At the intersection of two challenging computational and technological problems, may lie the key better understanding and manipulating quantum randomness
Machine learning, that AI subfield that allows Alexa and Siri to parse what you say and self-driving cars to drive down a city street, could benefit from quantum computer-derived speedups, say researchers. And if a technology incubator program in Toronto, Canada, has its way, there may even be quantum machine learning startup companies launching in a few years.
Quantum machine learning, in brief, represents the intersection of two challenging computational and technological problems: quantum computing and machine learning. The first involves essentially hacking the laws of physics to trick quantum systems like interacting atoms in laser ion traps or tiny elements of superconducting circuits to run multiple threads of quantum algorithms at once. The reason for going to the trouble is simple: Quantum computers crunch numbers faster than any conventional computer could do. The second technical hurdle involves running learning algorithms on vast fields of data—for instance, the legions of images used to teach a self-driving car the difference between a bicycle and a motorcycle as seen from all angles and lighting conditions.
What they have in common, says Peter Wittek, research fellow at the Institute of Photonic Sciences in Castelldefels, Spain, is that both quantum technologies and machine learning systems are essentially probing different kinds of high-dimensional variables and datasets through dense clouds of noise and uncertainty. And the overlaps between these two fields, he says, could be substantial. Even partially realized quantum computers might still find machine learning applications.
“To build universal quantum computers… is a big engineering challenge,” says Wittek, who’s also academic director at the Creative Destruction Lab startup incubator affiliated with the University of Toronto’s Rotman School of Management. “But it turns out, for quantum machine learning, you need something less. So there are learning models for which you can use either small-scale quantum computers or quantum technologies that are not universal quantum computers. So quantum machine learning could be the next big application for quantum technologies after quantum cryptography and quantum random number generation.”
Wittek, author of the 2014 book Quantum Machine Learning: What Quantum Computing Means to Data Mining, says the field of quantum machine learning took off after the 2008 creation of a quantum algorithm called HHL (after its three creators Aram Harrow, Avinathan Hassidim, and Seth Lloyd). HHL solves linear algebra problems with applications in some of those high-dimensional-variable problems from machine learning.
To be clear, Wittek says, conventional machine learning is very good and only getting better in many applications across industries like finance, transportation, and medicine. Most machine learning applications, he says, will not be knocked off their perch by a quantum machine learning system, even if Google, IBM or another research lab can one day build a fully functioning and practicable quantum computer. Yet conventional machine learning has weaknesses.
For instance, standard machine learning algorithms, Wittek says, have a hard time generating purely random numbers. Monte Carlo machine learning algorithms, often used in financial applications, require purely random numbers for optimal results. But pseudo-random numbers are often the best a classical computer can generate. Quantum systems, on the other hand, are purely random by definition. So a quantum machine learning has a foothold there.
And, says Nathan Wiebe, a researcher at Microsoft’s Quantum Architectures and Computation Group in Redmond, Wash., quantum machine learning systems will work especially well when the input is not the 0s and 1s of classical data but rather the qubits of the quantum computer.
“If you think about a quantum computer, how do you understand what’s going on inside one?” Wiebe says. “The vectors that describe it exist in an incomprehensibly large space. There’s no way you can go through, read off every single entry of those vectors and figure out if the machine is working properly.”
Developments in quantum machine learning, he says, have already led to protocols that could help smaller quantum devices interface with larger quantum computers. In other words, quantum machine learning might help us peer into and better understand otherwise inscrutable many-qubit quantum computers.
“As we start readying up and enjoying this rapid expansion of [quantum computer] technology we’ve seen in the last little while, more and more doors are going to open up and potentially lead one day to running things like the HHL algorithm,” Wiebe says.
According to Scott Aaronson, professor of computer science at the University of Texas at Austin, HHL has led to more hype than actual near-term hope in the field. As Aaronson argued in a skeptical 2015 review of quantum machine learning research, the words caveat emptor should be tagged to any citation of HHL today. He cast a wary eye on all the approximations and likely shortcuts one would need to take to employ HHL in a working quantum computer in the near future.
“Almost all the quantum machine learning algorithms that have been published over the last decade are really frameworks for algorithms,” Aaronson says. “They’re algorithms that don’t start with the classical problem that you would like to be solved and the answer to that problem.”
But that skepticism hasn’t dulled interest. Wittek says that, despite the technical objections, the number of applicants to this year’s quantum machine learning bootcamp and startup accelerator in Toronto exceeded expectations. Something has inspired entrepreneurs to try their hand at turning bold ideas into, possibly, working technology.
“Incorporation must be done by November, so these will be real companies,” Wittek says. “And the hope is by next summer we’ll have companies raising money.”
As Einstein once observed about quantum physics, God may not play dice. But angel investors, these companies can only hope, do.
Source: IEEE Spectrum