 The Clarinet BBoard
|
Author: Mark Charette
Date: 2018-09-07 23:16
I was going to say it's funny ... but it's not. I work in AI, and pattern matching is anything but perfect. It's pretty good - but when you have potentially billions of images and sound recordings to look at an accuracy of 99.9% isn't near good enough.
|
|
Reply To Message
|
|
Author: Paul Aviles
Date: 2018-09-08 22:27
I don't fully understand what Sony and others are doing. Is there a licensing revenue stream at stake? I recently attended school where the youngins "torrented" (downloaded for free) everything. I only see this validating their choice. Big companies need to feel fortunate for the numbers willing to pay for streams and ownership that they have..........before they lose that morsel. If I understand this correctly that is.
.............Paul Aviles.
|
|
Reply To Message
|
|
Author: kdk
Date: 2018-09-08 22:59
This is absurd sounding, of course, because, though SONY may have rights over use of its own recordings, it doesn't own the rights to the music. SONY can only protect the specific performances on those recordings it owns.
There *has to be* more to this than than the little blurb in boing boing includes. James Rhodes is (even if not a household name) apparently an established concert pianist. He may have made this recording under SONY's auspices or SONY may have bought the rights to the recording from whoever owned it originally. Cory Doctorow doesn't really say.
If it's the case that SONY and other record producers can identify their own protected recordings with AI algorithms, then people who post them will have to beware. And performers who make recordings will need to consider protecting their own interests when they sign recording contracts. If the algorithm is faulty, then that's a whole other issue. But claiming ownership of the music itself seems silly if true or an exaggeration on Doctorow's part.
Karl
|
|
Reply To Message
|
|
Author: gwie
Date: 2018-09-09 00:02
This has been happening on YouTube for quite awhile now.
A software algorithm is used to analyze and compare the waveform and flag things that match. It can be done very quickly on large amounts of audio data, however, the quality of the matching is only as good as the algorithm that performs that task. Computers aren't "smart;" they only perform the instructions that we give them, and automating some tasks like this give us a level of accuracy far below what a human operator could accomplish.
The irritating factor is that the companies involved disable or limit the access to the content automatically if it gets flagged, and the person who controls that content then has to prove that it isn't infringing. This is extremely frustrating because a lot of classical music is in the public domain, and the algorithms often times are NOT sophisticated enough to distinguish between different performances of them.
Each season, I post dozens of performances of music by Haydn, Mozart, Beethoven, Mendelssohn, etc. from my Summer Music Festival on YouTube and Facebook, then spend another whole day submitting the same explanation over and over when a number of them get flagged for potentially infringing on copyright: "I am the rights-holder of this recording of this performance. The music performed is in the public domain, the sheet music is a public domain source, and [name of company] has no legitimate copyright claim to this recording in any way, shape, or form."
Post Edited (2018-09-09 07:41)
|
|
Reply To Message
|
|
The Clarinet Pages
|
 |
.jpg){kind=small}
.jpg){kind=small}
.jpg){kind=small}
.jpg){kind=small}
.jpg){kind=small}
.jpg){kind=small}
