January 15, 2021
Two CES 2021 panels addressed the current state and anticipated advances in quantum computing, which is already being applied to problems in business, academia and government. However, the hardware is not as stable and robust as people would like, and the algorithms are not yet up to the task to solve the problems that many researchers envision for them. This has not stopped entrepreneurs, major corporations and governments from dedicated significant resources in R&D and implementations, nor from VCs and sovereign funds making major bets on who the winners will be.
One panel, titled “AI and Quantum Cyber Disruption,” was a discussion with Dr. Vikram Sharma, founder and CEO of QuintessenceLabs in Australia, and Pete Tortorici, director of joint information warfare at the DoD Joint Artificial Intelligence Center.
The other panel was titled “Quantum Computing — Making it Real,” and was an energetic discussion between Dr. Joseph Broz, executive director of the Quantum Economic Development Consortium (QED-C); Katie Pizzolato, director of IBM Quantum Strategy and Applications Research; and Dr. Eleanor Rieffel, senior research scientist and Quantum Artificial Intelligence Laboratory lead at the NASA Ames Research Center.
Sharma declared that we are in the second phase of the quantum portion of the fourth industrial revolution. The fourth industrial revolution is driven by the confluence of AI, 5G, robotics, autonomous, immersive media, and IoT. The first phase of the quantum portion involved passively harnessing quantum effects. The second phase, of which we are now on the cusp, involves engineering quantum states and effects that don’t exist in nature, similar to leap required in materials science to make super-materials. By manufacturing these quantum effects, we are opening up a whole host of new capabilities in computing, imaging, sensing and cybersecurity.
QuintessenceLabs is focused on the cybersecurity piece. Quantum computing can protect against breaches. Planning a quantum implementation can also help the organization think about what a quantum-safe security posture would look like. The general consensus is that a cybersecurity adversary with quantum capabilities will appear within 5-10 years. This will particularly impact public and private key encryption.
NASA Ames’ Rieffel pointed out that the National Institute of Standards and Technology (NIST) has already recommended larger key sizes in response to quantum development, regardless of what the quantum approach ends up being. For online transactions, symmetric key encryption is relatively safe against quantum crypto, but she agreed that public key encryption is easily cracked by it.
QED-C’s Broz said that companies can start planning immediately for quantum security. He noted that ex-Google head Eric Schmidt said at a conference last week that companies need to start thinking about quantum security now.
We have not yet seen a ‘near peer’ use AI in an attack, DoD’s Tortorici added, but it is a known and expected threat. He pointed out that AI can both build distrust into crypto systems and breach systems.
QuintessenceLabs is working on 3 areas: 1) using quantum to generate true random numbers, 2) having very good key management to handle hundreds of millions of keys as we move to a highly encrypted IoT world, and 3) the frontier world of quantum key distribution.
Discussing the fundament infrastructure, Rieffel said that the hardware needs to be much more robust and fault-tolerant to do something useful. If what you really want is certified random numbers you may be in luck in the next couple years, but if you want to do materials science simulations or optimization, it will be a longer wait. She added that what we learn about quantum algorithms has often fed into classical algorithms. Quantum has inspired better classical algorithms.
One business application for quantum computing, said Broz, is a situation where it can solve critical and difficult problems, unlock new potential in markets, and then businesses can calculate their increased revenue potential and the return on that quantum investment. IBM’s Pizzolato added that the IBM cloud-based approach is perfect for driving that value chain.
Pizzolato explained that quantum computing will only be good for a certain subset of problems. It should be integrated into traditional workflows so it is in concert with its classical computing counterpart. Rieffel added that the front-end programming is the same as traditional computing. The virtual infrastructure in the cloud can look at the code and determine if it should be routed to a quantum or a traditional computer.
Sharma and Tortorici were asked what recent development in their field stands out and impresses them.
Sharma said that in December 2019 Chinese researchers came up with a solution within minutes to Gaussian Boson Sampling (GBS). In 2011, MIT professors posed the problem, saying that solving it would demonstrate quantum advantage. The same problem would take 2.5 billion years to solve on a supercomputer.
Tortorici mentioned the DALL-E application on OpenAI GPT-3. The application can create a picture in the style of Salvador Dali from a string of words. It appeared to be able to ascribe context to the words. It even knew how to draw the shadows for the images. That level of inference is of high interest to his team for tactical military applications, he noted.
The moderator asked what do you expect to see in the next year.
Broz made three predictions: 1) there will be a cloud-based demonstration of quantum supremacy this year that will validate useful algorithms, 2) there will be big strides in translating customer problems into quantum algorithms, and 3) there will be leaps in middleware development in the next year.
Rieffel anticipates exciting advances in hardware and algorithms, but she doesn’t expect those advances to translate into significant improvements in the applications. Also, she doesn’t see the hype around what company will “win” quantum computing going away.
Pizzolato thinks that rapid advancement in use will spur a lot of software tool development. “I would be surprised and pleased if there were meaningful advances in correcting the errors in these systems,” she said, referring to Rieffel’s earlier comment that the systems need to be more robust, stable and fault-tolerant.