Dennis Collopy
Dennis Collopy is a Senior Research Fellow at the University of Hertfordshire (UH) in the UK. He specialises in Music and IP Right related research that has included various studies for UK Music and the Intellectual Property Office (IPO). As well as almost two decades of academic experience, including co-founding the Music Industry Management course at UH in 2007 and the International Music Business Research Association (IMBRA) in Vienna in 2012 , Dennis has spent over 4 decades in music working across artist management, record labels and music publishing having been MD of Riva Music (signing the Clash and John Mellencamp), BMG Music Publishing (signing Steve Earle and Maria McKee and working with Eurythmics and Clannad), EG (working with KLF, the Orb and Robert Fripp) and currently his own Menace Music Management, whose focus is the management of rights for artists, producers and songwriters including Matt Aitken, Frankie Miller, Slowdive/Mojave 3’s Neil Halstead, Steve Edwards, Lisa Millett and Gary Benson. Dennis was formerly a director of PRS and board member of the Music Publishers Association.
Phone: 01707281398
Address: School of Creative Arts, University of Hertfordshire, College Lane Campus, Hatfield, Herts, AL10 9AB
Phone: 01707281398
Address: School of Creative Arts, University of Hertfordshire, College Lane Campus, Hatfield, Herts, AL10 9AB
less
InterestsView All (33)
Uploads
Papers by Dennis Collopy
In relation to the use of AI in the enforcement of IPR, the number of challenges exceeds the number of opportunities, and this appears mainly a result of the number of fundamental issues relating to AI in the enforcement of Patents and Trade Secrets.
There are also increased concerns about the reliability of Machine Learning (ML), as a common form of AI, including a study that identified common methodological issues relating to the use of ML in the quantitative sciences.
Yet we remain confident of the ability of AI /ML to offer scalable solutions to assist the enforcement of some if not all the IPR under consideration. We also stress that AI/ML itself is constantly improving.
We note the emergence of adversarial ML where bad actors can exploit vulnerabilities to exploit AI systems and alter their behaviour to serve a malicious end goal. These attacks can involve poisoning (of the training data) or evasion attacks, many of which go unnoticed until there is a ML critical failure.
We cannot recommend the increased adoption of the technology without emphasizing the drawbacks involved that include concerns about the following:
• Non-transparent AI (aka black box AI) cannot be interrogated in the same way as systems that enable a full audit trail.
• We note that certain firms (including the largest tech platforms) creating
and deploying AIs admit they do not fully understand the workings of their own AI’s.
• Inherent and systemic biases can damage the reliability and indeed veracity of the findings produced by the AI.
• The variable quality of data used to train an AI is a fundamental issue even for those IPR sectors successfully using fully automated rights enforcement systems.
• Ensuring the human in the loop approach is sustained throughout the enforcement process requires training and additional financial resources.
• The limits of image recognition as well as NLP need to be considered in
any enforcement process especially one involving summary judgements as are commonly seen in the copyright sector.
• Whilst some automated enforcement systems are seen as quite successful (notably YouTube’s Content ID) none are perfect.
We recommend the careful piloting of any new AI based IPR enforcement system to determine whether the technology is operating within ethical, moral, and legal boundaries to achieve its primary purposes.
• to assess the role that social media plays in the sale and distribution of counterfeited and pirated physical goods from six representative sectors: alcohol, cigarettes, clothing, footwear, perfume and watches.;
• to estimate recent levels of counterfeiting within the UK;
• to understand the extent to which this is moving online; and
• to gauge how it is helped to do so by online social media platforms.
• The study specifically aimed to assess the scale, impact and characteristics of infringements, as well as opportunities for IP infringement.
What is ‘stream-ripping’?
Stream-ripping is the obtaining of a permanent copy of content that is streamed online. The process can be carried out on audio and audio-visual content and, in either case, it is possible to create an audio-only copy of the music. Once a copy is created and saved, it is possible for a user to listen to it offline and share it between their devices.
Our key findings
In nearly two years (Jan 2014 – September 2016), the use of stream-ripping services increased by 141.3%, dwarfing the growth seen for other types of music-specific infringing sites.
In the time period of just one month, the use of stream ripping sites made up the majority (68.2%) of the total usage across the 50 most popular music-only infringing sites.
Advertising is the main funding model associated with stream-ripping services.
YouTube is by far the most popular source of content for stream-ripping services (used by 75 of the 80 services surveyed).
57% of those UK adults surveyed claimed to be aware of stream-ripping services.
Those who claimed to have used a stream-ripping service were significantly more likely to be male, ABC1 social grade, and between the ages of 16 to 34 years.
Within the reasons driving stream-ripping the most common response was that the music was already owned in another format (31%), with wanting to listen to music offline (26%) and on the move (25%) the next most commonly given responses. Unaffordability (21%) and feeling official content is overpriced (20%) coming in after these reasons.
Source and other information https://prsformusic.com/what-we-do/influencing policy/stream-ripping
https://www.gov.uk/government/news/illegal-streaming-threatens-copyright-progress
In relation to the use of AI in the enforcement of IPR, the number of challenges exceeds the number of opportunities, and this appears mainly a result of the number of fundamental issues relating to AI in the enforcement of Patents and Trade Secrets.
There are also increased concerns about the reliability of Machine Learning (ML), as a common form of AI, including a study that identified common methodological issues relating to the use of ML in the quantitative sciences.
Yet we remain confident of the ability of AI /ML to offer scalable solutions to assist the enforcement of some if not all the IPR under consideration. We also stress that AI/ML itself is constantly improving.
We note the emergence of adversarial ML where bad actors can exploit vulnerabilities to exploit AI systems and alter their behaviour to serve a malicious end goal. These attacks can involve poisoning (of the training data) or evasion attacks, many of which go unnoticed until there is a ML critical failure.
We cannot recommend the increased adoption of the technology without emphasizing the drawbacks involved that include concerns about the following:
• Non-transparent AI (aka black box AI) cannot be interrogated in the same way as systems that enable a full audit trail.
• We note that certain firms (including the largest tech platforms) creating
and deploying AIs admit they do not fully understand the workings of their own AI’s.
• Inherent and systemic biases can damage the reliability and indeed veracity of the findings produced by the AI.
• The variable quality of data used to train an AI is a fundamental issue even for those IPR sectors successfully using fully automated rights enforcement systems.
• Ensuring the human in the loop approach is sustained throughout the enforcement process requires training and additional financial resources.
• The limits of image recognition as well as NLP need to be considered in
any enforcement process especially one involving summary judgements as are commonly seen in the copyright sector.
• Whilst some automated enforcement systems are seen as quite successful (notably YouTube’s Content ID) none are perfect.
We recommend the careful piloting of any new AI based IPR enforcement system to determine whether the technology is operating within ethical, moral, and legal boundaries to achieve its primary purposes.
• to assess the role that social media plays in the sale and distribution of counterfeited and pirated physical goods from six representative sectors: alcohol, cigarettes, clothing, footwear, perfume and watches.;
• to estimate recent levels of counterfeiting within the UK;
• to understand the extent to which this is moving online; and
• to gauge how it is helped to do so by online social media platforms.
• The study specifically aimed to assess the scale, impact and characteristics of infringements, as well as opportunities for IP infringement.
What is ‘stream-ripping’?
Stream-ripping is the obtaining of a permanent copy of content that is streamed online. The process can be carried out on audio and audio-visual content and, in either case, it is possible to create an audio-only copy of the music. Once a copy is created and saved, it is possible for a user to listen to it offline and share it between their devices.
Our key findings
In nearly two years (Jan 2014 – September 2016), the use of stream-ripping services increased by 141.3%, dwarfing the growth seen for other types of music-specific infringing sites.
In the time period of just one month, the use of stream ripping sites made up the majority (68.2%) of the total usage across the 50 most popular music-only infringing sites.
Advertising is the main funding model associated with stream-ripping services.
YouTube is by far the most popular source of content for stream-ripping services (used by 75 of the 80 services surveyed).
57% of those UK adults surveyed claimed to be aware of stream-ripping services.
Those who claimed to have used a stream-ripping service were significantly more likely to be male, ABC1 social grade, and between the ages of 16 to 34 years.
Within the reasons driving stream-ripping the most common response was that the music was already owned in another format (31%), with wanting to listen to music offline (26%) and on the move (25%) the next most commonly given responses. Unaffordability (21%) and feeling official content is overpriced (20%) coming in after these reasons.
Source and other information https://prsformusic.com/what-we-do/influencing policy/stream-ripping
https://www.gov.uk/government/news/illegal-streaming-threatens-copyright-progress
Source: https://musicbusinessresearch.wordpress.com/2014/10/14/5th-vienna-music-business-research-days-in-retrospective/