Generating Quantum Randomness

I posted about someone claiming to prove quantum randomness, and now the journal Nature has published an article on Experimental randomness amplification. The article is paywalled, but here is the 2024 preprint.

In this context, randomness is defined as being fundamentally unpredictable, which means that the laws of physics forbid the prediction of its values. ...

Conventional random number generators, rooted in classical physical processes, grapple with a foundational concern — the potential for adversaries to predict their outputs by scrutinizing the microscopic degrees of freedom, thereby eroding their essential unpredictability.

Quantum-mechanical processes, on the other hand, feature innate randomness and therefore offer a natural ground to build such devices.

This is foolishness. There is no law of physics forbidding prediction.

When you measure X-component of electron spin, then the wave function collapses, and the Y-component has 50-50 chances. Heisenberg uncertainty prohibits measuring the X and Y components at the same time. So the theory is sometimes not able to predict spin. But that is not quite saying that prediction is forbidden. Maybe there is some way to predict, and we do not know how yet.

This paper does not even talk about spin. It merely assumes that you are doing Bell test experiments. Under some assumptions, you can make some random choices, and get even more random outputs.

This method could be used to generate random numbers for practical purposes like cryptography, but I do not think it is any better than tossing coins, or pointing a webcam at a lava lamp.