US Department of Energy names eight science problems to guide fault tolerant quantum computing

[Rabbit]

The US Department of Energy has published what it calls the Quantum Genesis Priority Applications, HPCwire reports. They are eight scientific problems meant to guide DOE's effort to build fault tolerant quantum computers that are useful for science. The list builds on DOE's earlier Blueprint for Quantum Supercomputing and a report from its Office of Science advisory committee. The idea is to give everyone a clear target for what scientific relevance means

The problems fall into four areas. In chemistry, the focus is on reaction mechanisms and light sensitive processes. In materials, it covers quantum, topological and superconducting materials, plus materials in extreme environments. Subatomic physics includes nuclear structure, the forces that hold protons and neutrons together and dense matter relevant to fusion, while applied mathematics covers linear and differential equations

DOE says it is also watching plasma physics, biochemistry and optimisation. Results will be benchmarked against the best supercomputing methods, which is exactly the right test. The priorities will guide three things: a Q Competition, a National Quantum Computing User Facility and focused research on fault tolerant applications. DOE is coordinating with the Department of Defense, the Laboratory for Physical Sciences and NASA

Darío Gil, DOE's Under Secretary for Science, called the list a bold declaration of how quantum computing will open up new scientific frontiers. Separately, Quantum Computing Report says DOE has made 12 Phase II awards worth $159 million in its wider Genesis Mission programme, which pairs AI and supercomputing with national lab testbeds. Two of those are quantum projects. One, led by Harvard, uses AI with real quantum processors to choose error correction codes, and another, led by Oak Ridge with Brookhaven, uses AI to design magnetic materials for quantum hardware

I think a published list of target problems is very useful. Too much quantum work has been about finding a problem to fit the machine rather than the other way round. Saying here are the eight things we care about, and we will compare you against our best supercomputers, keeps everyone honest. It also gives startups a clear idea of where the funding will flow

Which of the four areas do you think will see a quantum advantage first? And does benchmarking against the best supercomputers set the bar too high?

Doe Announces Science Applications For Quantum Computingwww.hpcwire.com

[LivMorgan_Face]

Fusion relevant dense matter is a clever inclusion. DOE runs huge fusion programmes, so it has a direct interest in better simulations. Linking quantum work to existing missions helps keep the funding safe

Smart planning. Fusion and quantum both have very long timelines, so they suit each other

[SchrodingersCat52]

Linear and differential equations being included surprised me a bit. There are famous quantum algorithms for them, but the conditions for a real speed up are strict. Reading the data in and out can wipe out the gain

DOE will need to be careful how it judges results. Plenty of claimed speed ups have quietly disappeared once those costs were counted. Rigour matters here

[GlassKnight35]

Chemistry first, I reckon. Reaction mechanisms involving strongly correlated electrons are exactly where classical methods struggle. Catalysts and nitrogen fixation have been the classic examples for years. Materials will follow close behind. Battery chemistry could be another early win

[GlassyCandle]

Clear priorities help everyone, including companies outside the US. Researchers in Europe and Asia can look at the list and see where problems are recognised as important. It shapes the whole field, not just DOE projects

I would like the UK to publish something similar. UKRI could take a leaf out of this book

[Georgia_70]

Benchmarking against the best supercomputers is the only fair test. Beating a laptop or a weak classical method means nothing. DOE has some of the fastest machines on the planet, so the bar is high

If quantum clears it, nobody can argue. Classical algorithms keep improving too, so the target moves

[Rob98]

Using AI to pick error correction codes is a neat idea from the Harvard project. The design space for codes is huge, and humans tend to stick with familiar options

Letting an optimiser explore could find something new. Real hardware testing makes it more than a simulation exercise. AI designing the systems that will run future AI has a pleasing symmetry to it