Google Claims Quantum Supremacy, IBM Says “Nope” — Unpacking What’s Important
(An artistic rendition of Google’s latest quantum chip from Google’s quantum AI lab blog)
The media has flooded the internet with Google’s claim of quantum supremacy. Google’s recent paper in Nature explains how its latest quantum processor, Sycamore, solved a random number generation problem in 200 seconds; the paper further claimed it would take the best supercomputer today 10,000 years to do the same. IBM immediately counterclaimed that Google didn’t properly run the experiment. High drama in information technology, indeed — Google and IBM duking it out over the future of computing!
The reality of this supremacy announcement is far more mundane and nuanced. I’ve spent a couple of years researching quantum computers and talking to all the major players in an attempt to sift through what is happening and what, if anything, our clients need to know about it. I’m going to set down in plain English here what I think is going on and what is most important. I won’t hash through the basics, as I’ve written a couple of reports and blogs you can read to get up to speed on that here:
- Why HPE Walked Away From Quantum Computing . . . And Where It Went Instead
- Who Should Care About Quantum Computing?
Further, all the quantum vendors have useful basic information on their websites. In this blog, I will explain quantum supremacy in detail, perhaps debunk a myth or two, and put the real achievement in a proper perspective.
Quantum Supremacy Does Not Mean Quantum Computers Will Replace Digital Ones
It is natural to assume that quantum supremacy is a universal proclamation that quantum computers are better than classical computers. One might also conclude that the end is near for our smartphones, encryption is no longer safe, and firms with quantum computers will soon start crushing their competitors with techno voodoo. While I’m saying these things tongue-in-cheek, many will naturally apply their understanding of digital or classical computers to quantum computers and draw conclusions that simply aren’t true. I tell clients that to understand quantum computing, you first need to forget what you know about today’s computers. They simply don’t work like that or do the same things.
Foremost, classical computers are universal, which means they can solve almost any type of problem we can express in terms of ones and zeros. Note I said “almost” any problem. There are certain classes of problems on the frontiers of computer science theory that we can’t practically solve with classical computers. Enter quantum computers that, in theory, could solve some of these. This means that quantum computers won’t replace today’s computers — they will work alongside them to eventually solve real-world, useful problems. For example, quantum computers will likely help model far more complex molecular states of materials in chemical processes. Big deal? You bet, because there are new drugs, new types of electric batteries, new products to be made, and so on waiting to be discovered if we had a working quantum computer — enter the supremacy debate.
Google’s Supremacy Claim Unpacked — An Important Milestone If It Stands
Supremacy is important because we need hard proof that quantum computers are better in some cases in order to justify further investment. We don’t have that proof right now, which kind of boggles my mind. Quantum supremacy for a specific type of problem is a piece of that proof. There are plenty of definitions of quantum supremacy online if you care to search, but they boil down to this:
Quantum supremacy means a quantum computer can do a computational task that today’s classical computers can’t practically perform.
In the recently published paper, “Quantum supremacy using a programmable superconducting processor,” Google claims to have done this. But what does this mean to the average enterprise? The truth is hidden in a bunch of technical details. Here is a breakdown in near normal-people-speak:
- Quantum supremacy can be claimed by using a real quantum computer to solve a made-up problem. The first thing to understand about Google’s work is that the problem it solved has no immediate real-world application. Google used a 54-qubit quantum computing chip (53 qubits working) called Sycamore to verify part of a random number experiment. As a matter of context, truly random numbers are useful in a large number of computing tasks; therefore, Google’s experiment was interesting. Google solved the problem of verifying that random numbers generated by a quantum computer are actually random. This is hard for the best classical computers to do as random number sequences grow, it turns out. Scientists say this type of problem is exponential. Google hailed its achievement as a Sputnik moment — Sputnik was not practical, yet it marked the beginning of the space race, but we saw that only in hindsight. Hindsight will tell if this is such a moment — for now, I’m keeping my hat on but clapping.
- The supremacy definition leaves the door open for debate. As I alluded to above, the phrase “practically perform” is key in the debate over Google’s claim. Google needed a really good classical supercomputer to show that its quantum computer is better. Ironically, it chose IBM’s latest supercomputer to run the test. Google claims that it would have taken IBM’s supercomputer 10,000 years to perform random number error validation computation that it did in 200 seconds. IBM countered by saying that Google didn’t properly optimize the supercomputer in running the test. If properly tuned, IBM asserted, it would take its own supercomputer a mere 2.5 days, not 10,000 years — a real, “You’re stepping into my house?” moment. Further, whether 2.5 days is a practical amount of time for this kind of computation (or is 200 seconds that much better?) is also up for debate.
Should You Care About Quantum Supremacy Or Quantum Value?
The above illustrates a problem with supremacy: It is theoretical and debatable, which means, to ordinary people, it’s not all that useful. It is just one proof point in a body of evidence we need to have confidence in the future of quantum computers. Furthermore, in considering your response to quantum computing, understand that:
- Supremacy for one problem type does not mean supremacy for others. Another problem with supremacy is that its claim only applies to one problem area, like random number generation. Just because we can prove supremacy there does not guarantee supremacy for other important areas like cryptography, for example. This means that if you are a supremacy tracker, you need to track it in every problem domain you care about. My advice is to watch supremacy claims and debates and use them as a part of your decision-making process.
- Quantum supremacy does not equal quantum value. The two terms quantum supremacy and quantum value illustrate the different approaches Google and IBM are taking. Google’s claim of achieving a quantum supremacy proof point is a nontrivial achievement. IBM is downplaying Google’s supremacy efforts in favor of the term quantum value. Value means that a quantum computer offers some benefit over a classical computer for solving some practical problem. It is a pragmatic definition. Achieving quantum value for a problem will be very important to buyers, but there is no agreed-to method for scientifically proving quantum value. It is a subjective measure. Enterprises will have to decide for themselves if quantum value exists as part of their investment decisions.
- Supremacy does not set a timeline for value. This is perhaps the most important point to understand about quantum supremacy and value. Except for Google’s contested claim of supremacy, no actual quantum computer can do anything that a classical computer can’t do better or almost as well. In other words, we do not know yet that all this hoopla about quantum computers will actually create a return on investment.
- We expect that Google’s supremacy experiment will be debated by the research community for some time. Others will try to replicate it, researchers will conduct new experiments, and, eventually, a body of proof may emerge that quiets debate. This is the scientific process, and it is going to take time.
- Just like supremacy, no quantum computer today has demonstrated value. Will we reach value before we stop debating supremacy? I don’t know. As far as timing, IBM thinks we might get to value in 10 years or so. We will likely see a number of supremacy claims and debates along the way. The truth is that this is all just educated guessing. Quantum computers are intriguing because all this could happen in so much faster than 10 years or things might take 20 years or more.
Toward Quantum Value — Are We “NISQing” Yet?
Google claims that its achievement heralds the beginning of the NISQ era. Unless you are deep into following quantum computers, this won’t mean much. It is important, however, so let’s unpack it. Here is what you need to know:
- Living up to the hype of quantum computing will require extremely powerful, “error-corrected” machines. For example, breaking today’s PKI encryption will require quantum computers with millions of qubits; today, Google has a 54-qubit machine. IBM has just announced one, too. If qubits double every year, it will take about 15 years to get into the right range. In addition to doubling qubits, a bunch of engineering work needs to be done, coupling these qubits together and suppressing the errors that tend to make qubits go bonkers and need resetting. I recommend that you think of error-corrected, universal quantum computers as the stuff of science fiction. It’s a prediction of a future that might happen, but in truth, the future could be quite different. We simply don’t know.
- NISQ means noisy intermediate-scale quantum computing that delivers value in the next five to 10 years. NISQ was coined by John Preskill to reflect a period of time before we have the error-corrected quantum computers. In the NISQ era, we will hopefully be able to use intermediate-scale, noisy quantum computers to deliver small bits of quantum value. We won’t know we are in the NISQ era, however, until somebody delivers value with a quantum computer. In my opinion, Google’s heralding of the NISQ era is premature. When a client tells me they have solved a real-world problem with a quantum computer, I’ll post it here on the Forrester blog that we have hit NISQ.
- We will most likely achieve quantum value first in the field of sampling. Sampling is an extremely important but highly technical and hard-to-explain technique that relies on random numbers to explore the behaviors of complex systems, among other things. It is an important research technique with applications in modeling and simulation, physical sciences, cryptography, machine learning — the list goes on. It works by using a random number generator to insert a large number of random inputs into a system. Researchers study the system outputs to draw conclusions about the system’s behavior. Classical computers can only generate quasi-random numbers that eventually repeat. Quantum computers can potentially generate much larger random number sequences, which can improve some experiments. Verifying that a large random number sequence is actually random is the problem Google claims to have solved. Google’s experiment means that we might actually get to quantum value in sampling first. We’ll see.
Summing It All Up: File Google’s Announcement Away And Keep Learning
Given all this info, what should your firm do about it? For most people reading our blog, the answer is “nothing,” except to try to understand what it means and watch out for other supremacy proof points and possible quantum value emergence. Here’s a few insights for the future:
- Quantum computing must be watched, because things could change very quickly. As I hope you guessed, quantum computing is at the frontier of research. We simply do not know what the future holds. This means that it is anybody’s guess how fast things will proceed. The quantum computer’s ability to increase in theoretical power as qubit count goes up is beyond our human mind’s ability to imagine. That said, there are tough engineering problems that firms such as IBM and Google, among many others, must solve. Firms that understand when to invest, what to invest in, and develop the talent to take advantage of this technology could shoot out ahead of the rest of the market like rockets — or they could waste valuable time and money.
- We believe that the accelerating pace of technology evolution will keep the rich rich. No longer do small, nimble disruptors that understand agile, cloud, and big data have the advantage. Everyone understands these and is working to be much faster, more agile, and more connected. Our research shows that firms with the power to be adaptive and drive innovation with technology such as quantum computers are growing at 4x industry average growth rates. They have the money and freedom from investors to keep accelerating innovation. It’s a classic cycle: In the foreseeable future, rich, tech-savvy companies will keep getting richer and more tech-savvy. This is why Google can throw so much money at quantum computing. And why technology laggards are right to be worried.
- Your big dilemma with quantum computing is finding the wherewithal to jump in and experiment now. I am convinced after several years of research that quantum computing will be as significant as the invention of the printing press. It’s going to take longer than many expect but will still happen faster than most firms can prepare for. Knowing this, you have to make smart decisions so that your firm is ready before quantum value is achieved in a business area that matters. That is a difficult challenge given all the current problems most firms need to pay attention to.
Want to talk more quantum computing? Have a good story to tell? Do you disagree with anything I’ve said? I’d love to hear from you.