“Quantum supremacy is important both in its own right and as a benchmark or step toward something further.
But my theory is that quantum supremacy cannot be achieved, and this is based on analysis of the stochastic behavior of samples coming from quantum computers in the intermediate scale.
It is a bit complicated, but if you take a quantum computer like the ones people build now all over the world, what they produce is a combination of something very sensitive to noise, which is useless for computation, and something stable to noise, which is a very primitive computational device. If you have this combination, you cannot achieve quantum supremacy. This is my theory.”
— Gil Kalai,
This is an insightful summary by science journalist Anna Kramer of our recent zoom conversation. Anna interviewed me alongside four other experts – Scott Aaronson, Sergey Frolov, Joseph Emerson, and Shivaji Sondhi for our view on quantum computational supremacy. See this article on “Aventine”. It is one in a series of articles titled “Five Ways to Think About….”
The five views
All of us agreed that achieving quantum computational supremacy is a notable yet elusive goal worth pursuing. Some of us pointed out that despite its grandiose name, it represents a fairly early threshold in the path to quantum computing. None of us offered the view that quantum supremacy has already been achieved (although Scott said that “Quantum supremacy can be achieved and then unachieved later”).
Scott expressed the view that we will eventually reach a place of a firm and decisive quantum supremacy (where it will be achieved on a routine basis). I expressed the opposite view that quantum supremacy cannot be achieved and offered a brief explanation in support of my view. The other three experts did not make firm predictions for the future.
Sergey made a general point about science that “we are given a few years per idea, and then if that doesn’t work, either we fade away or need to come up with a completely different idea for what we might want to do.” This is very different from my experience as a mathematician.
Adding some balance and diversity
Personally, I would have been happy to see female experts among the participants contributing their opinion. To do some justice both to the many eminent women scientists in quantum information theory and to the prevalent optimistic view about the future of this endeavor, I would like to refer the readers to Dorit Aharonov’s brilliant lecture described in this post where, among other very interesting topics, she shares her optimistic perspective on the future of quantum computing.
Sergey Frolov and I
Of the four other experts I know Scott and Sergey personally. Sergey and I have crossed paths in the context of the replication crisis (aka the reproducibility crisis; it is the fourth item in this post), which is relevant to several experiments in quantum computation (and all over science). In our last meeting in Tel Aviv, Sergey kindly gave me a T-shirt from a recent conference on Reproducibility in Condensed Matter Physics that he organized.
Some more details of my view
Here is a somewhat more detailed (but still short) description of my view that I prepared for the interview.
Quantum supremacy refers to demonstrating on a quantum computer something which is extremely hard or even impossible to produce with a classical computer. Achieving quantum supremacy would be a notable landmark on its own and it could also serve well as a useful benchmark (among others) for progress towards the creation of good quality quantum error correcting codes that are needed for large scale quantum computation. In reality, the claims about classical hardness are much trickier than we thought and the claims by Google and other groups from 2019 and 2020 about quantum supremacy were largely refuted. Specific assertions and predictions regarding past and future progress offered by the Google “supremacy paper” also turned out to be false.
The key for both quantum supremacy and quantum error correction are substantially better “quantum gates”. My theory asserts that this is physically impossible and consequently both a convincing demonstration of quantum supremacy and of good quality quantum error correcting codes are not possible. My argument is based on analyzing the stochastic and computational behavior of quantum computers at the intermediate scale. For a large range of error rates the computers will express a combination of computational primitive robust outcomes along noise sensitive components of no use to any computation. This theory will be tested in the next decade and we can expect conflicting and confusing evidence in this period.
The temptation to immaturely present fantastic experimental breakthroughs is large, and, in my view, specific experimental claims should be carefully scrutinized. With Rinott and Shoham I carefully scrutinized the Google 2019 experiment and in view of our study I tend to regard this and other related experiments from 2019/2020 as “mock-up demos” rather than as complete scientific demonstrations. Putting quantum supremacy aside, the question of what can be achieved with superconducting quantum computers in the 10-30 qubit range is still unsettled and specifically, the gap between Google’s and IBM’s experimental claims is not understood.
There are recent interesting experimental claims also for ion trapped quantum computers, neutral atoms quantum computers, and other technologies. These claims should also be carefully scrutinized and in the meanwhile be taken with a grain of salt.
Update: There was a nice discussion over Facebook. Gali Weinstein criticized the term “mock-up” in this context and wrote: “The definition of a scientific experiment is broad, encompassing a wide range of activities, from preliminary investigations to highly controlled and reproducible studies. Even experiments with flaws, limitations, and the need for further refinement still fall within the domain of legitimate science as long as they are conducted with the intent to explore, test, and validate hypotheses using systematic methods.”