- 6 months ago
Qubits or not qubits ? State of the art of Fault Tolerant Quantum Computing in 2025
Category
🤖
TechTranscript
00:00Lovely to see you, wow, what a crowd we have today here at the Discovery stage, welcome, my name is
00:05Michael Rickwood, I will be your host today, and we do have a wonderful day here at the Discovery stage,
00:13lots of things to see, but I know of course why you're all here, but please do stick around, we
00:19have robots, gaming, health tech solutions, and this afternoon we will also have a segment from L'Oreal.
00:27But now let's take a more measured approach as we talk about something extremely interesting, complex, and that is of
00:39course quantum computing.
00:42Now we will have a segment called Qubits or Not Qubits, State of the Art of Fault-Tolerant Quantum Computing
00:51in 2025.
00:56I'm going to read this carefully.
00:59This report is the result of a collective effort by international renowned French experts brought together by the National Academy
01:08of Technologies of France to provide a comprehensive overview of the current and future state of error-corrected quantum computing,
01:20emphasizing industrial and technological aspects.
01:23So I would like to invite our two speakers please to the stage to take you through it, Olivier Ezrati
01:31and Catherine Lambert.
01:37Welcome.
01:46So, good morning ladies and gentlemen, we are very pleased with Olivier to present you this new report of the
02:02Academy of Technology devoted to quantum computing and exactly to fault-tolerant quantum computing.
02:08So, this report has been driven by the Academy of Technology.
02:13So, what is the Academy of Technology?
02:16It is a public institution which aims to provide to society to enlighten the technology and the issues they raise.
02:28So, we write and we provide reports and opinions.
02:34So, why fault-tolerant quantum computing?
02:37It seems very important to understand and to give an overview to the large public of what are the challenges
02:46besides fault-tolerant quantum computing.
02:48And that was the subject of this report and that was the subject of this report and it seems that
02:54it is a very original report and perhaps some unique report through the planet.
03:02So, there are today a lot of organizations, public and private, small and large, who are developing quantum computing.
03:13But the question is, what is really the possibility of such computers?
03:19So, that is the objective of our report.
03:24And the first question is, what about the quantum computing paradigm?
03:30So, we have the classical computers, we can simulate but only simulate quantum computers.
03:36And we have also analog quantum computers with the physics system.
03:46And so, we can simulate physics but we can solve only combinatorial problems.
03:52So, with digital quantum computers, that is we use qubits, we use quantum algorithms, so we use these qubits, a
04:01lot of qubits.
04:02And so, we are able to have perhaps extra competition with the classical computers.
04:13And the first generation is named NISC.
04:18With NISC, we are, because the problem of quantum computers is the noise inside these computers.
04:25And because there are a lot of source of errors.
04:29And with the first generation, the NISC generation, we have the possibility to correct this error.
04:36But the application and the use case of these computers are limited by the number of qubits we can access.
04:46And the new generation is the fault-tolerant quantum computing.
04:50So, with such computers, we are able to build universal, reliable and scalable quantum computers.
05:01And this is the subject of our report.
05:05And in order to do such a report, we, the Academy Technology was, as defined, a working group with, of
05:19course, some members of the Academy.
05:21But also international experts from the international French experts.
05:29And we have also the support of the French quantum ecosystem.
05:34That is both the experts from public institutions.
05:39But also the start-upers.
05:42Also the people from the industry.
05:44And also the ecosystem with Gen-C and Terratex.
05:49The HPC ecosystem.
05:51And it is important to have in mind that this working group also has the support of France 2030.
05:58and with Neil Arbrook.
06:01So, if you want to know what is this report, it's very easy.
06:06Don't hesitate to download the report on the website on the Academy.
06:15It is in French, but we will have very quickly a new report in English.
06:23And what is very important is that such a report is a very long journey.
06:29And such work is a very long journey.
06:32And this report is only the first step.
06:35So, we are continuing to work on this.
06:39And we aim to have an annual, a new report with an annual basis.
06:45By the way, we started two years ago.
06:47Yes.
06:47So, we had to update the report as we were producing it.
06:51Because there are so many news happening in that field.
06:53Yes.
06:53And this report is splitted in five parts.
06:58The first one is to better understand what is a quantum computing
07:04and what could be the quantum advantage.
07:07The second one is devoted to error correcting codes
07:11because without such codes it is not possible to build a quantum computer.
07:16The third one is to present the different physics
07:21besides the qubit technologies.
07:24And we present this physics.
07:26We present only the mature technologies.
07:32And another point is the scalability.
07:35That is how to manage all these qubits.
07:40How to manage the links between these qubits.
07:42So, that is the scalability.
07:44And the next one is only an initial analysis,
07:49technical and financial analysis,
07:51in order to have an industrial ecosystem
07:55devoted to fault computing computers.
08:02and for this, I give the floor to Olivier for the following.
08:07So, for starters, you have to remind us that what is the benefit of a quantum computer, potentially.
08:13The typical benefit that was thought out a long time ago is speed-up.
08:16So, how do you accelerate classical computing that's difficult to do right now?
08:22So, that's one of the advantages.
08:23But the other one that we looked on is also to improve the quality of the results.
08:27For example, when you do a chemical simulation, you need to have a better accuracy.
08:30When you do an optimization, when you solve an optimization problem,
08:34you want to have a better result, particularly if you are based on heuristics and statistics.
08:39So, but it's difficult to do that.
08:42It's very difficult for many reasons.
08:43One is there are few algorithms that are known today which brings an exponential speed-up.
08:48So, there's a lot of work to do in that space.
08:51We have a couple of ones and it happens that the most famous one that brings an exponential speed-up
08:56is not very useful.
08:57It's the one that breaks internet, sure.
09:00The other thing is it's expensive, it works well, but it's not good for data ingestion.
09:08So, the common wisdom is, okay, we're going to do big data.
09:12It doesn't work like this because the speed of the gas is not that good.
09:15And so, it's good for complicated problems, but not problems which require a lot of data, like an LLM, for
09:23example.
09:24The other thing is the qubits are imperfect.
09:27That's the reason why we want to do fault tolerance.
09:30The qubits are imperfect, there's noise, and if you want to make a comparison between a classical system and a
09:35quantum system,
09:36in a classical system, you make an error out of 10 power 18 to 20 operations, so very rare.
09:42In the case of a quantum system right now, we do an error for about 100 to 1,000 operations.
09:50It's a lot. So, when you accumulate those errors across all the operations,
09:54it creates so much errors that the end of the result, the calculation is not good.
09:59And the last one is, if you want to correct the errors, you need to use so-called error correction
10:04codes,
10:04and there are many ways to do that. And those correction codes, they had a lot of overhead.
10:09So, you need to multiply the number of qubits by at least one, two, three, four orders of magnitude of
10:15qubits.
10:15So, you multiply by the same order of magnitude the size of the cooling, the size of the electronics,
10:21the lasers that control the atoms or the ions in the case of those systems.
10:25So, it makes the system more complicated.
10:27So, in the end, the challenge is this mix of software design, hardware design, physics, all together.
10:35And the reason why it's complicated, it's all of that simultaneously.
10:39So, we need to fix a lot of things, and it creates a huge challenge.
10:44So, the main, I would say, the main topic of our report, thanks to different contributors,
10:51including Maziar Amirahimi, who is over there, was to look at error correction.
10:55So, how do you make sure that error correction works well at scale?
10:59So, there's some lingua to understand there.
11:02We have so-called physical qubits, which are the qubits that are in the systems,
11:05and they are prone to errors.
11:08And based on that, we add, let's say, a layer of software that corrects some errors,
11:14and it creates a so-called logical qubit.
11:17So, a logical qubit is a set of physical qubits, which, from the view of the developer,
11:22will have a much lower error rate.
11:24But it depends.
11:25It depends on the size of the algorithm.
11:27So, if you have a very large algorithm with a lot of operations,
11:30you need a large logical qubit.
11:32If you have a small algorithm, you need a smaller logical qubit.
11:36So, it's very dependent on your needs and the business deal and the application.
11:41But on top of that, even when you have done that, you need much more.
11:45You need not only to correct the errors, but you need to do that in a fault-tolerant manner.
11:50And one way to explain that, for example, which is simpler,
11:53is you need to do that faster than the error show up.
11:57Because if you're too small to correct your error,
12:00it's like if you are sinking on the sea.
12:02So, you need to correct fast the error.
12:05You need also to correct various sorts of errors.
12:08There are errors that happen on a single location.
12:10There are sometimes errors which happen simultaneously in different locations.
12:13You need to correct all of these.
12:15So, fault-tolerant is even more complicated than just correcting a single error in a single location.
12:21So, now when you look at the current market landscape,
12:25you have about 100 companies in the world who are trying to do that.
12:29It's a lot.
12:30It's a lot of companies.
12:32Of course, the famous ones like IBM and Intel and Google and others.
12:36But you have a lot of startups.
12:38And out of these startups, we have how many?
12:40Six French companies.
12:42So, we have six French companies that we studied a lot.
12:45And those companies, they are spread over the different kinds of qubits.
12:47So, we have Pascal for atoms. We have Alice and Bob for superconducting qubits.
12:52Candela for photons. We have Cobley in Grenoble for silicon qubits.
12:56So, they use equivalent of CMOS circuits to handle electron spins on this.
13:00And we try to figure out what were the pros and cons of these technologies.
13:05And at this point in time, you can't say which one is going to win.
13:08But we have some ideas of their scalability challenges.
13:14And if you want to bet on one or the other, you will be in for a surprise.
13:20Because if you look at the various kinds of ways to evaluate the quality of these qubits,
13:24nobody is perfect.
13:26You have red, green, yellow, orange, everywhere, all over the place.
13:29And so far, we don't know who's going to win.
13:32And we could elaborate on that. Of course, we do not have the time.
13:35But what you have to know is the future of quantum computing is probably going to be heterogeneous.
13:41It means that some of those technologies may be good for memory.
13:44Some of these may be good for communication, typically photons.
13:47And some may be better for computing, with fast computing, like superconducting qubits or silicon qubits.
13:52So it's probable that in the future, there's going to be a mix and match of all these technologies.
13:59Then we have the challenges.
14:01Yes.
14:01So what kind of challenges did we identify?
14:03Yes, because the question is to have such a computer.
14:09The world will be very long to have a manufacturing, an industrial manufacturing quantum computer
14:15and some useful applications.
14:17The one is long and it is part of a lot of challenges.
14:21And there are some challenges here.
14:25So we have identified, of course, challenges about the scaling in order to have a million of qubits.
14:34And with the connectivity and zero correction, we have also the challenges because we need to have strategic technologies
14:43in order to build these computers.
14:46That is, for instance, cryogenic electronics or lasers.
14:50And we need also to develop algorithms, software engineering in order to test, to design, to test and to implement
15:02this
15:03in order to have useful application for industrial use cases.
15:06And that is very important also is the hybridization with HPC, with the high performance computing.
15:15Another point is the benchmarking methodology in order to compare all these computers.
15:20And, of course, all the other technologies, that is AI, that is HPC, they are continuing to advance and to
15:31make progress.
15:32So it is important to have this competition with the quantum computers.
15:36And the last, and it is not the least, it is the skills and the funding in order to be
15:42able to build such a computer.
15:44I would add one thing, is the interconnect, because one of the specifics of our report is the interconnect thing.
15:52Because we know that it's going to be difficult to put all the qubits we need in a single processor,
15:57whatever the technology.
15:58And so we uncovered all the technology challenges to interconnect various quantum computers.
16:04And there are many ways to do that.
16:06For example, in supercomputing qubits, you can use flexibles with microwave cables.
16:10In the case of photonic qubits or ions or others, you can use photons, optical photons.
16:16And it's very challenging.
16:18And we happen to have a company in France doing that, Relink.
16:21So we really dig into that and you will not find that in most of the reports that have been
16:25published so far.
16:26So it's one of the specifics of that report.
16:30So, that is only the presentation.
16:32If you want to know more, there is a report on the website of the Academy.
16:36But if you have any question, we have time to answer to your question.
16:40And we have also three other persons from the working group who are able to answer to your own.
16:45There's one guy you were there who will ask questions.
16:47Of course.
16:47All right. Okay. Thank you very much.
16:49So, are you excited about quantum computers?
16:53Yes?
16:57You are. Okay, good.
16:58So that then begs the question, why are they excited about quantum computing?
17:03Huh? Computers.
17:04Well, probably because they think and they're right to think about that.
17:07That it can solve interesting problems.
17:09Yeah.
17:10Because we don't do so-called techno-solutionism.
17:14We don't do a technology just because it's fun.
17:16Because it's a scientific challenge.
17:18Yes.
17:18It's interesting because, potentially, those systems could solve problems in chemistry, energy, or in drug design, in optimizing various processes
17:27in various companies.
17:29That's the usefulness of those systems that is appealing, I would say.
17:32So, basically, quantum is going to allow us to really push research into all of the disciplines, whether it be
17:37physics, chemistry, in sciences, right?
17:40Yeah.
17:41Anything else?
17:41There's a duality over there.
17:43Okay.
17:43Because quantum computing is going to help, first, mostly researchers, industry researchers,
17:49academic researchers, industry researchers, to develop new stuff, advanced science.
17:54Yes.
17:54And, on the other hand, there are other algorithms, typically optimization algorithms, which are going to be useful mostly for
18:00industry and business operations.
18:02Okay.
18:02It's a different kind of solution.
18:04Yes.
18:05We dream of an advantage of quantum computing, so there is this dream, so it's important to have this.
18:14So, what I understand, I mean, we're talking about something that's going to be, I mean, at a performance level,
18:19I mean, vastly superior to what we're used to today.
18:21We're talking about speeds which are much faster, a calculation, resolving complex calculations at a much faster pace.
18:30I mean, just from a kind of layman's point of view, because I'm not in computer science, can we compare
18:36what the kind of computers we're using today
18:38to what a quantum computer might look like in, say, and it's very difficult, I know, to put a timeline
18:44on this,
18:45because, I mean, no one knows when we're going to really start seeing these.
18:48But could we use an analogy to say, okay, so today's computers are basically like a hot air balloon, and
18:55quantum computer is basically like a SpaceX, or, you know?
19:00Well, it's a tough one, because we like to use classical benchmarks to compare things, so like gigaflops per second,
19:08and so on.
19:08In the case of quantum computing, you compare not the speed, you compare the result. That's benchmarking. So you look
19:16at what you can achieve.
19:18But the best way to compare is not to compare. Let me explain why.
19:22Okay, right.
19:23It's to identify problems, but there's no classical solution. Well, it's not possible. It's not possible to get a good
19:30chemical accuracy. It's not possible to solve a combinatorial problem.
19:34And then we can do that. So you're saying that quantum can actually begin to address problems that we cannot
19:40solve on our own ourselves?
19:41Yeah, it's making the impossible possible. That's the goal.
19:44Does that mean that people might...
19:45Because it's not just improving the speed by a factor of, let's say, 100% or whatever.
19:49It's extending the reach of computing. That's the thing.
19:51So basically, it might help us to solve climate change, because we're not doing a very good job of it
19:55right now.
19:57Well, climate change, well, to some respect, there are some applications which could impact indirectly climate.
20:02Yeah.
20:03But we shouldn't oversell that.
20:04The other thing which I'm involved in myself as part of, as being a co-founder of the Quantum Energy
20:09Initiative,
20:09is to make sure that the new products we create are good with respect to the energy they consume.
20:16So it's a challenge, because the size of the systems may be large.
20:19And so we need to make sure that there are technology options to make sure that the systems don't consume
20:25too much.
20:27Okay, all right, all right.
20:28So I have a lot of personal questions. There are some questions I have here.
20:33All right, what are the barriers to getting this built over the next...
20:38Can you give us any kind of timeline of when we might start seeing Quantum?
20:43We need to have access to these computers and to develop what is very important,
20:49to have also the possibility to develop algorithms adapted to these computers.
20:54Yes.
20:55So we need a lot of research today in order to have such competencies.
21:02And the other very important is to have skills, because it's important to have people who are working on this.
21:11So the skills are very important in order to have more computers and to be able to use them.
21:18Okay, so basically the things maybe holding it back as development is that you need the right skills,
21:23the right talent, people to work on it.
21:25Okay, so a careers opportunity for people there to...
21:29We need you to help develop it.
21:31There's one way to respond to that.
21:32Material as well, right?
21:34Material.
21:34There's one way to respond.
21:35Most of the large companies in that space and the startups, including the French startups,
21:39they have a so-called roadmap.
21:41They explain what they want to do.
21:42They say in X number of years, we're going to have this and that.
21:46So if you take Candela, Alice and Bob, Pascal, just here in France,
21:50they have a roadmap for the next 10 years.
21:53So we will be able at the Academy of Technology to compare their plans and what they do, actually.
21:59Okay.
21:59And IBM is the same.
22:01So we can look at what they do.
22:04And what's interesting is it's a very open world, I would say.
22:07I mean, the technical and scientific challenges, they are known.
22:09They are shared by the academic world.
22:11There's a lot of exchange.
22:12It's an international endeavor.
22:14We all are looking for the same thing.
22:17And it's a worldwide quest, I would say.
22:20Not just single companies.
22:21But what…
22:22Okay.
22:22So I think I read recently that some of the big, big tech giants like Google, for example,
22:27are making huge, huge research into this.
22:29I mean, are they not going to be driving this forward?
22:31Is there going to be a race between the huge tech organizations to get this?
22:36It's a race, yes?
22:37Yes.
22:37Yes, it's a race.
22:39If you look at the size of the investments, the investments for those large companies,
22:43they are not that large.
22:44I mean, if you compare that with large European companies, the ratio is not one to one thousand.
22:50It's maybe one to two, one to three.
22:51It's not that large.
22:53And we bet also on different technology choice.
22:56So in France, for example, we have Alisson Bob.
22:59They selected the technology which reduces the cost of error correction.
23:03Because there's some part of the correction that's done in the hardware.
23:06So IBM and Google didn't choose that.
23:09So we have a differentiation in the technology.
23:12So with that, we can probably, for example, develop systems which may be cheaper, develop faster.
23:20And I mean, the cost of a system is more important if you do the same stuff.
23:24Yes, but it is interesting.
23:25So I see, from a European level, we have an interest as well to research this, develop it, and push
23:30it forward.
23:31Can we get some questions from the audience?
23:33Anyone?
23:34Yeah, please.
23:35We'll just bring a microphone now for you.
23:37Oh yeah, perfect.
23:38Hello.
23:39Can you hear me?
23:40We're speaking to the mic.
23:41Can you hear me now?
23:41Okay.
23:42So which of the realization technologies for implementing quantum computing do you believe that is the most advanced right now?
23:49And we'll have the bigger potential for solving real problems, like for example, chemical problems, something like that.
24:00I didn't get the question.
24:06Do you want to just repeat?
24:08I didn't understand the question, sorry.
24:10Let me rephrase that.
24:11What's the question?
24:12So you mentioned ion traps, you mentioned...
24:14These are the kinds of qubits we have.
24:15Yeah, all the different implementations just to realize a quantum computer.
24:21Which do you believe that is the most advanced right now and the one that will win?
24:25Okay.
24:26You've seen the slide.
24:27There's no answer.
24:27Yes.
24:28I mean, one is most of them are here in the room.
24:32I want to keep friends.
24:35Because in our work at the academy and in our daily job, we work with all those companies.
24:39So we won't say this one is going to win.
24:41Most of them have high potential and a lot of challenges.
24:46It's a mix of engineering, scientific challenges, sometimes software challenges.
24:51We have to understand that.
24:52It takes time to understand that.
24:54But you can't say this one is going to be better than that one.
24:56No, it's difficult today to answer to you.
24:59But we know that it's important to have all these physics and all these technologies in order to know in
25:07two years
25:08what could be the best one and what could be the best one for such applications and such use case.
25:16And by the way, there's another answer, which is we are lucky that there are so many choices.
25:20Because we are in the early stage of that development.
25:23And it's good we have all these companies and all these technologies because we don't know yet.
25:27It's better to have a kind of hedging your bets.
25:31So many bets around.
25:32There's enough money to do that as of now.
25:36So it's good we have all these options.
25:38And maybe sometimes one is going to hurt the other.
25:40So let's wait.
25:42Thank you.
25:43Any other questions?
25:45Yeah, over here.
25:46All right.
25:49Please come.
25:52All right.
25:54Thank you.
25:55Can you hear me?
25:56Yeah, yeah.
25:56Please, please come forward.
25:59So I'm not a tech expert, but I know that AI started in the 60s, but it entered our daily
26:06life only.
26:07a few years ago, especially after ChatGPT, and it is also entering the whole industrial landscape.
26:14My question for you is, do you know or can you say in how many years quantum computing will enter
26:20the industrial landscape on a similar scale to what we see today with AI?
26:26That's difficult.
26:27If you have an answer.
26:27That's a difficult question.
26:29Yeah, yeah.
26:31So it's difficult to say the number of years, but perhaps a few years.
26:37No, I don't know exactly when will be the date, but I think it will be in the future and
26:44not too far.
26:46Well, everybody is expecting a ChatGPT moment.
26:50We probably won't have this because it's not a consumer product.
26:54So it's going to be business problems solving, but the size of the market depends on algorithm design.
27:01So we don't know yet.
27:02Yeah, it's a bit like that.
27:03It's a bit like that.
27:04Just a remark.
27:06Of course, it's not possible to answer exactly to the question, but it's also normal that this question arises in
27:14the mind of everyone.
27:15Okay.
27:16And a way to answer this question is to consider what are doing the agency, the national agency in France,
27:25for instance, with the Proxima project or in the U.S. with the DARPA agency, what they want to do
27:32in the five coming years.
27:34So it appears that at least for these two big programs, the national program in the U.S. or in
27:41France, it appears that 1931 or 1932 is a key date to examine what has been achieved at that time.
27:51Okay.
27:52And to choose.
27:532030?
27:54Sorry?
27:552030?
27:562032.
27:582032.
27:58Okay.
27:58You heard it here, guys.
28:002032.
28:002032.
28:03Just actually, I have a last question very, very quickly because that one was, he drew on the example of
28:08AI and ChatGDP.
28:10So, now, AI is being embraced today.
28:15We know its benefits, but we also are very, very aware of its risks and the uncertainties around it.
28:20What are the uncertainties and risks around Quantum?
28:25Well, the main risk that people are talking about is breaking the internet.
28:29Yes.
28:30Yes.
28:31And crypto.
28:31Because potentially, it could do that with a famous algorithm called the Shor Integral Factoring
28:36algorithm.
28:37So, it could break asymmetric public keys on the internet.
28:40Yes.
28:40But, we know that the resources that the quantum computer would be needed to do that are enormous.
28:47The last estimations were about one million superconducting qubits from Google, another one with 100,000 physical qubits from Alice
28:56and Bob.
28:56These are the two best estimates we have right now.
28:58We know it's going to take time to do that.
29:00But, hopefully, we have a solution.
29:03I mean, there's a solution to that.
29:05It's called post-quantum cryptography.
29:07It's one of the solutions available.
29:09It's been standardized by the East in the US and, let's see, in Europe.
29:12We are starting to deploy those solutions, which will change the existing classical cryptography that's being used on the web
29:20everywhere.
29:20I mean, even my VPN operator has the solution for that.
29:25Okay, listen.
29:25We really have to stop there.
29:27We're being told off.
29:28Ladies and gentlemen, I'd like to thank our guests, Olivier and Catherine.
29:34Thank you.
29:38Can they find out more information about you from a website?
29:42Yeah.
29:42Easy to Google.
29:43Easy to find you.
29:44Okay.
29:45Thank you very much.
29:47Excellent.
Comments