Deep the Beatles! is, without a doubt, the least accessible of the currently released DADABOTS albums, but it's nonetheless interesting.
Unlike other DADABOTS albums, which seem to prefer metal as source material (possibly to make a point that their neural network can do any kind of music, even the most extreme ones), this one uses a much more typical source for deep learning algorithms: The Beatles
While music generated from The Beatles would seem more accessible than, say, music generated from Krallice or The Dillinger Escape Plan, this release is not at all structured as a typical music release. This is because, instead of being an attempt to write a Beatles album, it's actually a demonstration of the "songwriting" process the neural network goes through, as evidenced by the song titles, which show just how tedious it is to make this.
The first songs on this album sound absolutely nothing like the Beatles. I honestly thought that the AI was trained exclusively on Revolution 9 based on how abstract the early tracks were. However, the further you listen, the more the Fab Four begin to emerge, albeit sounding like they just dropped the world's entire supply of acid spoiler By the end, it kinda sounds like someone is playing Beatles music very loudly in the other room, and what you hear is the sound when it's only able to travel through the space between the door and the ground beneath it.
Yeah, that doesn't sound that appealing, but trust me, this a very interesting experience.