The claim is modest: we know what you will like next. The claim is enormous: we have modelled your taste so precisely that the margin of error is negligible. Both statements are in the playlist description. Neither is examined.
What concerns me is not the recommendation itself — most are harmless — but the tone. The confidence. The absence of any acknowledgement that the thing recommended replaces something not recommended, and that the cost of the replacement is never shown.
Refusal is the first act of taste. Every algorithm understands engagement. None of them understand refusal. This is not a limitation of the algorithm. It is the algorithm feature, presented as neutral.