawrence Berkeley National Laboratory in California. Both were tasked with calculating the most likely behaviour of a collection of particles, such as atoms with spin, arranged in a grid and interacting with each other.
The difficulty of this problem increases with the number of particles. For up to a certain number of particles, researchers can solve the relevant equations on a computer exactly. Supercomputers and approximation methods can handle the calculations when larger number of particles are involved, but eventually computations become so complex that conventional computing fails. The IBM researchers wanted to test Eagle together with a set of methods they developed for mitigating the effects of noise in each of these cases.
When the supercomputer could complete the calculations, the results of the two methods were in agreement. But when the complexity was increased beyond a certain point, the supercomputer failed, while Eagle was still able to deliver a solution. Even though there was no way to test if Eagle’s result was correct, Kandala says good agreement with established calculations up to that point gave his team confidence that the quantum computer had passed the test.
The IBM team does not claim to have achieved quantum supremacy, which would mean proving that Eagle’s performance is impossible to match for any conventional computer and algorithm. They didn’t test the quantum computer against every existing conventional approach. Kandala says this was not their goal as he expects classical computing methods to continue to improve, but he and his colleagues wanted to test the usefulness of the quantum processors available now.
He says that the methods they used for these calculations could be adapted to a broader set of problems such as calculating the behaviour of materials or molecules that are of interest in biophysics and chemistry.
Xavier Waintal at CEA Grenoble in France says demonstrating that a quantum computer is truly useful beyond an experiment designed specifically to show computational power is a tall order. He says that Eagle’s hardware performance is impressive and is an important milestone, but he is sceptical of how much utility researchers can get from the demonstrated ability.
“Demonstrating that a device is doing something that absolutely cannot be simulated on a classical computer is notoriously difficult. You would have to prove it for every existing technique as well as every conceivable technique that doesn’t even exist yet,” says Mark Howard at the University of Galway in Ireland. But the competition between quantum and classical computers has historically improved how researchers use both, he says.
IBM’s team is already looking to repeat its experiment for more complicated calculations and with larger, and therefore potentially more powerful, quantum computers. The hope is that with further development, quantum computers could assist in creating new materials for batteries or fertilisers, or new compounds for medicines.
“We know classical computing is not going anywhere, but we are excited about getting utility out of our quantum computers. We now have a good tool with over 100 qubits, and the more we use it the better we will get at making it even more useful,” says IBM’s Katie Pizzolato.