|Quantum Computers Too Slow To Crack Bitcoin?|
|Written by Mike James|
|Wednesday, 26 January 2022|
The huge amount of effort going in to building a quantum computer is based on the idea that it can compute useful things in a usefully short time. Now an analysis by researchers at the University of Sussex suggests that quantum advantage may not be enough - we may need really big hardware.
The ideal quantum computer isn't the same thing as a real world quantum computer. We cannot build a perfect machine and as a result we need to implement error correction to get useful results. The need for error correction slows things down and increases the number of qubits needed. As with all computing, there is a time/space trade off and using more qubits can reduce the time taken for the computation. What the latest results show is that the amount of hardware needed to create is way greater than we might have imagined to deliver reasonable times.
The first result presented in a paper authored by Mark Webber, Vincent Elfving, Sebastian Weidt and Winfried K. Hensinger of the University of Sussex is in the more important, but less glamorous, area of quantum chemistry. All of chemistry is determined by quantum mechanics - the only problem is that we cannot do the sums fast enough on a classical computer. A quantum computer, on the other hand, has always been thought to be the ideal way of getting to the solutions faster. Assuming reasonable figures for error correction rates and cycle times, the latest results suggest that to get a computation time of 10 days you need 750000 qubits at best. This is way larger than the tens of qubits currently achievable.
The second result is more headline grabbing. Bitcoin uses SHA-256 for its encryption. This is an elliptic curve algorithm that was created by the NSA and it is widely used and cracking it would be an achievement. To make the cracking worthwhile for bitcoin it would have to be done in around 10 minutes when the keys are vulnerable. The new research suggests that to meet this time requirement you would need 1.9 billion qubits. If you relaxed the timing to one hour, you would still need 317 million qubits and even allowing an entire day reaching the 13 million qubits needed is still the stuff of science fiction.
I have long held the view that most of the enthusiasm for quantum computers is generated by the idea that they might make cracking difficult encryptions possible in a practical time scale. All this talk of using them to further science and technology, and chemistry in particular, is partly a way of making the enterprise seem more worthwhile. If this is correct the idea that you need 2 billion qubits to crack SHA-256 in ten minutes is a huge blow. It doesn't matter much if quantum computers prove that they are faster than any classical computer on some toy problem if the real problems need silly amounts of qubits to be effective - and it has to be said that the number of qubits that this research suggests are necessary is very silly indeed.
Of course, there are lots of assumptions in the paper and hence there is lots of scope for disagreement. I'm sure that there are going to be lots of people modifying the assumptions or proving that they are unrealistic or too restrictive, but for the first time we have been provided with a look at the real world effectiveness of real world quantum computation, and it doesn't seem promising.
Artist's rendition of Google's Sycamore processor mounted in the cryostat. (Forest Stearns, Google AI Quantum Artist in Residence)
Mark Webber, Vincent Elfving, Sebastian Weidt and Winfried K. Hensinger, University of Sussex.
or email your comment to: email@example.com
|Last Updated ( Wednesday, 26 January 2022 )|