[dj43nq]

The pizza in the article looks horrible. But random.

I don’t see why the author states randomness is a hard problem then proceeds to show various ways of creating it via imagery. Lava lamps also work. https://en.m.wikipedia.org/wiki/Lavarand

[wahern]

Because they never explain or explore the particular qualitative definition of randomness that is difficult: unpredictability. Computers do indeed suck at this, which is why you need to attach specialized devices to feed it unpredictable bits.

Random is an overloaded term. You can algorithmically generate statistically random distributions which are nevertheless easy to predict given some previous state. That form of randomness is useful for many things, but lethal in cryptographic systems. Moreover, proving that some distribution is unpredictable--and that this unpredictability is preserved within the larger system--is really difficult.

To be even more technical, there are different ways to categorize unpredictability: quantum, thermodynamic, mechanical, computational, etc. The former relate more to how you physically generate unpredictable states from physical phenomena while the latter relate more to qualitative degrees of unpredictability. A thermodynamic generator may only be mechanically unpredictable (i.e. unpredictable because it's impractical or perhaps impossible to capture the necessary state needed to predict the next sequence), or it could be unpredictable in the quantum sense (there exists no state that you could ever capture sufficient to usefully predict the next element in the sequence), depending on how you model the fundamental physics. These distinctions matter when you want to analyze the consequences of, e.g., using a real-time video feed of a lavalamp to feed a CSPRNG.