Industry Committee on March 29th, 2022

[Dr. Raymond Laflamme]

Ah, we're finally going to talk a bit about straight science and math.

The people who are doing algorithms classify them as two kinds: the easy one—the one that takes a very small amount of resources—and the hard one, which takes a large amount of resources. Of the two, if I want to be a bit more technical, one takes an exponential amount of resources compared to the other.

The algorithms that are resistant to quantum computers are the ones for which we haven't found a quantum algorithm that turns a hard problem into an easy one. The usual one, which is used for logging into your bank, called the RSA algorithm, is hard for a classical computer; that is, if you want to factor a large number with many bits in it, it takes an exponential number of time in the number of bits of the number that you want to factor.

We have a quantum algorithm that takes that algorithm to an easy one on a quantum computer, but there are other mathematical problems that are used to encrypt information that have remained hard on both a classical and a quantum computer.

[The Chair, Liberal, Joël Lightbound]

Thank you for your answer.

I'd like to ask you one last question.

Many of you have said that governments around the world are well aware of the risks that quantum computing can pose, primarily for national security. In fact, intelligence agencies are well aware of it.

Is the private sector, such as banks and other industrial institutions, also aware of the potential risk?

The question is for you, Dr. Laflamme, but Mr. Janik, Dr. Yazdi and Dr. St‑Jean will also be able to answer afterwards.

[Dr. Raymond Laflamme]

In Canada, the private sector, including banks, knows the potential of quantum algorithms. The question for banks, for example, is when they will have to change the algorithms they use today for quantum computer‑resistant algorithms.

[The Chair, Liberal, Joël Lightbound]

I guess it would be a little premature to make such a change now, wouldn't it?

Would it be too expensive?

[Dr. Raymond Laflamme]

That would be premature, because those algorithms are approved by the Canadian government and the U.S. government through the National Institute of Standards and Technology. The U.S. is involved in the approval process for some of the algorithms. The algorithms used in Canada are closely related to the U.S. algorithms. Once the U.S. approves them, companies will start changing their systems.

[The Chair, Liberal, Joël Lightbound]

Thank you very much.

Now I'll give the floor to Mr. Dong, who has one last question. We have only about two or three minutes left.

[Han Dong, Liberal, Don Valley North, ON]

Okay, I'll keep it short. Thank you very much, Chair.

I always have this wonder about AI versus quantum computing. I heard that you can't truly realize AI without quantum computing. I want to hear from the experts on this. Speak to process and the nature of the quantum principle.

Is that true? Perhaps you could educate us a little.

[Dr. Raymond Laflamme]

I'll make a comment and leave it to my colleagues to say more.

I have one little correction. AI is independent of quantum. There's a lot of AI technology on classical computers, which is being used in many different areas.

Then there is the question of whether quantum can really help AI. I think the question has not been resolved yet. There are indications that it could help, but we are not totally sure. There are research programs in Canada. I have one of my research staff looking at some of these issues right now, and other places around Canada will do that—

[Han Dong, Liberal, Don Valley North, ON]

I'm just going to ask something.

I'm looking at it from the perspective of ethics legislation when it comes to AI. If quantum computing will be used in AI technology, is there any concern with regard to the ethics legislation aspect?

[Dr. Raymond Laflamme]

It would be the same consideration as what you have for classical AI. There would not be a fundamental difference from the ethics point of view.

You might have a difference in how much AI you can do and the speed at which you can process that information, but the result would be fundamentally the same, and the information would be used or manipulated with AI ways of manipulation.

[Han Dong, Liberal, Don Valley North, ON]

Does anyone want to jump in with additional comments on this with respect to AI?

[Rafal Janik]

Maybe I'll just make one comment.

We have a very large team focused on quantum machine learning and quantum AI applications. A large amount of our software stack is actually focused on that.

I would say there are two types of quantum machine or quantum AI. The first is what we do today. That is really only thought to have an advantage when the data itself is quantum, so when you're looking at materials, chemistry and these types of problems, and maybe it'll be in a few other places, but that's what's believed.

The other is what I would call the end game quantum computer, the one we're all dreaming of when we close our eyes. That one will be able to accelerate fundamental basic operations that will actually speed up machine learning and AI applications. However, this is something that is further out than just a fault-tolerant quantum computer. There are a few additional components to that type of quantum computer that probably put it outside the 15- to 20-year type of road map. I'd say for the next decade or two, the impact of quantum and machine learning AI will probably be limited to problems with quantum data.

[The Chair, Liberal, Joël Lightbound]

Thank you very much, Mr. Dong.

I want to thank all of you for this great meeting. I think I can speak on behalf of all members of the committee and thank you kindly. It's been very interesting and we've learned a lot, and I think it's important for Canadians to better understand this technology and for governments to be prepared, so thanks for your insight. It's helpful in our study.

I wish you all a very good end of the day.

This meeting is adjourned.