From my background in neural networks, and my understanding of how
they work, I would say that your trained network is a derived work if the weights are learned from a specific training sample. If it does not learn from a specific sample it is not a derived work from that sample. It is not sufficient that a specific image is in a training batch, it must also trigger learning.
The goal for the training is partial or full reconstruction of properties necessary for some operation. It is perhaps easier to see with an example: Imagine you have photos of Picasso's art, and you try to estimate whether some art is within his Blue Period. Is the network derived work from Picasso's art? And how does that compare to whether the network itself is a work of art, is it still derived from another (set of) work of art?
I would say training of neural networks is a derived work built on the training samples, but I would not say itself is a derived work of art, even if it can be copyrighted.
The most common solution these days seems to be to encode what is called "eigenfaces", and then use those to create a kind of hashes for similarity detection. Eigenfaces are a kind of arketypes of how a face look like, and mixing such faces creates new ones. It is a bit similar to those flipover albums the police uses, but allowing a lot more of gradual changes. Face detection is often visualized as vector points in movies, but this is not how eigenfaces work. Or rather, it creates vectors, but not as dots on a portrait.
Note that ClearviewAI goes a good bit longer than just learn some variant of eigenfaces.
On Sun, Jan 19, 2020 at 4:30 AM Benjamin Ikuta benjaminikuta@gmail.com wrote:
Thanks for that.
Pardon me if I've missed something, but that seems to imply, but not directly state, that AI training is a derivative work; could you clarify that?
On Jan 18, 2020, at 2:58 PM, Ryan Merkley rmerkley@wikimedia.org wrote:
[My comments are my own, and don’t reflect or suggest any official position from WMF]
The NBC story linked below come out about a year ago. Around the same time, when I was CEO at Creative Commons, we published a statement and updated FAQs that attempted to respond to questions being asked about permitted uses and attribution related to the licenses.
CC’s statement (March 2019) is here: https://creativecommons.org/2019/03/13/statement-on-shared-images-in-facial-... https://creativecommons.org/2019/03/13/statement-on-shared-images-in-facial-recognition-ai/ The FAQs are here: https://creativecommons.org/faq/#artificial-intelligence-and-cc-licenses https://creativecommons.org/faq/#artificial-intelligence-and-cc-licenses
r.
Ryan Merkley (he/him) Chief of Staff, Wikimedia Foundation https://wikimediafoundation.org/
rmerkley@wikimedia.org mailto:rmerkley@wikimedia.org @ryanmerkley https://twitter.com/ryanmerkley +1 416 802 0662
On Jan 18, 2020, at 2:14 PM, John Erling Blad jeblad@gmail.com wrote:
There are several reports of face recognition going mainstream, often in less than optimum circumstances, and often violating copyright and licenses
https://www.nytimes.com/2020/01/18/technology/clearview-privacy-facial-recog... https://www.nbcnews.com/tech/internet/facial-recognition-s-dirty-little-secr... https://www.ibm.com/blogs/research/2019/01/diversity-in-faces/
In my opinion building a model for face recognition is a derived work, and as such must credit the photographers. That pose a real problem when the photographers counts in the millions and billions. Even a 1px fine print would be troublesome!
What is the official stance on this? Is it a copyright infringement or not, does the license(s) cover the case or not?
John Erling Blad /jeblad
Wikimedia-l mailing list, guidelines at: https://meta.wikimedia.org/wiki/Mailing_lists/Guidelines and https://meta.wikimedia.org/wiki/Wikimedia-l New messages to: Wikimedia-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l, mailto:wikimedia-l-request@lists.wikimedia.org?subject=unsubscribe
Wikimedia-l mailing list, guidelines at: https://meta.wikimedia.org/wiki/Mailing_lists/Guidelines and https://meta.wikimedia.org/wiki/Wikimedia-l New messages to: Wikimedia-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l, mailto:wikimedia-l-request@lists.wikimedia.org?subject=unsubscribe
Wikimedia-l mailing list, guidelines at: https://meta.wikimedia.org/wiki/Mailing_lists/Guidelines and https://meta.wikimedia.org/wiki/Wikimedia-l New messages to: Wikimedia-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l, mailto:wikimedia-l-request@lists.wikimedia.org?subject=unsubscribe