In this video:
- Dr. Talia Gershon, Research Staff Member, IBM Research
IBM Researcher Dr. Talia Gershon wants you to know:
“Classical computers are really bad at optimization and chemistry when the problems have so many variables they are just starting to get interesting.”
By optimization, she means “finding the best possible solution from the many possible ones.” Seating guests at a function is a good example of this challenge.
In chemistry, classical computers meet the end of their usefulness when trying to simulate beyond an eight-atom-cluster molecule. In Talia’s example, she shows the size of one of these clusters, iron sulfide, then shows the tiny part it plays in the nitrogenase enzyme, responsible for fixing nitrogen (reducing nitrogen to ammonia).
These two problems have one thing in common that they share with many other big data puzzles – exponential scaling. For a deeper explanation of how exponential scaling can affect a solution, read about the wheat and chessboard problem. Talia thinks quantum computing can help us solve problems that include exponential scaling.
The two effects that exist in quantum mechanics that could make a Q-computer able to tackle a problem with as many variables as rice grains in a stack as high as Mount Everest are:
- Superposition: Classical information is a series of zeros and ones, but in quantum systems, information can exist in a superposition of multiple states – the classical 2-4-6-8 becomes 2-4-8-16 in the quantum realm
- Entanglement: When entangled, two particles will provide information on each other even if you only measure one of them
These two properties allow developers to change how they run algorithms. Like for optimization: Normally you have to consider each person individually (in the table-setting case) and compare each one to all the rest. In quantum systems, you go into a superposition of all the states, create a phase wave of all the possible answers, and use interference patterns (like with noise canceling headphones) to cancel out all the solutions that don’t match your requirement of optimal.
She also discusses a new metric, quantum volume, designed to reduce error rates in computational prediction.
Resources for you
- Explore the Q experience library
- Read more about Talia’s experience
- Learn more about quantum computing and IBM Q
- Mateo Bengualid provides a programmer’s perspective on quantum computing
- Explore the QISKit, an SDK and API for Q computing; it comes with a developer’s guide to using the QISKit
- Try Bluemix for free