Kreuzberg is a borough in Berlin with 150k inhabitants, and it used to be divided into two postcodes: 36 and 61.
36 is the cooler, poorer, more political, younger half of the borough, it really was a center for non-conformists after reunification, but it's being gentrified these days.
'I’ve tried to come up with a rhyme that works in English: “36’s on fire, 61’s just tired” was one solution, but it doesn’t carry the matter-of-factness of the German.'
For a AI paradigm that gives the neural network no explicit information about phonetics, the ability of GPT to produce rhyming phrases never ceases to amaze me - especially in a language which is as infamous for its inconsistent orthography as English. Has there been any research attempting to work out how such subtle patterns are learnt by the AI?
36 is the cooler, poorer, more political, younger half of the borough, it really was a center for non-conformists after reunification, but it's being gentrified these days.