Google's new quantum processor could soon outperform classic supercomputers
With the quantum computing race heating up between Google, IBM and Intel, it feels like we're hurtling towards quantum supremacy, that milestone when a quantum computer outperforms a classical one for the first time. Bringing us ever closer, Google has now unveiled Bristlecone, a new quantum computer chip with the record-setting power of 72 quantum bits (qubits).
Traditional computers perform their calculations in binary, so every bit of data is represented as either a zero or a one. Thanks to the quirky science that is quantum mechanics, a qubit can be in a superposition of both, effectively representing both a zero and a one at the same time. That means the power of a quantum computing system scales exponentially – two qubits can represent four states at once (00, 01, 10 and 11), three qubits represent eight, and so on.
As a result, quantum computers are great at performing simultaneous operations, processing all of these states at the same time where classic computers would have to run through each in turn. That means that, theoretically, a quantum computer made with a 49-qubit chip (like Tangle Lake, a processor Intel unveiled at CES in January) could outperform our current best supercomputers at certain types of operations.
Google's latest chip prototype, Bristlecone, boasts a staggering 72 qubits. These are arranged in a square array, and get their quantum nature through superconductivity, which allows them to represent multiple states by conducting current in two directions at once.
But it's not just a matter of cramming in more qubits and putting them to work tackling the biggest mathematical problems in the universe. Qubits are notoriously fragile, and outside fluctuations can introduce memory errors that undermine the whole calculation. And these are hard to detect – as Schrödinger illustrated with his infamous feline friend, as soon as you peek in the box you collapse the wave function into one state or the other, garbling the information. That could be great for quantum encryption systems that alert you if someone tries to hack in, but it's not so helpful when you just want to read the output of a quantum computer.
To get around that problem, a few years ago the Google Quantum AI Lab developed a quantum error correction (QEC) technique, and demonstrated it in a system with nine qubits. The QEC works by checking combinations of data and measurement qubits, allowing the system to indirectly measure the information without affecting it. On the 9-qubit system, the method achieved error rates as low as 1 percent for readout, 0.1 percent for single-qubit gates and 0.6 percent for two-qubit gates.
The Google team is now aiming to achieve similarly low error rates on the Bristlecone processor's 72 qubits. To measure its performance, the researchers have developed a benchmarking tool that deliberately introduces a single error into the system, then compares the sampled output distribution to the results that a simulated classical computer came to.
The team says that eventually, the Bristlecone should be more than capable of reaching that watershed moment of demonstrating quantum supremacy. If it does manage to outperform a classical supercomputer, the processor could become the basis for building larger scale quantum computers.
Source: Google Research