This is related to a theorem that I have proved and its relation (or not) to an existing result. Essentially, I have shown that PAC-learning is undecidable in the Turing sense. The arxiv link to the paper is https://arxiv.org/abs/1808.06324 I am told that this is provable as a corollary of existing results. I was hinted that the fundamental theorem of statistical machine learning that relates the VC dimension and PAC-learning could be used to prove the undecidability of PAC-learning.