Podcast: The Google-IBM “quantum supremacy” feud

Was it a breakthrough or a snooze? In October 2019, Google scientists announced they’d achieved “quantum supremacy,” the long-sought proof that a computer built around the strange properties of quantum mechanics can, at least in certain cases, carry out calculations exponentially faster than a computer built around classical bits. Researchers at IBM, one of Google’s main rivals in the race to commercialize quantum computing, pre-empted them with a claim that Google had exaggerated its quantum computer’s advantages and that quantum supremacy wasn't a meaningful achievement anyway. MIT Technology Review’s editor-in-chief, Gideon Lichfield, visited both companies on a quest to understand their disagreement and learned that it goes much deeper than it seems to.

Inside the race to build the best quantum computer on Earth, from the March/April 2020 print issue, p. 38

Here’s what quantum supremacy does—and doesn’t—mean for computing, September 2019

Quantum supremacy from Google? Not so fast, says IBM, October 2019

An exclusive interview with Google CEO Sundar Pichai on achieving quantum supremacy, October 2019

Our quantum explainer series:

Audio ID: This is MIT Technology Review.

Gideon Lichfield: What’s going on here is that IBM isn’t just skeptical that Google achieve quantum supremacy in this particular instance. It just thinks quantum supremacy is not very important. And what I was trying to understand was why. Why did they think that?

Wade Roush: For decades we've been promised quantum computers. With their almost mythical power, these machines could solve hard problems and unlock new breakthroughs in science. Last fall Google claimed it had taken a big step towards building the first useful quantum computer, and IBM immediately shot down that claim. So what's really going on? Technology Review’s editor in chief Gideon Lichfield explains why the rivalry between these two tech giants goes even deeper than it appears, and why the dispute over quantum supremacy matters for the rest of us. I’m Wade Roush and this is Deep Tech.

[Theme music]

Wade Roush: Now, in a minute, I’ll talk with Gideon about exactly what it is that Google accomplished last October with its experimental quantum computer, called Sycamore, and why IBM was not impressed. But first, I think it helps to acknowledge right up front that quantum computing is weird. It’s built around behaviors that are absolutely real at an atomic scale, but that seem a little unreal to us at our human scale. So, to get us ready to talk about this stuff, I want to take you first to downtown Boston, where I got a friend of mine to help with a musical demonstration.

Wade Roush: Tell me your name and tell me where we are.

Heinrich Christensen: My name is Heinrich Christensen and I am the music director at King’s Chapel in Boston. And that’s where we are right now, in the organ loft.

Wade Roush: King’s Chapel has a beautiful pipe organ, and I went there to see if Heinrich could create sonic analogies to three of the weirdest ideas in quantum computing. So, you know how a traditional computer operates on bits that are either on or off representing a one or a zero? I asked Heinrich to represent that by just playing two separate notes.

[Organ music]

Wade Roush: Think of the low note as a zero and the high note as a one. The first weird but true idea in quantum computing is called superposition. The heart of a quantum computer is a collection of quantum bits or qubits, and if you can keep a qubit isolated from the outside world, you can get it into this state of superposition where it isn’t a zero or a one. It’s kind of both at the same time. Now you could represent that by playing the high note and the low note simultaneously.

[Organ music]

Wade Roush: But the math of quantum computing actually says that when a qubit is in a state of superposition, you have to describe it with a kind of smear of probabilities between 0 and 1.

Heinrich Christensen: Right. So that would sound like this.

[Organ music]

Wade Roush: It’s not until the end of a computation when you measure a qubit that this smear of probabilities collapses back into a classical one or zero. The second weird idea in quantum computing is called entanglement. If two quantum particles or two cubits are entangled, their properties or fates are linked up in a way that lets them act in unison. And that’s what makes quantum computers exponentially faster at some jobs than classical computers. And when I say exponentially, I mean that literally. If you have some number of entangled qubits, call it n, and they can represent two to the nth states at the same time. So, two qubits can represent four states. Three qubits can represent eight states. Four qubits can represent 16 states, five qubits can represent 32 states and so on. I would have asked Heinrich to play 32 notes, but he ran out of fingers. The point is that a quantum computer with just a few dozen cubits could in theory do certain computations faster than the world’s most powerful classical supercomputers.

Wade Roush: There is one last phenomenon that makes quantum computing different from classical computing, and it’s called interference. It’s like waves in a pond overlapping. I asked Heinrich if he could play two notes on the King’s Chapel organ that were so close together that we could hear the sound waves interfering.

[Organ music]

Wade: What you’re hearing there is a pulsating change in volume as the notes from the two pipes interfere constructively and then destructively. And as it turns out that you can program a quantum computer to use an analogous type of interference to amplify the correct answers and cancel out the wrong ones. Listen for it again.

[Organ music]

Wade Roush: Thank you, Heinrich.

Heinrich Christensen: Thank you!

Wade Roush: Now, the analogy between music and quantum computing is not what any computer scientist or physicist would call precise. So please don’t take anything you just heard too seriously. But now I think we’re ready to meet Gideon. For his feature story in the March-April issue of MIT Technology Review, he went to a Google lab in Santa Barbara, California, and an IBM lab in Yorktown Heights, New York. And he talked with the scientists building some of today’s most advanced quantum computers.

Wade Roush: Gideon, thanks for being on the show.

Gideon Lichfield: Thank you, Wade.

Wade Roush: You’ve been to both Google and IBM to see their quantum computing labs. Why did you go to see these guys?

Gideon Lichfield: So last September, a paper leaked online that was written by researchers at Google that said that they had achieved this thing called quantum supremacy. They’d gotten a quantum computer to do a calculation that they reckoned the most powerful classical supercomputer on the planet would take 10,000 years to do. And they had done it with a quantum computer in three minutes. So the paper leaked. Google wasn’t quite ready to publish it, but a month later, they did, in fact publish it. And they invited me and a bunch of other journalists down to their lab in Santa Barbara to see the computers and to talk about what this discovery meant.

Gideon Lichfield: Two days before we were all due to show up in Santa Barbara, IBM published its own paper in which it said Google had basically got it wrong. And this classical supercomputer wouldn’t take 10,000 years to do the calculation. It would take only a couple of days. So we were there to witness this Google milestone, which they’re describing as something like the Wright Brothers, the first flight of the Wright Brothers’ Flyer for quantum computing. And IBM is saying no, this wasn’t the Flyer. This was just, you know, this was the Wright Brothers testing that their engines started or something like that.

Gideon Lichfield: So there was this immediate face-off, this battle between the two giants over not so much about who got there first, but whether or not the achievement was really what Google was saying it was. After that, I got very interested in why IBM was so intent on debunking Google’s claim. And I talked to them. In fact, around the same time of the Google announcement, and then I went down to visit their lab later.

Gideon Lichfield: What’s going on here is that IBM isn’t just skeptical that Google achieved quantum supremacy in this particular instance. It just thinks quantum supremacy is not very important. It thinks that that proof, that moment of demonstrating that you’ve got a quantum computer to do something way, way faster than classical one is not actually very relevant. And what I was trying to understand was why. Why did they think that? Why was something that to everybody else seems kind of obvious—you got a quantum computer to do something nobody had ever done before—why isn’t that an achievement? IBM really deeply believes that that is the wrong thing to be talking about. That it’s not a significant milestone. And I wanted to understand why.

Wade Roush: When you go and visit these labs, what did you see when you walk into these places? Can you kind of paint a picture for us of a Google facility or the IBM facility or both?

Gideon Lichfield: So what you see in these labs, principally, I mean, there’s a lot of equipment lying around and, you know, measuring devices and stuff. But the main thing you see is a cylindrical steel drum, probably a bit bigger than an oil drum. And it’s hanging from a scaffolding rig that is meant to damp vibration. And when that drum is taken off, what you see is the thing that they call the chandelier. It looks kind of like a chandelier. Somebody once wrote about it and called it a steampunk chandelier. It’s this multitiered thing full of brass and on wires and loops of stuff. And what it is, is a cooling system. It’s a dilution refrigerator. And it cools things in successive levels. At the very top of the fridge. It cools things down to about 4 kelvin, 4 degrees above absolute zero. And then with each successive level down, it gets colder and colder until at the very bottom it’s 15 millikelvin, fifteen thousandths of a degree above absolute zero. And inside that is a small silicon chip. And that is where the qubits, the actual quantum computer sits.

Wade Roush: When you go into one of these labs and you see this steampunk extravaganza chandelier thing, do you come away thinking, ‘Wow, that’s incredibly cool, we’re on the edge of a revolution?’ Or do you come away thinking, ‘Man, that looks like something out of a bad movie? It’s going to take forever to get real quantum computing.’

Gideon Lichfield: When you look at one of these things in the lab, it looks very homebrew. But I think what you get the sense that this is what the early days of the technology looks like. When I was at IBM lab and Jerry Chow was showing me around, he was pointing to some of the machines that they have. And he said, look, this already looks much more sleek than the rat’s nest of wires that you have in some of our earlier machnes.

[Cut to recording from Gideon’s visit to IBM’s Thomas J. Watson Research Center in Yorktown Heights, NY]

Jerry Chow: So this is one of our primary research labs, where we’re doing a lot of the throughput of devices to make them better.

Gideon Lichfield: How many machines do we have in here?

Jerry Chow: We have five machines in here. The pumping you hear are the pulse tubes for the refrigerators.

[Cut back to studio interview]

Wade Roush: Right. So my understanding is that both IBM and Google are using the same core technology to embody their qubits, using these things called Josephson junctions.

Gideon Lichfield: They’re both using the same basic technology. So we’re at the point with quantum computers that we were, let’s say, with vacuum tubes back in the old days of computing where people are trying all sorts of different ways to build a qubit, to build a basic element of computing. And there are I don’t know, what, 10 or a dozen completely different ways of making qubits right now. There’s only a couple that are really in the lead, but there are many, many different ways of trying to do it. Many in other words, all of these are different ways of making a simulated atom. So IBM and Google have both chosen something that is called a superconducting transmon qubit, which consists of this thing called a Josephson Junction. Basically what it is, is it’s two little strips of metal that are superconducting when they’re kept very cold. And then there’s a very, very thin gap in between them about a nanometer wide. And the way that electrons move across that gap is basically what creates the quantum behavior.

Wade Roush: When you were in Santa Barbara, how did the Google folks react to the fact that IBM had basically tried to puncture their balloon a couple of days before? What were they saying and feeling about IBM coming along and saying, ‘Wait, hold up, guys. Maybe it wasn’t quite as astonishing as you’re saying.’

Gideon Lichfield: They were, at least on the surface, unbothered, but it was clear that they were a little bit bothered. So first we have this press conference. The Google team is out there talking about what they achieved and why it’s important. And then one of the first questions from a journalist is, ‘OK. So what do you think about IBM’s claim that you guys didn’t really achieve anything that significant.’ And I remember that Hartmut Neven, who is the head of the Google quantum lab, said something that basically didn’t address the question. He kind of dodged it. And it was clear to me that he just didn’t want to go into this detail. Later, I spoke to John Martinis, who is the guy in charge of the hardware within Google’s team. And I asked him the same question. What about this IBM paper? Do you do you think that that that claim is significant or not?

[Cut to recording from Gideon’s visit to Google’s Santa Barbara lab]

John Martinis: I’m kind of surprised what they’re doing, because I think it’s clear to most people that this is a big advance. So, you know, it’s nice that they did it. And, you know, we’re opening up our software so that they can model the thing. We’d like for them to actually test it. And if they validate things we’ve done, hey, you know, that’s great.

[Cut back to studio interview]

Wade Roush: He’s saying, ‘Oh, well if they’re saying they can actually do this calculation in two and a half days, Show us. Do it.’.

Gideon Lichfield: Exactly.

Wade Roush: All right. They haven’t done it, by the way, right?

Gideon Lichfield: They haven’t.

Wade Roush: OK. So we’re talking about very complicated machines and very deep math and very hard physics. But at some level, it seems like we’re also just talking about language. And I wanted to ask you to explain where this term quantum supremacy even comes from, and why has it become so contested?

Gideon Lichfield: So when John Preskill coined this term quantum supremacy in 2012, it was still a little controversial whether we would ever be able to build a quantum computer that could do something faster than a classical machine, because you sort of don’t really know what’s going on inside the guts of these things. You can only do all kinds of experiments to try to deduce it from its behavior from the outside. So Preskill was saying if we can demonstrate in just one specific case that a quantum computer is way, way faster than a classical machine we will have proved that it’s possible. And that will put at least that debate to rest and then we can get on with developing them.

Wade Roush: So from that perspective, Google really did achieve, quote unquote, quantum supremacy. They met Preskill’s challenge.

Gideon Lichfield: Yes, they did. And pretty much everybody in the quantum computing world that you speak to, except the people at IBM, will agree that this meant something, that there was a significant milestone achieved.

Wade Roush: So when IBM comes along and says, ‘Sure, you may have achieved quantum supremacy, but how practical is that? And we could probably do that on our giant Summit computer anyway. Just give us a couple days,’ what are they really saying at IBM?

Gideon Lichfield: IBM’s objection to Google’s achievement has many levels. So at the most basic level, or the most superficial level, rather, it’s a semantic one. They don’t like the term ‘supremacy’ because they think the public will misinterpret it as meaning that now, quantum computers can do everything faster than classical ones. OK. It’s a fair objection. Beyond that, what they say is that achieving quantum supremacy in this one narrow case doesn’t really prove anything. And so IBM is focused on something that it calls quantum advantage. This sounds like a semantic distinction, but it’s not for IBM. The idea is we shouldn’t be looking for one particular moment of quantum supremacy is as a milestone. What we should be doing is just trying to continually build better quantum computers, make them bigger and make them faster and gradually increase the number of cases in which they can do some things somewhat faster. It’s not that they’re going to destroy beat all classical computers into the dust. It’s that they’re going to be a bit faster, fast enough for it to be economically worthwhile to use them on certain problems. And so that is what IBM means by quantum advantage. It’s a gradually growing number of cases in which quantum computers have an advantage. Their philosophy is that what IBM is there to do with quantum computers is to deliver products that will serve its customers and help them achieve higher efficiencies or to work faster. That, I think, is what underlies this otherwise rather hard to understand dispute between two companies about what from the outside seems like just a matter of terminology.

Wade Roush: What are the stakes here for the rest of us? Why does it matter whether Google or IBM are a little bit ahead at the moment in the quantum computing race?

Gideon Lichfield: So what’s at stake in quantum computing? The promise is that quantum computers will be able to do certain things that classical computers basically cannot. And the kinds of applications, the kinds of useful applications that are most often talked about involve things like modeling chemical reactions or weather patterns. And this could be important because particularly in things like drug discovery and material science, we’re running into a bit of an innovation wall. It’s getting harder and harder to discover new materials and new drugs that can move medicine forward or move, for instance, battery technology forward. And at the moment, the way that we do this in the lab is, scientists play around with molecules that they think might be promising and do experiments on them and work their way through the space of possible molecules. You can do some of this kind of modelling now with supercomputers and AI but the idea with quantum computers is that they might be able to actually accurately contain the model of a molecule of a complex molecule and really predict exactly what it’s going to do. And so that could just bypass a whole lot of the lab work. It could allow you to explore a much, much larger number of potential drugs or potential materials and identify which ones are actually going to be useful. So for overcoming this innovation gap or this slowdown in a lot of the science that’s really important to us as a society, quantum computers could play a big role.

Gideon Lichfield: Now, why should we care whether Google or IBM comes out ahead? In some sense, I don’t think we should. I mean, ultimately, these are two very big companies. One represents the Silicon Valley culture of innovation and agility. One represents the staid, institutional, steady as she goes. But each of them is also trying to evolve away from what they have been in the past. So I think the only thing that matters, maybe the thing that is relatively important here, is simply that there is competition between them and between other companies as well to build the first quantum computer. The fact if we get progress in this in this field, it’ll be because these giant companies with hundreds of millions of dollars to spare are throwing resources at the problem and trying to solve it. Whether or not IBM believes in quantum supremacy, I think it’s going to have to attain quantum supremacy again and again on its computers in order to make them viable, to make them useful to its customers. Whether or not Google believes in quantum advantage, it’s going to have to keep on increasing quantum advantage in order to keep on making its computers better and faster and more useful to its customers. So they may hate each other’s terminology, they may hate each other’s concepts, but I think they’ll end up following much the same much the same route.

Wade Roush: The March-April issue is the TR10, the 10 Emerging Technologies issue, and quantum supremacy is on the list. So, why?

Gideon Lichfield: Because we thought that it was actually a significant achievement. In other words, to some extent, I guess we buy Google’s narrative. People have been talking about quantum computers for a long time. We’ve actually featured them in the top 10 list in the past. But this really felt like a milestone, a step that brings them significantly closer. And the TR10 is all about identifying breakthroughs that we think are going to have an important impact in the next three, five, maybe 10 years. And this just felt like one of them.

Wade Roush: If that’s your threshold, that there will be a practical impact in the next three to 10 years, what you’re saying is you feel like we’ve reached that level. We were at that point now where quantum computing could become something that has a real-world impact within three to 10 years.

Gideon Lichfield: Yes. So with Google’s achievement of quantum supremacy, we’ve entered what people call the noisy intermediate scale quantum era, the NISQ era. And what this means is we can now build quantum computers that can do probably something useful that will have a few hundred qubits, but will be noisy, meaning that they’ll be susceptible to errors and to stopping working after a few seconds because of those errors. Nobody really knows what those will be useful for, but it’s a fair bet that there will be some applications that that they can be useful. And so something with a few hundred qubits, which we might be able to see built in the next three to five years, let’s say, could actually have a practical application.

Wade Roush: So this is really one to keep an eye on.

Gideon Lichfield: I think it is.

Wade Roush: Thank you, Gideon.

Gideon Lichfield: Thank you very much, Wade.

Wade Roush: That’s it for this edition of Deep Tech. This is a podcast we’re making exclusively for subscribers of MIT Technology Review, to help bring alive some of the people and the ideas you’ll find in the pages of our Web site and our print magazine. But the first four episodes of the show cover our annual 10 Breakthrough Technologies issue. So we’re making those episodes free for everyone.

Wade Roush: Deep Tech is edited by Michael Reilly, with editorial and production help this week from Jennifer Strong and Jacob Gorski. Our theme is by title card Music and Sound in Boston. Special thanks this week to Doreen Adger, John Akland, Elizabeth Bramson-Boudreau, Linda Cardinal, Angela Chen, Heinrich Christensen, Kyle Hemingway, Katie McClain and Eric Mongeon. I’m Wade Roush. Thanks for listening. And we hope to see you back here in two weeks for our next episode.