Music: how does it work? 
Generative algos are a great tool to answer this question. But merely mimicking music on a statistical level teaches us nothing useful.
-
-
This is phenomenalogical, but you might enjoy http://www.apa.org/pubs/journals/releases/aca-aca0000090.pdf …
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
same could be said of some neural NLP/LM stuff, right?
-
yes. Treating language as a pattern matching problem is at best a way to achieve prestidigitation tricks (e.g. captioning, QA).
- Show replies
New conversation -
-
-
Someone had NIPS had a fun take on this. Argued that gen models give you "a sampling from all the world's music".
-
But to have a coherent composition you have to choose other sources of variance to kill. And our models aren't good at that yet.
End of conversation
New conversation -
-
-
I used to think Novelty was the answer. However, most people like to look at, and listen to things they are already familiar with.
-
isn't that just improvisational jazz?
End of conversation
New conversation -
-
-
ML has never really been a tool for gaining human understanding. Flip-side of that is you can get a lot done without understanding
-
I think once we start making AIs specifically for improving human understanding, they'll perform systematically worse than SoTA.
End of conversation
New conversation -
-
-
I was looking at this myself. Seen something with jazz and Rnn. Seen wavenet. Nothing cool with gans. Am I missing something?
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.