> Willow performed a standard benchmark computation in under five minutes that would take one of today’s fastest supercomputers 10 septillion years — a number that vastly exceeds the age of the Universe.
What computation would that be?
Also, what is the relationship, if any, between quantum computing and AI? Are these technologies complementary?
Is it really fair to call that "computation"? I am definitely not an expert, but it seems they are just doing a meaningless operation which happens to be trivial on a quantum computer but near-impossible to simulate on a classical computer.
To me that sounds a bit like saying my "sand computer" (hourglass) is way faster than a classical computer, because it'd take a classical computer trillions of years to exactly simulate the final position of every individual grain of sand.
Sure, it proves that your quantum computer is actually a genuine quantum computer, but it's not going to be topping the LINPACK charts or factoring large semiprimes any time soon, is it?
It seems like they went from lighting a candle to firing off some fireworks. Impressive, but there's still a long way to go until they can land a human on the moon - let alone travel to Alpha Centauri like some people are claiming.
Yes, this is exactly what it is doing [1]. The area of research is Noisy Intermediate Scale Quantum, and has arisen specifically to prove that quantum supremacy is possible in practice. It is currently the focus of pretty much all quantum computing research because attempts to produce a generalized quantum computer have all failed miserably. Existing practical quantum computers (like DWave) perform various annealing tasks but have basically proven to be inferior to probablistic algorithms computing the same task.
To date all attempt to produce valid claims of quantum supremacy via this channel have failed on closer inspection, and there is no reason to assume otherwise in this case until researchers have had time to look at the paper. There's a number of skeptics in the quantum computing field that believe that this is simply not possible.
It's different from your hourglass in that the computer is controllable. Each sampled random circuit requires choosing all of the operations that the computer will perform. You have no control over what operation the hourglass does.
It won't be factoring large numbers yet because that computation requires the ability to perform millions of operations on thousands of qubits without any errors. You need very good error correction to do that, but luckily that's the other thing they demonstrated. Only when they do error correction, they are basically combining their system down into one effective qubit. They'll need to scale by several orders of magnitude to have hundreds of error corrected qubits to do factoring.
> Also, what is the relationship, if any, between quantum computing and AI? Are these technologies complementary?
Ongoing research.
The main idea of quantum machine learning is that qubits make an exponentially high-dimensional space with linear resources, so can store and compute a lot of data easily.
However, getting the data in and results out of the quantum computer is tricky, and if you need many iterations in your optimization, that may destroy any advantage you have from using quantum computers.
"Also, what is the relationship, if any, between quantum computing and AI? Are these technologies complementary?"
AI is limited in part by the computation available at training and runtime. If your computer is 10^X times faster, then your model is also "better". Thats why we have giant warehouses full of H100 chips pulling down a few megawatts from the grid right now. Quantum computing could theoretically allow your phone to do that.
Actually about 20 W — if you ignore the 80 W used by the rest of the body (which seems debatable). And clearly far more than this was required to 'train' the human brain to the level of intelligence we have today.[1] But this still probably doesn't take away from your point. The human brain seems to be many orders of magnitude more efficient than our most advanced AI technology.
Though the more I think about this, the more I wonder how they really would compare if you made a strictly apples-to-apples comparison.
What computation would that be?
Also, what is the relationship, if any, between quantum computing and AI? Are these technologies complementary?