People have been debating whether a quantum computer presents a realistic threat to the Bitcoin network for over a decade. It was a serious topic of conversation over 13 years ago when I first discovered Bitcoin myself. 

There has been quite a lot of progress, both in terms of theory and real-world engineering, since long ago when I was just a bumbling idiot trying to figure out what was going on here. 

Two major milestones have been reached since then that make a material difference in the likelihood of a viable quantum computer actually being produced sometime in the next decade or so. That doesn’t inherently mean that it will reach a point of ubiquity, or even relative ease of access for those with large amounts of capital. 

But it is very possible that a number of viable machines will be produced in the near future. 

Error Correction Improvements

The first major improvement has been in error correction. To account for the inherent noise in working with things at this kind of tiny scale, to get a logical qubit that is useful in computation in practice requires the use of multiple redundant physical qubits. 

The prior state of the art way of doing this was surface codes, a way of bundling multiple physical qubits together in a grid and using some of them as check qubits that periodically “check on” their neighbors to ensure no internal errors in the superposition have occurred (without collapsing the superposition). Each grid’s empty spaces need to be filled with check qubits. 

This check qubit requirement creates an extra overhead that can get close to 1,000 physical qubits per logical qubit in total, and it gets bad at scale because check qubits can only check on the qubits immediately next to them. So every grouping of qubits needs to have checkers in equidistant spacing. 

Quantum low-density parity-check (qLDPC) codes remove this bottleneck, allowing check qubits to check other qubits at large distances (either through traces interwoven to communicate across chip sections, or by physically moving atoms like with the neutral atom design) across the device. This has allowed a 10x reduction in the amount of physical qubits necessary to produce a reliable logical qubit. 

That is not something to sneeze at. While it might not be a fully functional machine making progress at gaining more efficiency, it is material efficiency gains in the engineering processes that underlie the production of a fully functional quantum computer. 

Progress In Proving Fundamentals

The second has to do with a more fundamental question around the assertion that adding more physical qubits leads to a reduction in overall noise in the system rather than an increase. This is really at this point still theory, and you have to keep in mind that to this day there has never been a fully functional quantum computer that has end-to-end performed a computation a classical computer is incapable of. 

Google performed an experiment using their Sycamore (and later Willow) chips to experimentally verify the effect of adding more physical qubits. To be very clear, this was not a demonstration of performing computations, but simply a demonstration of storing information in memory without it decaying. 

They demonstrated through the use of logical qubits composed of a bundle of 17 physical qubits, a bundle of 49 physical qubits, and a bundle of 101 physical qubits that the logical error rate, the frequency of data corruption, decreased as the physical qubit count went up. This test passed a critical threshold, where the logical qubit being created out of the independent physical qubits maintained coherence longer than any individual physical qubit it was composed of. 

Now again, this is not a jump to a fully functional quantum computer performing computations that classical machines are incapable of, but it is material progress proving one of the fundamental assumptions underlying quantum computers. 

AI

These aren’t the only things that we are finding better solutions to in this problem space either. Artificial intelligence has become a big component in these systems. It is being used in the actual process of reading and decoding information from a quantum computer, a big bottleneck for actually making use of it at scale. 

AI is also being used in the development of new quantum algorithms optimized for these types of machines, and given the recent spate of AI helping to solve (or even disprove existing conjectures) major problems in the field of mathematics, this isn’t really that crazy of a leap to consider the possibility of major breakthroughs brought about by AI. 

They are being put to the same use in actually designing the actual physical quantum circuits that are built using different architectures. This is a very complex problem, actually, finding the optimal way to lay out quantum gates in a physical space to minimize noise at the quantum level, without creating so much empty space that you introduce latency, inefficiency, and other problems to solve. 

This is a factor that very well could hypercharge progress at solving the necessary fundamental problems. 

Outlook Ahead

Ultimately, in my opinion, this comes down to one question: does the assumption that adding more physical qubits reduce noise actually hold when it comes to computation and the active manipulation of quantum information. 

If that assumption does hold, and isn’t experimentally disproven sometime in the near future, then I think there is a very realistic case for a viable quantum computer being produced in the next ten years. 

There is a massive amount of resources being thrown at this problem, significant (but not overwhelming) progress at solving pieces of the problem, and if there fundamentally is a way to do something, human beings usually figure it out. 

I’m not saying that it’s time to panic, but don’t discount the possibility. 

This piece is featured in the latest Print edition of Bitcoin Magazine, The Quantum Issue. We’re sharing it here as an early look at the ideas explored throughout the full issue.