What we talk about when we talk about qubits

Ed. note: This story was originally commissioned by Scientific American. After a difference in editorial vision, we came to an agreement and decided to kill the piece. I think there are still some valuable conversations to be had about the topic, so I’m posting a mostly-unchanged version from the second round of edits. The title, for those unfamiliar, is a riff on Raymond Carver’s What We Talk About When We Talk About Love.

Update (2/17/23): A preprint by Kobrin, Schuster, and Yao finds that three key signals claimed as evidence by Jafferis et al. are mirages. In particular, the signals for thermalization and teleportation* match the SYK model only because Jafferis et al. take the average over their data—a fact they do not clarify in the Nature paper. At the level of individual functions, their signal is not in agreement with the SYK model. I went into more detail here, but the upshot is that the case for Jafferis et al. being deliberately misleading is now even stronger.

Correction: The signal for teleportation in Jafferis et al. is not an average from the data. Clarification here.

Alice is talking. She is a physicist, and so Bob, a computer scientist, listens. “Do you remember my neighbor—you know, the one who’s always asking about quantum mechanics,” she asks. Bob nods over the phone, even though she can’t see him, and Alice continues. “Well, the other day he brought up all this—” she gestures inarticulately “—stuff about wormholes and quantum computers. I tried to explain, but I got stuck just describing a qubit.” Bob finishes his drink and shakes his head. “You didn’t want to trot out the standard line and tell him it’s both 0 and 1,” he asks. Alice bites her lip. “No. No, I thought I’d tell him it’s neither 0 nor 1 before we measure it,” she says. “State’s not definite, right? So we shouldn’t say it is both.” Bob scratches his chin, following the logic but not necessarily its conclusion. “Sure, but then you haven’t answered the question of how to plainly say what it is, only what it isn’t,” he prods. Alice frowns. “But I can’t very well tell him the qubit’s a linear combination of complex vectors in a two-dimensional Hilbert space.” 

Here, Alice and Bob, perennial participants of thought experiments, demonstrate a typical challenge in talking about quantum mechanics—its rejection of the familiar logic of possibilities. Finding plain, accurate language requires navigating a minefield of quantum conundrums. Unsurprisingly, as knowledge of quantum computing has spilled out of academic journals and labs to a curious public, miscommunications have abounded about the nature of this strange new tech. 

The latest quantum drama began late in November with a heavily discussed Nature article, tantalizingly entitled “Traversable wormhole dynamics on a quantum processor.” How could research so redolent with science fiction potential—legitimated by prestigious institutions including Google and Caltech—not lead to immediate attention? 

Quanta declared, in a 4,000-word article: “Physicists Create a Wormhole Using a Quantum Computer.” (The headline was subsequently changed to clarify the wormhole was “holographic.”) The New York Times also stated that the scientists made a wormhole, but qualified the study as “Physicists Create ‘the Smallest, Crummiest Wormhole You Can Imagine’.” Subsequent pushback led to headlines such as “No, physicists didn’t make a real wormhole. What they did was still pretty cool.” The “wormhole” study quickly became the talk of physics twitter, and frequently came up, in this reporter’s experience, at the Q2B conference in Santa Clara.

To be clear: outside of very speculative mathematics, there is no evidence whatsoever that wormholes actually exist in our universe. Even if they did, current theories of gravity strongly suggest such spacetime portals are unlikely to ever be stable enough to be traversable.

I should tell you, when my editor at Scientific American sent me the embargoed press release on Thanksgiving with the subject line “Hmmm,” I responded dubiously. “They used 9 qubits! What the hell could that possibly tell you?” We chose not to cover it—initially.

That skepticism was shared by many expert onlookers. Harvard theoretical physicist Matt Strassler ran through a list of objections on his blog: “Did physicists create a wormhole in a lab? No. Did physicists create a baby wormhole in a lab? No. Did physicists manage to study quantum gravity in a lab? No. Did physicists simulate a wormhole in a lab? No.” Peter Woit, a mathematician at Columbia University accused the authors of organizing a “publicity stunt.”

In a field deluged by claims of breakthroughs and electrified by the promise that esoteric advances in physics can eventually be converted to profit, the clamor about a putative “wormhole” on a quantum computer is not a shock. Rather, it’s a representative—albeit extreme—case of the trouble with talking about quantum computing.

As one of Raymond Carver’s discursive, slightly inebriated characters might have put it, “I was going to tell you about something. I mean, I was going to prove a point. You see, this happened a few months ago, but it’s still going on right now, and it ought to make us feel ashamed when we talk like we know what we’re talking about when we talk about [quantum computers].”

The trouble begins on a fundamental level: We do not know how things happen in the quantum realm. We can describe the results of measurements, but the ultimate nature of reality remains out of grasp and up for grabs because of this uncertainty.

An adherent of the many worlds interpretation, which asserts that every possible outcome of any quantum measurement unfolds across infinite parallel universes, might say that a calculation in a quantum computer involves the interference of wavefunctions throughout all those universes. Whereas a proponent of the traditional Copenhagen interpretation—which is agnostic about what can be known prior to measurement—might simply shrug. Because quantum mechanics is without a clear mechanism, our best descriptions of “quantumness” are by necessity riddled with equivocations and half-truths. 

Consider that even a colloquial term like “computer” presents difficulties. The most rudimentary classical computers inside “smart” refrigerators and coffee machines can store memory and carry out—without error—lengthy instructions via silicon-based transistors. But the analogous parts of a quantum computer, which has qubits instead of bits, quantum logic gates, and uses quantum algorithms, only offer superficial similarities that conceal vast differences in architecture and ability. 

Unlike their classical counterparts, quantum computers store information for a fraction of a second with ephemeral pairs of electrons, or ions suspended in magnetic traps, or even light bouncing between mirrors. Moreover, to protect the sensitive quantum states of qubits from computation-scuttling thermal interference and other environmental errors, most devices have to be chilled with liquid helium inside fridges a few degrees above absolute zero. At this stage in their development, quantum computers rely on tangles of wires and a bulky copper anatomy that makes them look like futuristic chandeliers rather than anything most people would recognize as a computer. 

Appearances aside, quantum computers do indeed compute, but not as your computer does. They do not send email or stream video or store documents. If anything, they are more like the bespoke apparatuses of the 19th century, such as Charles Babbage’s Difference Engine. A remarkable feat for its time, Babbage’s computer used thousands of hand-cranked mechanical gears to do one thing: calculate basic mathematical functions. A “quantum difference engine” is a more accurate description of what today’s quantum computers are capable of, but even that requires a caveat—Babbage’s machine would far outstrip them at addition or subtraction. 

Even without the intention of hype, it is extremely easy to perpetuate misconceptions about quantum computing. Researchers who command quantum technology thus almost inevitably are seen as techno-wizards summoning eldritch forces from some spooky realm. For many readers, invoking “quantum” is not so different from invoking magic; more D&D than R&D.

What to do? One possibility, proposed by IBM researcher Charles Bennett, is to reverse the typical thought pattern, and think about the classical world in quantum terms because the world is fundamentally quantum, and classical reality is just an approximation. Consider the inverse, Bennett asks. “A classical bit is a qubit which can only take the values of 0 or 1,” he explained during a talk at IBM. From this perspective, it’s not quite so eerie.

Quantum message discipline is sorely needed. Clarifying their headline change, Quanta noted that “The researchers behind the new work — some of the best-respected physicists in the world — frequently and consistently described their work to us as ‘creating a wormhole.’” 

When reached for comment, the study’s co-leader Maria Spiropulu pointed to a Caltech FAQ in her team’s defense. “Did we claim to have produced or observed a wormhole deformation of 3+1-dimensional spacetime?” the FAQ asks. The answer it then offers is a firm “No.” The FAQ further elaborates that what the researchers saw was only “consistent with the dynamics of a traversable wormhole.” The Nature paper is less absolute. There, Spiropulu and her co-authors describe their research as “the experimental realization of traversable wormhole dynamics on nine qubits” and discuss how they used a quantum teleportation protocol to “insert another qubit across the wormhole from right to left.” 

And in a video produced ahead of publication by Quanta, several researchers spoke gushingly, saying, “This is a wormhole” and “I think we have a wormhole, guys.” This is at best, deeply misleading.

What the researchers actually did was use nine qubits to compute a crude version of the Sachdev–Ye–Kitaev (SYK) model, which is used to calculate the dynamics of some quantum systems, like graphene. Simulating the full SYK model required too many qubits, so the researchers substituted a barebones, simplified version that would fit on just nine qubits. The SYK model is interesting to theorists because it is thought to be mathematically equivalent to descriptions of two-dimensional gravity in a universe with the opposite curvature of space than our own. This equivalence, theorists hope, may mean the SYK model has something to tell us about quantum gravity in our own universe. But the fact remains that the supposed bombshell dropped by the Nature paper is a calculation of an approximation of a model conjectured to be equivalent to a lower-dimensional gravity for a universe that isn’t ours

None of which is to say the study is without value. John Preskill, a widely-respected theorist at Caltech uninvolved with the work, tweeted that, “Because only 9 qubits are involved, this same computation can be done easily (and more accurately) with a conventional computer. Nevertheless, it’s interesting and rather unexpected that such a simple model can capture some novel features of quantum gravity.” 

“The physical world is quantum mechanical,” Richard Feynman wrote in 1981, “and therefore the proper problem is the simulation of quantum physics.” But exactly simulating an atom with just a few dozen electrons is impossible for classical computers because every additional electron exponentially increases the computing requirement. This restriction does not apply to a quantum computer because it follows the same laws as the quantum systems may simulate. The distinction between simulation, experiment, and computation becomes treacherously blurry, especially when one’s “computer” consists of a bunch of ions tasked with calculating the behavior of ions.

Last year, the mathematician Richard Borcherds responded to this quandary with a deceptively simple challenge for the most devout acolytes of quantum computing: Do better than his teapot. 

Borcherds’ snarky argument is that dropping a teapot and calculating how it shatters is ultimately a quantum problem because the teapot is made of quantum bits and bobs like protons and electrons. Teapot shattering is not abundantly useful, but no worse than the contrived problems various research teams have “solved” with their quantum computers to claim quantum advantage. Because today’s quantum computers have no more than a few hundred qubits, they’re unable to simulate the teapot—yet the teapot, by virtue of being itself, is extremely good at simulating the “quantum” problem of its shattering.

Borcherds’ “teapot challenge” bluntly asks a question that strikes at the heart of the wormhole drama and much of today’s quantum “breakthroughs”: What precisely are today’s quantum computers doing? The authors of the Nature study call it a “quantum experiment” because they “measured observables of the physical system”—that is, they measured the qubits of the quantum computer. 

Natalie Wolchover, a Pulitzer Prize-winning science writer at Quanta, argues that when a quantum computer simulates a toy model of quantum matter, such as the SYK model, it is really “creating” the quantum system it asks about. “It’s profound but somehow I can’t put my finger on what it means about the difference between ‘real’ and ‘simulated,’” she wrote in an email. (Full disclosure: Wolchover has been my editor at Quanta.) 

No consensus on that distinction yet exists within the quantum computing community, but last year a trio (two physicists and a philosopher, respectively)—Dominik Hangleiter, Jacques Carolan, and Karim Thébault—published a book suggesting a potential framework for thinking about the question. “The question is, what kinds of inferences can you draw?” says Hangleiter, a quantum researcher at the University of Maryland. “The term ‘simulation’ sort of implies that you have some other system in mind.” Under this framework, there are important differences to distinguish between two kinds of simulation: emulation and computation.

Emulations aim to simulate a physical system in nature, like a water droplet, while computations attempt to simulate theories that describe nature, like fluid dynamics equations. Both simulations can be “validated” or “unvalidated” depending on what we can learn from them.  Validated simulations tell us directly how the world actually is, whereas unvalidated simulations rely on analogies that merely tell us how the world might be. For instance, although bouncing oil droplets  mimic certain quantum phenomena, they cannot tell us how the quantum world actually works. But they can tell us how raindrops actually bounce..

Under this framework, Spiropulu and her coauthors’ use of nine qubits to study a scaled-down approximation of the SYK model is an “unvalidated computation”. In the Quanta video, the researchers claimed their research was “the first sign that you could see gravity on a quantum computer.” To explain what they saw in the qubits, they claimed the appearance of a wormhole. “It’s not clear to me how either of those claims is substantiated,” Hangleiter says.

Even if we lived in a different universe, in which the SYK model was equivalent to a description of gravity, this research still would not be evidence for the existence of wormholes. At best, it would be consistent with wormholes; how nature could be, not how it is. 

Ironically, a more prudent description of the Nature research might have been “Teleportation on a quantum processor.” Unlike wormholes, quantum teleportation is a well-established physical phenomenon which the researchers used in their experiment to ferry quantum information across qubits.

Suppose Alice (Remember Alice? It’s a story about Alice) is in San Francisco and she wants to send the state of a qubit to Bob in New York. So she asks Charlie in St. Louis to send them a pair of entangled qubits. Then, calling Bob on the telephone, she gives him instructions so he can properly measure his qubit. When he does this, the state of Alice’s qubit is teleported to Bob. 

“You know, Bob,” she begins over the phone. “We never did figure out how to tell my neighbor about qubit states. How in the world are we going to tell him we just performed a quantum teleportation protocol?” 

Bob presses the phone between his ear and shoulder as he prepares dinner. “Well, it’s not what he’s thinking. It’s just transporting information, and it needs a classical channel, so there’s no faster-than-light communication,” he muses over a bubbling pot of pasta. “How about we tell him there were ‘no wormholes necessary’?” 

One thought on “What we talk about when we talk about qubits

Comments are closed.