Phasecraft is speedrunning...
...and that's why DARPA is interested.
“Speedrunning” is when you complete a video game, or part of it, as quickly as you possibly can. You map the most efficient path, skip around the normal order of events, and maybe exploit some bugs. I’m not going to pretend to know all the tips and tricks, otherwise I’d probably be too busy playing to write this. But the point is that speedrunning goes well beyond winning the game, and is about finding the fastest, most efficient path to achieve the same outcome.
Phasecraft speedruns algorithms.
DARPA established the Quantum Benchmarking (QB) program to determine which applications are NOT potential use cases for quantum computers. We can’t prove what will be useful, but we can prove what won’t be useful. Out of the application classes that have NOT been disproven, the question arose whether a quantum computer can be built that can actually execute them.
This led to the Quantum Benchmarking Initiative (QBI) and, yes, I’m doing a lot of compression here. Anyway, there are two ways to make utility-scale quantum computing more plausible: improve the hardware, of course, but also optimize the algorithms.
The latter is where Phasecraft comes in. The textbook algorithms are not optimized for execution, plus the costs can vary significantly across platforms, so Phasecraft’s mission is to provide the best possible algorithms with the resource estimates for their execution. When further efficiencies can no longer be found, the speedrun is complete, that’s the new record, and those are the resource estimates that DARPA will have to work with. Each QBI-participating team, together with DARPA and Phasecraft, will have to agree that an algorithm will actually execute, thus giving the utility-scale system actual utility.
To avoid conflicts-of-interest, Phasecraft is not working with the teams directly. To avoid favoritism among the diverse modalities, a number of different metrics are employed. Efficiency versus speed is considered, as is error correction overhead. Altogether, Phasecraft’s algorithms will help put points on a graph that will help DARPA better estimate when value will be able to be delivered. It is reasonable to assume that different applications might favor different modalities.
The algorithms will not be published immediately, but Phasecraft suspects the algorithms will be published at some point.
Why Phasecraft?
Phasecraft thinks it’s the only commercial organization involved in this algorithmic speedrunning and I can’t dispute that. The company credits its track record of reducing the costs of algorithms, especially in chemistry and materials modeling and in combinatorial optimization problems. Some approaches involved plugging quantum into classical frameworks to overcome their limitations. The company also designs new algorithms, but the underlying themes in both scenarios are extreme efficiency and resource cost reduction.
Conclusion
Phasecraft hopes that QBI becomes a springboard to more government work in the US, UK, and elsewhere, plus commercially. The company is looking for genuine quantum advantages, which do not necessarily require rigorous theoretical proofs. It is looking for practical advantages, including through heuristics, even when such advantages can’t be proven mathematically.
You’re not supposed to be able to complete Super Mario Brothers in less than five minutes, for example, but apparently you can. You use all the tips and tricks in your repertoire and you prove that it’s possible nevertheless.
Epilogue
Did you know that more than 500 people have completed 1985’s Super Mario Brothers in less than 5 minutes?
Image generated by Google’s language model AI.