3 ms·
Though I am not a lawyer, I will jump to a conclusion that this strategy will fail. First, algorithmically generating all length 12 sequences of 8 notes (2^36
by dieselerator 7y ago
Though I am not a lawyer, I will jump to a conclusion that
this strategy will fail.
First, algorithmically generating all length 12 sequences
of 8 notes (2^36 = 64 G results) is just math, not music
composition, even if you encode the results in midi format.
Thus, not subject to copyright protection (in the U.S. anyway).
Next, a catalog of 64G entries is not human searchable.
It is an unreasonable to state a person searched or could
have the catalog, and copied an entry. Perhaps a program
containing certain search criteria could scan the catalog
looking for entries with "musical" attributes. However,
such a program would not need the catalog, because it
could generate every entry independently.
Since the output of this effort is neither subject to copyright
nor human searchable, the effort will not mitigate copyright
law suits. It seems a fruitless exercise. I suppose it was
worth thinking about, though.
And that is my opinion.
- lowdose 7y agoMaybe this is the best attack on copyright we have. It at least proves the smartest in this case have good intentions.
- helloyoshiworld 7y agowhat do you mean by “not human searchable”? we’ve indexed some pretty large resources before