Quantum machines promise big leap forward in computing
In everyday life, it is hard to imagine that something can be in two places at the same time. Quantum mechanics, however, describes this unique behavior at the submicroscopic level. Particles such as photons or electrons, can also exhibit both particle and wave behavior, depending on the mechanism of observation.
Using ultra-low temperatures and superconductivity, quantum computers take advantage of these fuzzy situations, particularly the phenomena of superposition, entanglement and tunneling. While conventional machines monitor an on-or-off state across millions of transistors, a quantum computer can accommodate both the on and off state, simultaneously. In other words, until they’re measured, quantum bits, known as qubits, can take on a zero or one state, at the same time.
Satya Nadella, CEO at Microsoft, analogously stated that a classical computer trying to navigate a maze would try each path one after another, while a quantum computer computes all paths simultaneously. Quantum computing has attracted its share of hype and skepticism in recent years, but is now becoming more tractable in the corporate IT sector. Aside from Microsoft, giants such as IBM, Intel and Google have investigated quantum computers because they harness the potential to solve complex problems with remarkable speed improvements.
Breaking new computing ground
Dr. Colin P. Williams wrote the first textbook in the field of quantum computing. Now Vice President of Strategy & Corporate Development at D-Wave Systems and based in British Columbia, Canada, Dr. Williams attended last month’s Rakuten Technology Conference and gave a presentation on the latest developments in the field.
“As a company, we have a grand ambition of solving the world’s hardest problems,” Dr. Williams stated to attendees. “The areas we want to focus on are machine learning and artificial intelligence, because they’re most naturally suited to our technology.”
D-Wave is known for being the world’s only manufacturer of commercially available quantum computers. The shed-sized, multimillion-dollar machines are best suited to complex optimization problems. However, they are not yet able to solve any computational problem and hence are not categorized as “universal” quantum computers. Nevertheless, that hasn’t road blocked NASA, Lockheed Martin or others from purchasing them. Earlier this year D-Wave announced its latest system, the 2000Q, which doubles the amount of qubits available for solving potential optimization problems that exist in domains such as machine learning and cybersecurity. Dr. Williams explained how a dilution refrigerator, a cryogenic device, maintains the D-Wave processor environment at near absolute zero, 180 times colder than interstellar space, in order to achieve quantum effects.
“You can create an object that’s actually small enough to fit in the palm of your hand, and it has more components than all the particles in the entire universe,” Dr. Williams noted.
From quantum medicine to e-commerce
Although general-purpose quantum computers are yet to be built, they could transcend a wide range of applications. E-commerce platforms, such as Rakuten Ichiba, could potentially increase the computational speed of natural language understanding or the learning rate of machine intelligence techniques, to provide users with a more human-like experience.
Since quantum computers are not yet practically capable of replacing the classical “von Neumann” architecture in all situations, it may be best to use the two methods in combination to provide more optimal solutions. For instance, a radiation treatment plan for a cancer patient was optimized with a hybrid system that produced a better solution than one based on classical computing alone. The result for the patient’s treatment was less collateral damage to the surrounding tissue.
In a unique business case, D-Wave recently released web-based machine learning services to the University of Toronto’s Creative Destruction Lab, where startups are using them to accelerate the training of probabilistic machine learning models. The models can learn from noisy or incomplete data and then later infer any missing data; such as extrapolating an entire face when given a picture of a partial face.
“With a quantum chip, it seems like you could get better samples in fewer tries,” Dr. Williams explained in an interview on the sidelines of the conference. “So you can train quicker and do a better-quality model as well. And that will revolutionize AI, because at the end of the day it’s the quality of the model that matters more than how long it took to train.”
Read more posts from the Rakuten Technology Conference here.
