- il y a 1 jour
Quantum Computing: What Leaders Need To Know
Catégorie
🤖
TechnologieTranscription
00:00Good afternoon. So I'm really pleased to have two such eminent experts on quantum computing here with us today at
00:11VivaTech.
00:12For those of you who don't know, Alain Aspey won a Nobel Prize in 2022 for his work on quantum
00:23entanglement.
00:24And Jay is in charge, is the vice president in charge of IBM's initiative on quantum computing.
00:36So we're going to dive right in to talk about how business leaders should prepare for the coming quantum computing
00:48wave.
00:49So let's start with something that many of us have already heard about, and that's the security issue.
00:57So the technology could be used to break today's encryption, such as RSA, leaving data that is currently safe vulnerable.
01:09Should business leaders be worried? How long will it take to upgrade all businesses to quantum safe cryptography?
01:20Would you like to start, Alain?
01:22Well, first it's very difficult to understand what you say because of this big, big noise.
01:28But yes, I think that all this quantum computation started because of the Shure algorithm, which could allow one to
01:42factorize larger numbers.
01:43So it would put in question the security of our codes.
01:48But it's clear today that a Shure algorithm would not be the first application of quantum computing.
01:55And so I think that what happens with the security and the cryptography is that we have to think about
02:06it in a different way.
02:07We have to think that it's very likely that many people in the world monitor everything and store the data.
02:18And if after a few years they can decipher, then it will be a big problem.
02:25And to counter that problem, there is only one thing I know, it's quantum cryptography.
02:31And I think that quantum cryptography is necessary even if we don't have the Shure algorithm immediately.
02:38But I don't think that we are here to mostly speak of quantum cryptography.
02:44We are rather to speak about quantum computing, right?
02:47Yes, absolutely.
02:48But the security issue is one that is top of mind for business.
02:53So, Jay, I mean, I know that there are standards that are already being worked on and that are being
02:59used.
02:59Maybe you could talk a little bit about that and also about the length of time that it takes to
03:05prepare.
03:06So, even if it's not the first problem, it will be a problem and businesses do need to prepare.
03:13Yeah.
03:13So, I can add to what Alan said.
03:16So, Shure's algorithm was definitely the algorithm that launched this field.
03:22It's a beautiful piece of mathematics and it breaks efficiently RSA, which is a lot of the encryption we use.
03:30But as Alan alluded, there's two ways to look at this problem.
03:35One way is there are new algorithms that are emerging that we, the mathematicians, feel very confident they will not
03:44be broken by a quantum computer.
03:46So, one way is to go to those algorithms that cannot.
03:49And as scientists, we're often very ambitious that the future is going to be quantum computers, quantum communication.
03:57And so, security will have to keep changing.
03:59And quantum key distribution and quantum encryption is a beautiful theory that will definitely have some impact.
04:07So, I would say, first, there are new algorithms emerging that are basically being standardized by most governments around the
04:16world that we should go to.
04:18And we should keep investing in quantum key.
04:20The other part of your question is, I don't think we should necessarily fear it because we have solutions.
04:29But how to change it is a big society problem because cryptography is used everywhere.
04:36So, the more complicated problem isn't going to quantum, quantum safe or quantum encryption.
04:43It's actually how do we actually change security.
04:46So, I know a lot of people say, oh, it's coming to the end because we'll get quantum computers.
04:51If we're proactive, I'm very optimistic that there are various ways to get around it.
04:56Maybe a last word on quantum cryptography before switching to quantum computer.
05:03I think that it is extremely encouraging to see all these papers which appear where people say they are able
05:11to distribute entangle photons under Manhattan, here or there, just by using the commercial networks of optical fibers.
05:21And, of course, I am especially delighted because this is the technique invented by Arthur Eckert, which is based on
05:29my entangled photons.
05:31You know, Arthur came to see me 30 years ago and said, do you know that with your entangled photons,
05:36I can do quantum cryptography?
05:37It was the first time I heard of it. So, I am happy to see all these people distributing entangled
05:43photons below Manhattan and below many places like that.
05:46So, maybe it's time to go to quantum computer.
05:51And I would just, it turns out that I'm just arriving from the island of Helgoland in the North Sea.
06:00Here, exactly there, exactly 100 years ago, Heisenberg discovered the first commutation relation, etc., etc.
06:11And so, in this conference, there were many people, and I have very good news about neutral atoms.
06:19It's not to say that you don't have good news, but I have very good news about neutral atoms.
06:25Michel Houquin was Quira, which is a competitor of Pascal, and Antoine Boës.
06:34I also gave a call to Antoine Boës yesterday night.
06:38And they are extremely optimistic that the logical qubit is really, we are very close to the break-even when
06:46we have really error-corrected logical qubits.
06:51So, this is good news, you know, because those of you who have heard me for months and years maybe
06:57heard me saying,
06:58we are not going to see that in my lifetime, okay?
07:03And I become optimistic, provided my cardiologist keep me alive a few years, probably I will see it.
07:13I agree.
07:18Jay, you want to comment on that? I mean, you also believe that quantum is at an inflection point.
07:24Sure. So, look, we've been as scientists battling understanding how light atoms interact, whether it's in neutral atoms or superconducting
07:39qubits.
07:39I remember when I first got into this field in 2004, it was understanding how a superconducting qubit would interact
07:48with a microwave cavity.
07:49And I was really excited when we saw it have long lifetimes.
07:53And then we've entangled two gates.
07:56But we've got to a point in the superconducting field that I have the same level of excitement.
08:02I don't see fundamentally any science roadblocks to build large scale, what we call large scale fault tolerant quantum computers.
08:11So, at IBM, we've pretty much laid out exactly what we're going to do over the next four years.
08:18And it's on us to go and show that we can engineer these large systems.
08:22And I share the same optimism.
08:26I think by the end of this decade, fault tolerant quantum computers will be a reality with hundreds of logical
08:33qubits doing hundreds of million operations.
08:36And the thing I would say is I would love to see more people doing the algorithm research because these
08:42things are actually going to come.
08:44Whether it's neutral atoms or superconducting, I may be a bit biased.
08:47I understand others may be biased the other way, but they're going to come.
08:52Well, you have to realize that I don't feel that as an harsh competition.
08:58I see that as a good and healthy competition.
09:01And there is plenty of healthy competition in this domain.
09:05For instance, take, again, neutral atom because I know it better than you.
09:11But when Antoine Bouhès, using his machine in Institute Optique, which is a machine at the origin of Pascal,
09:18he solved the problem of condensed matter called the easing problem.
09:23A problem that you cannot solve with a classical computer by brute force.
09:29And he shows he can do it for 200 or 300 particles.
09:32And then the people of classical computing try no kind of approximation.
09:39And with their approximation, they can do the calculations that they could not do and they can compare.
09:44This is exactly what I call healthy competition.
09:48Because on the one hand, Antoine Bouhès has a result, but he has nothing to compare to.
09:53And on the other hand, the people of classical computing can compare.
09:58So it's benchmarking. And I think that really the efforts of supraconducting and neutral atoms have to go hand in
10:09hand,
10:09especially on the side of the algorithm and of the mathematics.
10:14It's clear that there is, if I understand well, a good collaboration between IBM and Pascal to train to have
10:23similar algorithm that you can try on one platform or on the other platform.
10:30I think this is very good news.
10:32Yeah, so it has always been our goal.
10:36I 100% agree that the art of science is about moving the bar.
10:41So any competition between classical and quantum is good.
10:46I think I would pull on the point to say I would caution everyone trying to think that every big
10:53experiment that's going to come out over the next few years is a flagpole event.
10:58Because I expect us to show as a field examples of quantum advantage or whatever term you want to use.
11:06Classical will respond and we'll keep seeing this.
11:08And as was said, I am optimistic.
11:14So fault-tolerant quantum computers will exist.
11:18They will be capable machines.
11:20We need algorithms to do it.
11:22But I think the more general statement is the future of computing is going to be what I like to
11:29refer to as heterogeneous accelerators.
11:31There will be quantum computers.
11:33They could be analog-based quantum computers.
11:36They could be superconducting-based.
11:39They could be neutral atoms.
11:40There will be HPC.
11:42And there will also probably be GPUs.
11:45We are probably in the most exciting time for algorithm discovery.
11:49And if we can make all of these systems work in a way that is easy, we actually enable this
11:57to happen.
11:58And so one of the announcements that we have made is working with Pascal.
12:02We have actually made it very simple to program on both of these systems.
12:07But I think this is just the start.
12:08There's no reason why we cannot extend this to every type of accelerator there is.
12:15Well, another thing that I know regarding neutral atoms is that the debate that people had two or three years
12:24ago like do we insist on doing simulation or do we go digital?
12:31Well, this is finished.
12:33Now the same platform in Pascal, the same platform can either operate in the analogous, I don't know how to
12:42say, analogical way or digital.
12:45And this I think is remarkable also.
12:48I think really we live a fantastic time where you can try many different things simultaneously.
12:55Or not simultaneously, but on the similar machine.
12:58Can we talk a little bit about how AI might help accelerate quantum and the reverse, that quantum could help
13:05accelerate AI?
13:07Okay.
13:08Well, AI, I don't understand well enough to know how AI can help quantum computing.
13:14But there is one thing that I have heard and which seems to me extremely interesting.
13:21It's the idea that quantum computers could produce data to train AI.
13:28And the reason why I think that this is a very good idea is that if you think of it,
13:35a quantum computer is something which is real physics.
13:39You have a machine.
13:40It's an experiment of physics giving data, real data.
13:45That's the way the world works.
13:48And so you can probably train and feed AI with this kind of data.
13:54I like the idea.
13:55I'm not an expert at all.
13:57Maybe it's crazy, but at least I like very much this idea.
14:01Because again, a quantum computer is a real physical machine.
14:08So, where we are seeing is what we're finding and what is interesting, and this is why I think we
14:18often say is AI versus quantum.
14:20I do think it's the wrong way to look at it.
14:24A quantum does a new type of math, a math that we're very excited to scale it as a competition.
14:29So, then you can ask the question, can you use that math in the AI, whether it's data or whether
14:35it's different types of calculations?
14:37Some of the work that I find exciting is people have shown that certain data that a classical computer would
14:45take exponential time to classify,
14:48quantum could classify it faster.
14:50That's a toy problem, not useful for anything, but it shows that this math that quantum computing offers gives us
14:58a different lens to look through AI.
14:59So, my view on quantum is as it matures and as people discover algorithms, whether it's putting data into it,
15:06whether it's doing the math, there's a lot more to be done.
15:10And I actually look forward to a future where you have quantum computers, quantum sensors, quantum communication.
15:17And imagine if you could put quantum data into a quantum computer, what could that do?
15:25Maybe a word about quantum communication, a word about quantum teleportation.
15:32I think it's absolutely crucial because if you have a system where you have several entities, if you have a
15:40machine producing a quantum result, it would be stupid if you need to use it for another calculation, not to
15:48transfer it as quantum.
15:50And this is all the idea of quantum teleportation.
15:54And for this, there is really one thing which starts working, but that demands to be much improved and developed,
16:03it's quantum memories.
16:04You need to make a network and to transfer quantum information.
16:10You need quantum memories.
16:11And when young people ask me, what would you advise us to do?
16:15I say, if you want something with a good result, not too long ahead, why don't you work on quantum
16:22memories and you will be at the heart of this quantum system.
16:29I am sorry because you cannot follow your equations.
16:35It's not a problem.
16:36You cannot control me.
16:39That's the fun.
16:41That's the fun part.
16:42So let me just ask a little bit.
16:46We talked a little bit about the timeline.
16:47You said IBM is confident it could have a fault-tolerant computer by the end of the decade.
16:54Startups and big technology companies are taking radically different approaches to quantum computing.
17:01Based on your knowledge of the market and the current development, which approach to quantum looks most likely to succeed?
17:10Well, I certainly cannot answer that question.
17:14From a superficial point of view, a startup is more agile than a big company.
17:20But on the other hand, when I visited Jay in New York, I had the feeling that his lab is
17:27not like a monster of a big company.
17:29He probably, his boss, gives me the possibility to be agile.
17:34So the question is more to have an entity which has not too much bureaucracy and which has enough flexibility.
17:44And if it belongs to a big group, that's okay.
17:48Except that I ask you the question.
17:49What happens in the IBM stock goes down?
17:53And he told me, no, no, no, I still will have money to fund that research.
17:59So I think it's more a question of being, he's still in basic research.
18:05And a startup is still in basic research.
18:08I mean, really exploring.
18:10He doesn't know.
18:11Pascal doesn't know.
18:13Quiera doesn't know what will be the machine 10 years from now.
18:17So we have to go step after step and see what happens.
18:21And this can be done only in entities which can evolve rapidly.
18:27But he may say, probably, he will confirm that he's able to evolve.
18:33Look, yeah.
18:34When you work in a technology, you're obviously committed to the technology.
18:37But I think we covered this before.
18:39It's better to focus on creating an industry first.
18:45And creating an industry we need to get algorithms to happen.
18:48So I actually think there's two things that need to happen.
18:53We need more algorithm research.
18:54We need to test these devices.
18:57Because there are many of them that are at a point now that they're beyond what we can simulate.
19:02Whether it's our systems, Pascal systems, Quiera systems, and various other startups out there.
19:08We need to get that going.
19:10We need to push the limit of science.
19:12But science is always about you have hypotheses.
19:16You build it.
19:17You test it.
19:17And you keep iterating.
19:21I don't think answering which question is the best technology from...
19:26I'm going to, of course, say mine because I've dedicated my life to it.
19:29But that's what scientists do.
19:31And when they see a roadblock, they will change.
19:34But if I give you the answer that I think is the correct answer is...
19:38Let's focus on actually bringing every technology into the hands of people as fast as possible.
19:43So we can see what they can do.
19:45You mentioned...
19:46I'd like to follow up on something you said earlier.
19:49You mentioned the combination of quantum computers and quantum sensing.
19:54What will that combination bring us that we can't do now?
19:59So there's always these ideas like...
20:03So today we have communication and we have sensing type things.
20:09Generally, if you want to make communication go further, you need something like a repeater.
20:14Or if you want to make sensing go further, you need to get to what's known as the quantum limit.
20:20Generally, this requires creating very exotic states in the machine or the sensor.
20:27If there are ways that we can combine the ability to do the gates and operations with the ability to
20:34put the data in...
20:35We can envision doing these type of things that are shown at some type of scale.
20:40So what I've always wanted to see is, yes, we're going to build quantum computers.
20:46Our roadmap says by 2029 we'll build 200 logical qubits, 100 million operations.
20:54Then I want to go to a couple of thousand logical qubits and a billion operations.
20:58You get to a point, even when you're looking at that, that you have to now think,
21:05I need to communicate that quantum computer to another quantum computer to allow me to keep scaling.
21:12And if I've got to now communicate a quantum computer with another quantum computer,
21:17yes, you can do teleportation, but you need the quantum degrees of freedom to interact.
21:25So you need the quantum computer to interact with the quantum optics
21:29and then do the teleportation Alan talked about.
21:33Since I see the time we are approaching the end, there is one thing which really excites me.
21:39It's a fact that my hope, my great hope is that neutral atom on one hand
21:46and your approach superconducting qubit on the other hand can address the same kind of problems
21:52and compare and do benchmarking.
21:56Because it's a real question. You have a machine which produce a result
22:00and you say, oh, there is no way with a classical computer to check it.
22:05But if there is another quantum computer and that you can compare and benchmark,
22:09I think that it is a remarkable situation.
22:12And this is what makes me optimistic.
22:16When I listen to you, when I listen to the result of Pascal, of Cuerra, of Boes, of Lucan,
22:22I think at least we have two different technologies which apparently work hand in hand, so to speak.
22:31Well, a little bit pushing, but not too much.
22:33I agree.
22:35Okay, so with just a little over a minute left, what do you advise businesses to be doing now to
22:44prepare for the shift to quantum computing?
22:47And how would you advise them to evaluate quantum computing for problem solving at their companies?
22:56Well, in a few seconds, you need to have as much effort on applied mathematics and quantum informatics as on
23:07hardware.
23:08The two must progress in parallel and the good thing is that on the side of quantum informatics and applied
23:15math, then really we can share.
23:17Yeah.
23:19So, I agree.
23:22So, for those that obviously use security, you should actually understand your security and have someone looking at that to
23:30understand what you need to do.
23:32In the quantum algorithm or computing side, our advice has always been if you're using a lot of compute and
23:42you're doing that to do a lot of compute to do your type of problems,
23:45you should start investigating the quantum computers to do it.
23:48If you're more a user of libraries and things like that, you have to stay informed.
23:55Because this technology is not, in my view, not going to happen.
24:01It's going to happen and you don't want to be playing catch up.
24:05But would I put many, many people on it if I can't get the best of it?
24:09The answer is no.
24:10So, depending on how much you use compute in your current business, that's how we would advise how much you
24:16should be dedicating.
24:17And they have to be applied mathematicians, quantum information type researchers and engineers.
24:26Yes.
24:26And we must give tools to people who have ideas.
24:31Oh, if I had a quantum computer, we must give them the tools.
24:34And the tool is not only the hardware, but also the libraries and everything.
24:39And the training of the students, of course.
24:42So, it's a very exciting time in quantum computing.
24:45And we've heard from, you know, two worldwide experts.
24:49Please give a nice round of applause for our panelists.
24:53Thank you.
Commentaires