October 29, 2019
Get Ready for Some Magic
by Paul Carroll
Clarke’s Third Law, posited by science fiction legend Arthur C. Clarke, says that “any sufficiently advanced technology is indistinguishable from magic.” Well, we’re going to see some magic, courtesy of the technology that led Google to claim “quantum supremacy” last week, and we should start adapting sooner rather than later.
The underlying technology, quantum computing, is as far from conventional computing as quantum physics is from the Newtonian view of the world and will have important implications for insurance.
Traditional computing is already pretty magical. Right, Siri? But traditional computing depends on a highly prescribed approach: Billions of transistors are in either an on or an off position, and problems are solved through an unbelievably fast manipulation of those 0 or 1 values, mostly in sequence. Quantum mechanics, meanwhile, operates in ways so mysterious that even Einstein was famously wary of the implications, and quantum computing is no different. Values don’t have to be binary: they can be both 0 and 1 at the same time. (Told you this was weird stuff.) And all the values work together at the same time, not in sequence.
The Google claim of “quantum supremacy” means it believes it has solved a problem with a quantum computer that could not have been solved with a conventional computer. To be precise, Google says it solved a problem in three minutes and 20 seconds that would have taken the most powerful IBM supercomputer 10,000 years. IBM cried foul, saying its computer could have solved the problem in 2.5 days if the problem was set up right, but, even in the best-case scenario, IBM was 700 times slower.
Quantum computing will take an estimated 10 to 15 years to establish itself, which allows time for us to adapt—but not loads of time, in some areas. Quantum computing will render trivial today’s approaches to encryption, which count on making problems (related to prime numbers) too hard to solve, and it takes about a decade to broadly replace one encryption scheme with a new one throughout industry. Quantum computing may require ending today’s reliance on passwords and other computationally intensive schemes, in favor of sampling of DNA, fingerprints, retinal scans or other biometric evidence, and the switch can’t start too early.
Richard Feynman famously said decades ago that chemistry isn’t Newtonian, it’s quantum, so any tool that’s really going to help us understand chemistry needs to be based on quantum mechanics. Et voila. Such a tool is now in sight, and being able to simulate the quantum behaviors of atoms could lead to all sorts of new materials, new medicines and new understanding of the basic behaviors of our bodies—for instance, while we talk about DNA sequences and can define them, how the strands of protein fold up is also hugely important and has been hard for conventional computers to predict.
Quantum computing could also lead to much better models for the development of hurricanes and, more generally, for potential natural disasters. While such disasters occur at a massive scale, not at the subatomic, quantum level, the intricacies of the massive number of interactions lend themselves to a quantum computing approach.
Lots of deep analytics in insurance, such as looking for fraud or identifying patterns that can help mitigate risk, also lend themselves to a quantum approach.
And fundamentally new technologies like quantum computing often produce convergences with other technologies that can rewrite the business landscape. Think, for instance, about quantum computing powering the AI that goes into driverless cars. You don’t think Google will hook up its “quantum supremacy” computer with the brain that powers all its autonomous vehicles?
Now, any technology that is expected to arrive some 10 years in the future can turn out to be mere science fiction—I’m still waiting for my flying car. And quantum computing has at least one clear drawback: It doesn’t provide a definitely right answer like conventional computers do. The quantum world is probabilistic, so quantum computers just tell you an answer is probably right. If you test the problem enough times—Google tested 1 million times in those 200 seconds in its “quantum supremacy” experiment—you can be highly confident, but you still likely want to check your work with a conventional computer.
So, don’t throw away your supercomputer just yet. But do start understanding quantum computing, and even experimenting. It’s coming, and it will make today’s AI seem like child’s play. Quantum computing will pull new companies and even new industries out of the proverbial hat.