Not AI per se, though still interesting. https://www.iflscience.com/with-qua...an-finally-generate-real-random-numbers-79601
In my experience any finite sample from an RNG will fail one or probably more than one of the Marsaglia/Die-Hard randomness tests. This includes supposedly true random bit sequences from Random.org. John von Neumann famously said that believers in these deterministic generators were "living in sin." That said, the authors have made their random output available and I have downloaded a few large samples. I will run my usual tests on them. FWIW, though the Nature article is not on Sci-hub yet, a pre-publication version can be found on arxiv: https://arxiv.org/pdf/2411.05247
http://www.cacert.at/cgi-bin/rngresults has many pseudo-random number generators that pass all their tests.
Post-human in 5... I know, I know. I'm looking forward to it. Until then : https://linux.die.net/man/4/urandom