Australia’s “disproportionately strong” body of quantum-computing research has convinced a former Google research pioneer to relocate to Sydney for a collaboration with local innovators making real progress in the push to develop usable quantum computers.
Late last year, Google engineer Professor John Martinis claimed that his team had achieved ‘quantum supremacy’ with a quantum computer using 53 qubits – a fundamental unit of quantum computing, which harnesses the curious behaviour of sub-atomic particles to churn through calculations so quickly that tasks like brute-forcing encrypted data become possible.
In just 200 seconds, that system could solve a test mathematical problem so complicated that even supercomputers like Australia’s Gadi – recently pegged as the world’s 25th-fastest machine – would take 10,000 years to complete.
Engineers at IBM, which has developed its own 65-qubit computer, claimed by IBM that Google’s results were based on data artificially engineered to produce a particular result.
The test calculation was “kind of a crafted problem,” Martinis told Information Age, “but it was a real problem that scientists and engineers could understand, and get government people to understand.”
“We were able to show that the field is on track, and that at least a quantum computer is powerful, which is interesting. The next step is to make it useful – and we have to build it bigger and better to do that.”
Centre of quantum gravity
As researchers better learn to control qubits, scaling quantum computers to useful size has a become a major research focus.
Martinis recently arrived in Sydney for a six-month stint with Silicon Quantum Computing (SQC), the venture stemming from the world-leading work pioneered by former Australian of the Year UNSW Professor Michelle Simmons.
Most quantum computers have relied on being cooled to near absolute zero, reducing molecular entropy and ‘charge noise’ that makes qubits unpredictable and affects the computer’s reliability.
SQC’s success in embedding individual atoms in silicon provides crucial predictability in the way qubits perform – and that, Martinis said, has put Australian researchers at the vanguard as they work towards a 10-qubit quantum computer by 2023.
“When I left Google, I decided that I would like to go to the various quantum computing ideas around the world,” he explained, “and working with SQC was at the top of my list.”
“What I like about the things going on in Australia,” he continued, “is that they build very clean systems with low noise – which presented a unique opportunity that meshed with how I like doing physics on clean systems.”
“If you have a noisy device, then the quantum bit doesn’t act like you think it should act, and it’s not going to work properly.”
The behaviour of two or three qubits can be mathematically modelled on a conventional computer – an approach that researchers are already using to write software for quantum environments like IBM’s cloud-based Quantum Experience and Honeywell’s own platform.
Once the quantum computer scales to around 10 qubits, however, there is no way to simulate its performance or check its results – meaning researchers must be able to trust that built-in error correction has adequately eliminated quantum noise.
Predictably error-free performance, Simmons told Information Age, is the key to being able to build consistently performing quantum processors with “very fast qubits, very stable qubits, and reproducible qubits” that consistently behave as required.
Yet in the quantum world, interacting with one qubit can often affect the neighbouring qubit – something that Martinis was able to rein in during his quantum-supremacy work, and which Simmons looks forward to applying to her team’s own work.
“There are all kinds of technical challenges to solve by making things very small,” she said, “and John joining us is applying some of the technical issues that he had to solve to get there, but he can see in a different way.”
Quantum meets the real world
Industrial partners are helping identify use cases for complex data-analytics problems in areas such as financial modelling and energy exploration – which recently saw Woodside Energy sign a long-term AI and quantum partnership with IBM.
That company’s Australian researchers, affiliated with the University of Melbourne, are supporting IBM’s plan to build a 1121-qubit Quantum Condor processor within three years.
Just as there are many ways to perform multiplication and long division, quantum systems require mathematical problems to be represented in completely new ways, says Anna Phan, a Melbourne-based IBM Research Labs research scientist whose work in ‘quantum machine learning’ is exploring how to do this.
“We have shown that with specific types of quantum correlations, you can’t efficiently get the same accuracy using classical machines,” she told Information Age. “But currently that requires an artificial data set; we want to see if we can find a real-world data set where the quantum correlations can’t be explained on a classical computer.”
Physics data sets derived from chemistry, molecular modelling, protein folding and similarly complex applications – long pushing the boundaries of conventional computers – seem particularly well-suited to modelling in the quantum space and “we believe these will possibly be the first use cases and applications of quantum computing,” Phan said, “particularly with the larger machines that are on our roadmap.”
Those machines will give researchers more options as they navigate the oddities of the quantum world, pushing towards building reliable quantum integrated circuits operating well past the capabilities of conventional systems.
“Quantum in Australia is disproportionately strong and it really is a powerhouse across the country,” Simmons said.
“The timeframe [for large-scale systems] will depend on how quickly we can go, and having John here is going to accelerate our ability to do that quickly. But it’s also great to get co-development partners to look at what their use cases are now – and what we actually build.”