Australian researchers to build apps for Google quantum computer

Initially, Google would be focusing on so-called “noisy intermediate-scale quantum” (NISQ) algorithms, he said.

NISQ is the low-hanging fruit of quantum computing that scientists hope will make the technology useful this decade.

At some point in the future, machine learning will be done with quantum processors.

— Hartmut Neven

NISQ algorithms typically require fewer than 100 qubits to operate, and can perform their computations in microseconds, before a nascent quantum computer becomes too unstable and error-prone to produce reliable results.

They typically focus on physical-world applications such as modelling molecules or the exotic states of matter, Dr Neven told The Australian Financial Review.

Google’s long-term goal, however, is to develop bigger, more sophisticated quantum algorithms that can be used in applications such as machine learning and artificial intelligence, making quantum computing useful to the company’s core business of selling ads, he said.

Google’s collaboration with the universities will be spearheaded out of the Australian Research Hub the advertising giant has just opened in Sydney, on the former site of this newspaper. It has contracted Marika Kieferova, a lecturer in quantum algorithms at UTS, to run the collaboration.

Australia’s Chief Scientist, Cathy Foley, said in a statement the collaboration was “a step in building Australia’s quantum industry here”.

‘Punching above its weight’

“Having Google’s investment in Australian quantum science is a testament to the world-class research that has been supported by the Australian Research Council for over two decades,” she said.

“Australia is punching above its weight in terms of investments and successes in quantum computing,” Dr Neven said, noting that some of the experts at QuAIL had been trained in Australia.

“There’s a broad range of [quantum] activities going on in Australia, and it was just natural for us, given the scarcity of global talent in this space, to tap into Australian collaborations,” he said.

One class of quantum computing algorithm the Australian researchers probably would not be focusing on was optimisation software, Dr Neven said.

Solving optimisation problems, such as figuring out what is the best route for a delivery van to take, was expected to be one of the major industrial applications for quantum computers, alongside drug discovery.

But in a huge blow to the industry, scientists recently figured out that the speed running optimisation software on a quantum computer would actually be slower than running it on a supercomputer, meaning there is no “quantum gain” at least until quantum computers have parts inside them that run much, much faster than they do now.

Logic switches that flip in nanoseconds inside a regular computer chip take 10,000 times longer to flip in a quantum computer, and given that quantum optimisation algorithms are not 10,000 times faster than classical optimisation algorithms, they’re not going to as useful as was once thought, he said.

Indeed, Dr Neven got Google into the quantum computing race in the first place on the premise that its core machine learning technology could be dramatically improved by quantum optimisation, so the news that it’s slower was a big disappointment, he said.

“You lose some, you win some. It’s still true that at some point in the future, machine learning will be done with quantum processors,” he said.