Can cat qbits save quantum computing?

A whimsically-named quantum company named Alice & Bob actually has a quantum chip in the Google Cloud marketplace. Its “cat qbits” solve a massive issue that affects all other quantum chips. And it might just make quantum computing actually matter.

In this episode of TechFirst, we explore the fascinating paradox of building a quantum computer with Théau Peronnin, CEO and co-founder of Alice and Bob. We talk about the unique challenges and potential breakthroughs in quantum computing, discussing how Alice and Bob’s quantum chip aims to overcome the common problems of bit flips and phase flips.

Théau explains the concept of a universal quantum computer, the importance of error correction, and the revolutionary impact quantum computing could have on science, technology, and industry.

Watch: cat qbits and quantum computing

(Subscribe to my YouTube channel)

Subscribe to the audio podcast

 

Transcript: can cat qbits save quantum computing?

This is AI-generated … it is not perfect.

John Koetsier (00:01.56)
Will cat qubits reinvent quantum computing? Hello and welcome to tech first. name is John Kutz here. Quantum computing sometimes seems a little bit like test the full self -driving. It’s huge, impressive promises every year for a decade, but nothing never really seems to change. Maybe that’s about to end. A whimsically named quantum company named Alice and Bob actually has a quantum chip in the Google cloud marketplace. Solves a massive issue that affects all other quantum chips.

just might make quantum computing actually matter. And we’re going to talk to CEO and co -founder, Tho Pronen. Welcome,

Théau Peronnin (00:39.762)
done thanks for having me today

John Koetsier (00:41.838)
Hey, super welcome to have you. You’re in Paris. It’s 5 PM. You’re still talking, still in the office. I’m pretty sure that’s illegal. I’m pretty sure the EU is going to persecute you or prosecute you or something. Uh, but thank you for joining us. I want to start off with a big general question. Why are you building a quantum computer? Why are you building a universal quantum computer?

Théau Peronnin (01:04.04)
Yeah, I guess the most important part of why for me it’s such an important question is the sense of wonder. mean quantum mechanics always feel a bit magical but when you think about it, it’s just the best description of nature we have.

And building a quantum computer is really that. It’s damning nature’s inner gear to try to leverage all that sense of wonder, all those exotic rules of the game, and actually crack mankind’s most challenging issues with those inner gears of quantum mechanics. And that’s really fascinating to me.

John Koetsier (01:49.066)
No question about it. Super fascinating. Should be to everybody. Maybe define a second. What is a universal quantum computer? We talk about quantum computing. We talk about quantum computers. We don’t talk about universal quantum computers. What do mean by

Théau Peronnin (02:04.373)
Yeah, what we want to emphasize by adding this word universal is that it’s a general purpose quantum computer. So here we need to pose for a second. So quantum computing or quantum computers are not meant for everything. I mean, it’s absurd to try to use a quantum computer just to do a multiplication or something like that. But by stating that we’re here to build a universal quantum computer, what we mean is

we’re here to address all quantum algorithm with a single machine, just like your CPU on a classical computer can run basically any algorithm. It might be more or less well suited for some of them, but it’s still a very general and that what drove classical computing for 50 years at least. And so we’re building the same thing for quantum computing. And maybe to start

what will come next in this discussion, the fact that some of those algorithm actually requires to be able to do a lot of quantum operations and do what we call very deep algorithms. And to be able to run those, you need a machine that remains quantum throughout this lengthy computation. And this is actually absolutely not easy, not trivial at

John Koetsier (03:29.982)
So we’re not going to be doing word processing on a universal quantum computer.

Théau Peronnin (03:35.811)
yeah, well, you might, use it at some point to train, an AI, but this is still, pretty unsure. what, what quantum computers are really meant for at the moment, bear in mind that we just, we’re just getting started in discovering or inventing a quantum algorithm. But what we, we understand from that, from today’s point of view on the capabilities of the machine, it’s a machine made for a problem that

Let’s say small data, big compute. Let’s say for example, given a molecule, what is, how would it behave, how would it react? Given an extremely large number, how you can factorize it to decipher an encrypted message, for example. So those are problems that that requires very few data input. At the same time to be solved on a classical machine could.

require billions of years of computing on the best supercomputer.

John Koetsier (04:37.9)
Interesting. And so probably pairing it with classical computing in some scenarios is likely the future. So you can have sort of a general purpose compute platform, but you can assign tasks to the part that will do it most efficiently and most effectively.

Théau Peronnin (04:53.621)
Yeah, the quantum computer is really the heavy machinery, the big muscle you bring just to break that wall of computation. And actually, people are, mean, often I hear that quantum computers will speed up things, but it’s actually not a matter of speed. It’s such a speed up. mean, we’re talking about billions of billions of billions of fold speed up.

So it’s rather a change of possibilities of what you can reach, what is solvable with such a machine. And then, yeah, obviously you need to interface with the classical world. So you need a whole bunch of classical computing around it. And you can think about that just like the current craze that is happening at the moment with GPUs. A GPU cannot make a computer in itself.

It will still require a CPU, a central processing unit next to it to run and be integrated in an ecosystem.

John Koetsier (06:00.382)
makes tons of sense. Now there’s a problem of course with quantum chips, right? We’ve got bit flip and phase flip. They’re kind of the bane of quantum computing. Talk about those a little bit and what you’re doing to solve them.

Théau Peronnin (06:14.825)
Yeah, so that’s a rather technical question. But first, I think we need to emphasize the fact that building a quantum computer in itself is a paradox. On one hand, you want your machine to behave quantum mechanically. And you, myself, we do not teleport. We’re not at several places at once. We live in a classical and noisy world. So if you want something to behave quantum mechanically, it has to be perfectly isolated from the rest of the world. No information getting out of

John Koetsier (06:35.391)
Yes.

Théau Peronnin (06:44.715)
threading your cat box for the experiment. Now, at the same time, you’re trying to build a computer, a machine you can program input data, output results. And to do that, you need to open channels to connect, to control it. So in essence, a quantum computer is sort of a paradox. And so what we’re seeing at the moment is the fact that today’s machine, the early days of quantum computers are

very promising, but at the same time, they’re definitely not delivering this exponential speed up, the whole change of computing error, they’re promising. And the reason is that the noise from our classical world comes and pollutes the quantum computation. It creates what we call decoherence, the fact that those fragile exotic state becomes classical. Now, there’s two ways

for a bit of quantum information called a qubit to become classical or to suffer errors, should I say. It can suffer a very classical one, which is the bit flip, which switches a zero into a one and vice versa. It can also suffer from a phase flip, which is a purely quantum error, switching the phase of a superposed state, zero plus one into zero minus one. Long story short, in quantum you don’t have one, but two fundamental possible errors. And they’re both equally important.

And just to state the magnitude of the challenge, in today’s machine, in today’s early day quantum computers, they usually make about one error every 100 or 1 ,000 operations. And so that might seem not too big. I don’t know how familiar you are. But to put in perspective, this is 100 billion of billion of time more errors than a classical machine. So they’re basically noise generating machines at this point, in some sense.

John Koetsier (08:35.693)
Yes.

Théau Peronnin (08:40.939)
And so we need to solve that. We dramatically need to solve that. And so the community actually came with a breakthrough in the late 90s, early 2000, which is the idea that you can apply method of error correction designed for classical communications to quantum computing. And by doing so, you can correct those errors potentially faster than they happen so that you can create arbitrarily good

quantum computers. The trick is, and this is really the tough part, is that to correct for those errors, you need a tremendous level of redundancy. And here again, let me illustrate with a figure. If you wanna run show algorithm to break RSA 2048 because you hacked Bitcoin or those kind of things, well, the algorithm you’d wanna run,

theoretically only requires about 6 ,000 qubits. But with the standard approach, because of that burden of quantum error correction, you’ll end up requiring 20 million of those. So that means that 99 .9 % of your qubits are not here to compute. They’re here to correct errors. This is just how massive error correction is in terms of part of what makes a quantum computer.

John Koetsier (09:52.408)
Ouch.

Théau Peronnin (10:07.851)
And so indeed, this is where Alice and Bob comes into play. And we find a way to directly embed by design error correction within the qubit, within the physical system that hosts quantum information. And by doing so, we dramatically simplify the machine, actually by 200 fold. So instead of requiring 20 million to run this sci -fi use case of breaking

internet basically, you’ll only need a hundred thousand. So this is a machine that one can envision within the next decade for sure. And what is very exciting is that this sci -fi use case is actually one of the toughest. You have some very impactful use cases you can start to tackle much earlier on as soon as you manage to correct for those errors.

John Koetsier (11:02.4)
Amazing I was thinking as you were talking early on and and the the the quandary of Us being classical and the quantum computer being quantum saying quantum is in heaven. We’re on earth and the two don’t really communicate Very go

Théau Peronnin (11:21.353)
Yeah, and if you want to wonder a bit in how mind -blowing this machine is going to be, actually just designing the first, what we call the first logical qubit, the first bit of quantum information without errors, that means that you have a piece of your machine that shares no causal link with the rest of the universe. That is perfectly decoupled.

So you have kind of created such a black box that is perfectly isolating part of your universe from the rest of it. And I think this is just from a philosophical point of view really mind -blowing.

John Koetsier (12:04.522)
It’s incredibly mind blowing. mean, Schrodinger’s cat comes to mind, right? Is it alive or is it dead? Is it a one or is it a zero or is it a zero minus one? I it’s perfectly decoupled, but I need to couple because I need to give it a problem and then I need to actually see, get the answer at the end. So I need to connect at some point, paradoxical.

Wow. Okay. So you’ve done some cool stuff on error correction, reduced the number of qubits that you need by a factor of 200. What’s the impact of that? And how many qubits do you need for maybe not breaking the internet, decrypting Bitcoin, but doing truly unique and useful things?

Théau Peronnin (12:51.519)
Yeah, so the way we play, Alice and Bob, is that we’re taking the challenges from the toughest to the simplest. And so we started with the biggest part of the challenge, which was escaping the coherence. And so we’re not completely done there. What we demonstrated and what are cheap available on a…

on Google Cloud let you do is to witness that we solved half of that problem. We corrected for bed flips at this point, over eight or nine orders of magnitude. it’s not completely done, it’s way good enough. Now, what we’re working really hard towards is with our Helium 1 chip, the next generation that hopefully we will put.

on the cloud within the next year or so, is to correct for the remaining error, the phase trip, for a bit of redundancy. And this is sort of that Sputnik moment of decoupling from the rest of the universe, just like Sputnik escaped gravity. Now, the challenges that come later on is to scale that machine, to add more qubits until we reach impactful use cases.

John Koetsier (14:08.333)
Mm -hmm.

Théau Peronnin (14:15.019)
For us, the first stable orbit is around 100 logical qubit. And given how efficient or how powerful our architecture is, this would only require about 1 ,500 cat qubits. So it’s a big quantum computer, but it’s definitely within the realm of today’s enabling technology. So one of our competitors, IBM, for example, is already operating about 1 ,000 qubit. So it’s within

John Koetsier (14:36.82)
Mm -hmm. Mm -hmm.

Théau Peronnin (14:44.203)
order of magnitude. But because each of our qubit is so powerful, you’ll get a 100 logical qubit. And with that, the first use case you can start to tackle are typically deep science use cases. understanding spin chains, doing deep physics. And this is very interesting for HPC centers, for government or universities that want to push the forefront of science. Then as you add more and more qubit, you start unlocking more and more families of use cases.

John Koetsier (15:07.03)
Mm -hmm.

Théau Peronnin (15:12.075)
starting with material science, chemistry, a bit of optimization, then more chemistry, which is called biology, then come all the finance with the Monte Carlo and all the stats. Then you have most of the engineering with a large matrix diagonalization and the sum assumptions. yeah, then finally you can break Bitcoins and the internet.

John Koetsier (15:41.103)
So if all things go as planned, within perhaps a year or two, you might have the phase flip solved, or at least solved to an acceptable level. And then maybe a couple of years after that, or a year after that, then you’re thinking, hey, we can release an actual machine.

Théau Peronnin (15:59.967)
Yeah, let me try at giving a timeline. But as a physicist, I have to say this is still might get a bit delayed. But the first prototype of logical qubit we’re targeting for late this year, early next year, will take time to write a proper scientific paper. So no rush there, but we’re definitely not that far. Now, the next

on our journey would be several tens of cat qubits. And with that, by 2026, we should demonstrate the first minimal viable product of a universal falter and quantum computer, where you have all the features you expect, how you do data, how you do logical gates, how you can build the whole stack together. And very bullish.

roadmap I have to say aims for an industrial impactful machine by end of 2028 early 2029 for this hundred of logical

John Koetsier (17:06.19)
How physically large will that quantum computer be?

Théau Peronnin (17:11.851)
actually pretty reasonable. So 1 ,500 cats might very well fit in one of today’s dilution refrigerator. So this is something we need to say. Those chips are cooled down at very low temperature, about 10 millikelvin, which is basically 100 times colder than outer space. So it’s really damn cold. And so now you see them in pop culture and some TV shows. You see those.

golden chandeliers, which is the inside of the quantum computer. But in terms of footprint, it only occupy, let’s say, three to four square meters. And then you have a whole bunch of classical control electronics. The orchestra that governs, pilots, send the signal in, analyze, digitalize the signal that comes out of the quantum computer. And this might take a handful

John Koetsier (18:01.408)
Mm -hmm.

Théau Peronnin (18:09.545)
racks of just like in a data center. Yeah, maybe three to six, something like that, depending on the progress we make.

John Koetsier (18:17.358)
What we see in the classical world when we want to make something truly powerful, whether that’s a supercomputer or just a cloud setup, is we see massive parallelism, right? And we put 100 ,000 GPUs together and we put some fancy software around them and wiring and control and all that stuff. And boom, we have a supercomputer and it’s super fast because it does stuff in parallel.

Does the same thing apply to quantum computing or not really?

Théau Peronnin (18:48.189)
Yeah, actually it gets much better in my opinion. So there is, I don’t want to spoil too much, but we should have a paper covering this full. But the key message there is that as soon as you can do quantum error correction, it’s fairly easy to interconnect quantum computers or quantum processing unit, either within the same fridge or between fridges. And what is absolutely remarkable in quantum is

as soon as you manage to interconnect them, they truly operate as one single large quantum computer. And that can be extremely powerful. So I’d rather say that what it lets you envision is a world where you can on demand scale the machine depending on the size of the problem you want to tackle. And then in terms of parallelization, the SOM quantum algorithm,

try to sample, try to throw dices and get statistics and here obviously you can parallelize. But this is not the case for all quantum algorithm at all. Some are absolutely deterministic. This is also a common misconception that quantum computing is a type of probabilistic computing. It’s not. Some algorithm might be probabilistic, but other might be very well certain or deterministic.

John Koetsier (20:12.91)
Super interesting. Super interesting. Okay. Do you view this as a bit of an arms race? You’ve talked about a timeline. Your timeline is somewhere around 2028, 2029. There are many others trying to do the same. There’s geopolitics involved and there is the real possibility of if you invent something like this, breaking the internet, you said, decrypting Bitcoin.

breaking all cryptography. How do you view this technological challenge in sort of a political social frame?

Théau Peronnin (20:51.935)
Yeah, so first let me get out of this joke of breaking the internet because actually there are known algorithms, classical algorithms to encrypt classical data that we don’t know how to break with a quantum computer. We absolutely won’t break all encryption. We’re just going to break. I mean, for Bitcoin or other, they just need to fork and change their algorithms.

John Koetsier (21:08.461)
Mm -hmm.

Just half. No big deal. 25%.

Théau Peronnin (21:22.183)
And since it’s going to take a bit of time to build such a quantum computer, so they have time to adapt. Now, in terms of arm race, I don’t know about the term arm. What I’m sure is that there is a bit of a Los Alamos feeling. Kind of, I don’t know if you’ve seen the movie, Oppenheimer, but in the sense that we’re a bunch of physicists tasked with the mission of pushing the frontier of quantum physics or quantum information science.

John Koetsier (21:40.194)
Yes.

Théau Peronnin (21:52.475)
to not only push the science, but also immediately produce a usable technology in a very short timeframe with pretty large resources, actually. So indeed, it’s exhilarating. mean, it’s very exciting. And in that phrase, you have some types of players very different. You have some of the largest or biggest companies out there, the FANG companies, and also some smaller startups.

But when you look into it, say for example, us at Alice and Bob, we’re about 100, we raised about 30 million today, we’ll soon announce a big series B. at the same time, when you look in those very large corporations doing quantum, they’re not that big in terms of lineup of physicists. There is a possibility they might get a bit surprised by the outcome.

Now terms of geopolitics, which was your question, I’m not really concerned about the dual use aspect of the technology. I think it’s very marginal. What is more interesting is the, I’d say, economic sovereignty question. At the moment in Europe, you have this remark that there is just a fraction of the global GPUs that are hosted in Europe. And all that question of who will control the infrastructure underlying tomorrow’s economy.

because today’s economy is governed by data and very soon the level of innovation enabled by quantum computing will definitely push some geographies faster than others. So that have direct close access, strong ecosystems and so for sure governments have relied that and they’re trying to make sure the quantum valley, if there is one, will be on their territory.

John Koetsier (23:48.174)
makes a ton of sense and you can view that through the lens of the digital markets act, which of course, think 95 % of the companies targeted by the digital markets act are American, big tech basically. Right? And so every couple of months there’s a $3 billion fine, there’s a $5 billion fine, there’s a $2 billion fine, those sorts of things. And there’s all this pull push and…

And Europe obviously wants to maintain some level of sovereignty and control. And if you get to quantum computing, then you unlock all the things that you talked about earlier in material sciences, in biology, in other things like that. And that is where there’s real world impact medicines or different processes for making maybe solar panels, who knows that are 70 % efficient rather than 30 % or 20 % efficient. Who knows, right?

so many possibilities from all that stuff. And, yeah, we already see the results globally of chip making being really actually centralized kind of in Taiwan and Korea, two very vulnerable countries. If you look at them geopolitically, right. there’s a, there’s a chunk in the U S there’s a tiny bit in Europe and there’s a bit elsewhere. Right. And that’s pretty much it. So interesting, interesting world.

I just have one question to end with. This has been super fascinating and a lot of fun. Why is the company named Alice and Bob?

Théau Peronnin (25:15.593)
Yeah, it’s the name comes sort of a private joke. it’s those are two placeholders using textbook exercise or for experiment actually widely used in a in encryption. But I didn’t know that back in the time I discovered Alice and Bob as characters in physics textbooks. And so they refer to point A and point B physicists trying to make the thing a bit more lively, I guess. But the reason we picked it as a company name was

that we actually try to avoid the word quantum. They’re so polluted by pop culture to mean magical. There’s nothing magical about it. It’s counterintuitive, but whether you like it or not, this is our best description of nature. so it’s a, yeah, it’s this paradox we talked earlier about of building a quantum computer is just like a very tough textbook exercise, which requires an elegant solution, which we’re trying to solve with

by design error correction.

John Koetsier (26:16.898)
I really liked that. And it brings to mind what you talked about kind of up at the top of our discussion, which is that quantum computing is you’re dealing with the fundamental substrate really of reality of, of, of everything. Right. And, and, and, and, and,

That’s fascinating. That’s amazing. That’s our reality sort of emerges, bubbles up from that in some complex way that you probably understand a million times better than I do and probably are still baffled by.

Théau Peronnin (26:50.687)
Yeah, I mean, it’s a never ending question. I think that the fact that we cannot explain easily what is quantum in general and the fact that it’s so counterintuitive actually in my very personal opinion boils down to the fact that we haven’t yet completely understood or yeah, you know, there is this faint mind quote. If you cannot explain it, it means you haven’t understood it yet. So that’s kind of the point where and

To come back to your very first question, why am I doing that? I think this is actually one of my motivation as a former physicist is to try to better understand that. And there is this quote by Gaston Bachelard, a philosopher from the early 20th century, which is, understand nature by resisting it and trying to fight the natural tendency of quantum states to become classical is the best way, in my opinion, to better understand quantum computing, quantum in general.

John Koetsier (27:51.352)
That’s a great place to end. Thank you for this time.

Théau Peronnin (27:53.909)
Thanks!

TechFirst is about smart matter … drones, AI, robots, and other cutting-edge tech

Made it all the way down here? Wow!

The TechFirst with John Koetsier podcast is about tech that is changing the world, including wearable tech, and innovators who are shaping the future. Guests include former Apple CEO John Scully. The head of Facebook gaming. Amazon’s head of robotics. GitHub’s CTO. Twitter’s chief information security officer, and much more. Scientists inventing smart contact lenses. Startup entrepreneurs. Google executives. Former Microsoft CTO Nathan Myhrvold. And much, much more.