University of Southern California

Director Q&A

Q&A with Professor Daniel Lidar

USC-Lockheed Martin Quantum Computing Center scientific and technical director

Q:  What’s unique about this particular computer?

A:  D-Wave has built the largest quantum information processor to date. It presents an opportunity to investigate a host of fascinating scientific questions concerning the quantumness of the processor and its computational power.

Q:  How will USC be using the D-Wave system?

A:  We are and will be using this chip to conduct basic and applied research in adiabatic quantum computing, in which a problem is encoded into the lowest energy (“coldest”) state of a physical quantum system.  We’ll also use it to train a new generation of quantum scientists and engineers.

Q:  How powerful is the processor?   For what applications is it best suited?

A:  The D-Wave One Rainier chip is designed to contain 128 superconducting flux qubits, arranged in a two-dimensional 4x4 array of 16 cells of 8 qubits each. Within each cell, each qubit is connected to four others. Depending on whether the cell is in a corner or not, the qubits in each cell make 8 or 16 connections to the qubits in neighboring cells. At this time 108 of the 128 qubits are functional.

The system is kept at the exceedingly low temperature of 20 milli-Kelvin, and is extremely well isolated from its environment. We don't yet know exactly how powerful the processor is, and are planning to study that very carefully. It's designed to solve optimization problems, particularly those which can be mapped to finding the lowest energy state of a system of interacting magnets.

This captures a very broad class of optimization problems known as "quadratic unconstrained binary optimization". This includes many problems of practical interest, such as unsupervised machine learning, constraint satisfaction, image recognition and segmentation, and even protein folding, a biological process that must proceed correctly for normal functioning.

Q:   How does D-Wave's technology differ from other quantum computers under development?

A:  Most quantum computers being developed elsewhere are based on the circuit model idea, where logic gates are applied to qubits, similarly to the way logical gates are applied to classical bits in standard classical computers. The D-Wave chip instead uses adiabatic quantum computing.

Q:  How do adiabatic quantum computers operate?

A:  Adiabatic quantum computers, proposed by Edward Farhi and his collaborators at MIT in 2000, operate by mapping the answer to some hard problem into the minimum energy ("ground") state of a quantum dynamical system.

An adiabatic quantum computer finds the answer, or ground state, by starting out in the ground state of a simple dynamical system.  It then slowly ("adiabatically") transforms this system into one whose ground state encodes the answer to that problem. The quantum adiabatic theorem guarantees that if the transformation occurs slowly enough, and if the computer is isolated from external disturbances, then adiabatic quantum computation will find the correct answer.

The adiabatic theorem also estimates the time requirement. The time scales in inverse proportion to the energy difference between the ground state and the next lowest energy state. This difference tends to shrink as a problem becomes harder, so solving a harder problem takes correspondingly longer.

However, since it can use quantum superposition and tunneling, it is expected that the solution time may grow more slowly for an adiabatic quantum computer than for classical computers. For some problems the time may grow much more slowly, and this is where we expect to reap the quantum benefit.

Q:  Which kind of research will you do with the D-Wave machine?

A:  We're interested in developing new algorithms that can run faster on a quantum adiabatic optimizer, like the D-Wave chip, than on standard classical computers. We’re also interested in the fundamental physics underlying the D-Wave machine, particularly to understand to what extent it is quantum or classical. And, since we realize that the D-Wave machine will perform imperfectly, we’re interested in the development of error correction strategies. We plan to perform studies in all these topics using the D-Wave machine.

Q:  Could these studies be done using classical computers? 

A:  Yes and no.  At 128 qubits, our D-Wave One chip is still small enough that even if it worked perfectly as a quantum computer, its results could be accurately checked on a classical computer. In other words, any calculation that can be executed on the D-Wave One chip can also be done with current technology on a classical computer.   We’ll exploit that fact by using the D-Wave One to perform calculations, then checking the answers using classical computers.

However, using the D-Wave One to perform calculations is not the same as figuring out its much more complicated internal dynamics. In this sense, it’s actually a very large quantum system, and fully simulating its internal workings would be very difficult for classical computers.

For that reason, we plan to also use the D-Wave One to perform basic science experiments. We’ll effectively be exploring the chip in territory where classical computers can’t make accurate predictions, which dangles the possibility of new scientific discoveries.  Of course, we’ll publish our results, and fully submit our findings to scrutiny by the scientific community.

Q:  Many quantum scientists don’t believe that the D-Wave machine is a real quantum computer, and that no one has created devices with more than a handful qubits.  What’s your reaction?

A:  To me, the D-Wave chip is a fascinating science and technology experiment with the potential to revolutionize computing. I’ve followed D-Wave for more than a decade, and have never felt there was a controversy regarding the quantumness of the D-Wave computer. 

If anything, there were gaps between the scientific community's perception of D-Wave’s technology and the company’s public disclosure.  After initially protecting their proprietary technology, I think they’ve been exceedingly transparent for the past three or four years.

I’ve read their scientific papers, visited several times and seen D-Wave’s development first hand.   In my view, their recent results leave little room for doubt that they have developed a system worthy of serious attention -- and scrutiny -- by the quantum computing community.

One of our first priorities is to design tests that can provide additional, conclusive evidence about the exact nature of computation taking place in the chip.

Q:  State-of-the-art quantum computers created at universities have so far been limited to about 15 qubits,while D-Wave’s Rainier chip has 128 qubits.  Why such a large gap?

A:  There are several possible reasons. First, D-Wave is a commercial venture able to retain complete focus on its goal.  That degree of focus is very difficult to achieve in a university, where the focus is on creating new discoveries and publishing results, not on building complete, integrated products.  We’re also highly dependent on students, who by definition come and go, versus having fulltime staff who stay for long periods and are already experts in their disciplines. 

Another reason is design choice. Almost all university-based development of qubit devices has centered on the circuit model, which – incidentally -- was invented about 15 years before the adiabatic model by David Deutsch. 

Most circuit-model-based research aims to build a universal quantum computer that can solve any problem.  Research focuses on designing ultra-high quality qubits and gates that, in principle, can be used as universal quantum computer circuit components.   That’s an admirable goal. So far, though, it has yielded only small-scale quantum computers.

Conversely, the D-Wave chip is a special-purpose optimization engine.  They’ve deliberately sacrificed the ability to solve general-purpose computing problems to focus on scaling up their engineering approach.  The result may be a much more powerful system in far less time, albeit for optimization, as opposed to general purpose problems. D-Wave’s next-generation Vesuvius chip is slated to contain 512 qubits.  I look forward to finding out how it performs.