I'd be more impressed if they had trained it on Kenny G and Pat Metheny liked the results.

I emailed the author about a week ago about using this work to (re)harmonise melodies, but as others have noted, the network with one training piece doesn't generalise.

I'd be surprised if a current gen LSTM will be able to generalise music or language rules well enough to be able to piece together music or sentences, long enough for a coherent story or a jazz piece that matches that of a competent human author.