Now, this is a 'KQRBNN vs kqp' three-move chess problem generated by the program, Chesthetica, using the Digital Synaptic Neural Substrate (DSNS) computational creativity approach. The DSNS does not use endgame tablebases, neural networks or any kind of machine learning found in traditional artificial intelligence (AI). It also has nothing to do with deep learning. Chesthetica is able to generate mates in 3, mates in 4, mates in 5, study-like constructs and also compose problems using specific combinations of pieces fed into it (e.g. composing something original using only three pawns vs. a knight). Read more about it on ChessBase. The largest endgame tablebase in existence today is for 7 pieces (Lomonosov) which contains over 500 trillion positions, most of which have not been seen by human eyes. This problem with 9 pieces goes even beyond that and was therefore composed without any such help.
