Earlier this month, Dr Sebastian Tomczak was hit by 5 copyright claim infringements for his 10-hour white noise video uploaded on YouTube.
These infringements were incorrectly identified by YouTube’s automated Content ID system which uses algorithms to find close matches of content between videos. The story has gained global media attention from the likes of the ABC and the BBC.
We caught up with Dr Sebastian Tomzcak, music technologist and lecturer at the Electronic Music Unit at the University of Adelaide, to talk about how something like this can occur and what has happened since the copyright infringements were made.
Q: Tell us about the 10-hour white noise track you made and what prompted you to make the track.
I created and uploaded this white noise video in July 2015. At the time, I was working on a project for the Cove Civic Centre (City of Marion) alongside Prof. Stephen Whittington (Elder Conservatorium) and Dr. Margit Bruenner (School of Architecture and Built Environment). One aspect of this project involved sound design that was long and continuous. I researched different kinds of sonic textures that would not be intrusive when listened to at low levels for hours at a time.
As part of this research, I generated ten hours’ worth of white noise, which I then uploaded to YouTube. I would see a video like this as a tool to help people with their concentration at work, or improvement of quality of sleep. The white noise itself has been digital generated.
Q: You have been hit with 5 copyright infringement claims against the video on YouTube. How does something like this happen?
YouTube has a Content ID system, whereby content creators (who are in the YouTube Partner Program) can upload music and audio. The system will then scan any new content (uploaded by other people) for matches with existing Content ID material – with the assumption that this may constitute a copyright infringement.
If a match is found, a Content ID copyright claim can be automatically generated, which means that the newly-created video may be demonetised or blocked. In my case, the white noise matched up with five separate pieces of audio found within the Content-ID system. As a result, these five copyright claims were placed against my video.
Normally, it is a manual process to dispute such claims – I would have to indicate why I thought I should retain the rights to show and advertise on the video. In this case – most likely due to the attention this received on Twitter – YouTube cancelled the five claims.
I think the fascinating thing about this case is that – unlike many other YouTube copyright cases – an A / B comparison is not required. Many people can imagine the absurdity when a randomised signal such as white noise is caught up in a copyright system.
Q: In a world that is becoming increasingly reliant on artificial intelligence, how can an algorithm such as this get it so wrong?
In my opinion, there are two issues at play in this case – and we can probably see these in the bigger picture as well.
I think that algorithms can, at times, reveal their imperfections due to being built by humans, who are themselves imperfect. I can completely understand YouTube’s necessity for algorithms to help police copyright claims, but cases such as this white noise video and similar incidents are demonstrations of an overbearing system.
We also need to be cautious of incentivising negative behaviours in any systems that we build. There are scenarios wherein the Content ID system can be misappropriated for financial gain. For instance, with some Content ID claims, advertising revenue will automatically be diverted from the uploader to the claimant, which in itself is based on automatic pattern matching. A little more human intervention may still be useful in these situations.
Want to hear more from Sebastian Tomczak?
Sebastian is part of the free Music Technology Foundations online course. This 6 week self-paced course will teach learners how to use creative technologies to make their own music and get a step closer to a career in music.
This course opens on the 15th February 2018.
Find out more here.
About Sebastian Tomczak
Sebastian Tomczak (A.K.A Little-Scale) is an Adelaide based chiptune artist, music technologist and lecturer at the Electronic Music Unit at the Elder Conservatorium.
Tomczak is a music technologist and holds a PhD from the University of Adelaide. He has an interest in hardware and software development, including physical interfacing and chip music. He has presented, performed and exhibited with his technology and music in Melbourne, Sydney, Los Angeles, New York, Mumbai and Belfast.