Annex I - Gestdem 2018/0793
Total Page:16
File Type:pdf, Size:1020Kb
Ref. Ares(2018)1522652 - 20/03/2018 Annex I - Gestdem 2018/0793 List of documents P. 02 1. Powerpoint Presentation by Audible Magic dated 11 July 2016 (Ref. Ares(2017)4595074) P. 14 2. Presentation by Audible Magic: Content Recognition Technology (Ref. Ares(2017)312988) P.28 3. Document - Audible Magic Services Description (Ref. Ares(2017)4595115) 4. Email from Mike Edwards to DG CNECT dated 7 April 2016 and attachment (PowerPoint P. 33 Presentation) (Ref. Ares(2017)4053880) 5. Email from Mike Edwards to DG CNECT dated 7 April 2016 and attachment (Reply to the P. 46 Copyright office) (Ref. Ares(2017)4053919) 6. Email from Mike Edwards to DG CNECT dated 12 January 2017 (Ref. Ares(2017)4054021) P. 54 7. Email from Mike Edwards to CAB Ansip dated 29 June 2017 and attachments (Presentation on Content Recognition Technology; Audible Magic-content recognition technology.pdf) (Ref. Ares(2017)3257224) P. 55 8. Document provided by Google – Comments on the development of the joint strategic plan on intellectual property enforcement (Ref. Ares(2018)1381117) P. 74 9. Document provided by Google – How Google fights piracy 2016 (Ref. Ares(2018)1381255) P. 114 10. Document provided by Google during presentation held in 2015 (Ref. Ares(2018)1381336) P. 175 11. Presentation of Blue Efficience at the European Parliament on 29 November 2016 (Ref. Ref. Ares(2018)1373424) P. 195 1 Ref. Ares(2017)4595074 - 16/08/2017 Doc. 1 Powering the Compliance and Licensing of Copyrighted Content on Social Video Networks Audible Magic Corporation www.audiblemagic.com 2 Audible Magic Overview Business Audible Magic is the defacto standard for copyright compliance services to online platforms and networks. The services identify Overview copyrighted music and television/film in user generated video uploads Online platforms including Dailymotion, Soundcloud, Spinnup, Customers Facebook, Twitch, Vimeo, Tumblr & Partners Content owners including Canal+, RTL, PIAS, Merlin, Universal Music, Sony Music, Warner Music, Fox, NBCU, Disney/ABC • Industry leader and pioneer of content identification technology • In production processing over 4B transactions a year Key • Accurate and robust technology with over 30 patents granted Strengths • Embedded in customer infrastructure/workflow • Services paid for by the social video platforms. No financial relationship with content owners Company • Founded in 1999 Background • Offices in Los Gatos, Berkeley, and London 3 Range of Content Identification Technologies ID Tech Description Effectiveness Cost/file Comment Not a good filter Matches words contained File Name <5% <0.001€ High false negative and in file name false positive rates Not a good filter File Hash Matches file bits exactly <10% <0.0001€ unless paired with other ID technology Good for identifying Matches embedded mark Watermark <10% <0.01-.05€ premium content. Little either visual or audible content is watermarked Standard today for Fingerprint Matches perceptual <0.005 - production. Range of (image, characteristics compared to 80-99% .05€ accuracy among audio, video) original technologies Not practical for primary Human Manual review <50% > 1€ identification process Review Good for counter-notice Effectiveness – How much of the targeted content is matched? 4 Content Identification Fingerprinting 1. Building a Reference Database Inherent Reference Database Characteristics of Registrations A1,B3,C4 A1,B3,C4 Fred Reference A2,B4,C8 Sally Database A6,B7,C4 John Fred S. 2. Matching Unknowns to the References A2,B4,C8 Match Reference Sally Database 5 Powering Social Video Industry’s Copyright Compliance Systems Online Platform Ingest Post/Publish Uploader • Content owners register their works into AM Content Registration databases Copyright • Online platforms integrate AM services to Compliance review at time of ingest (less than a day) Studios • AM notifies online platform of matches and Reference business rule Music Labels/ Database • Online platform blocks or allows posting of Publishers videos • Pricing is based upon number of files analyzed per month. Affordable for small online platforms. 6 Trusted 3rd Party Mediating Between Content and Online Platforms 7 Discussion Topics •Why online platforms pay for content filtering •Take Down, Stay Down Services •Fair Use •Evolution from compliance to licensing 8 Ability to Implement a Take Down Stay Down Service Online Platform Ingest Post/Publish Uploader • Use for Content • Previously taken down copyrighted Registration content Copyright • Content which has been deemed as Compliance inappropriate (sex, hate, or violence) Studios • Ensures exact content does not reappear Registered Take Down on the service Music Labels/ Content Stay Down Publishers • Reduces cost and volume of repeated notice and takedown • Implemented by Myspace in 2008 and Dailymotion today 9 Content Filtering and Fair Use • Industry practice: Online platforms have counter- notice/dispute resolution practices • User can dispute the blocking of their files • Human review • Role of Content Filtering • Reviewed file can be fingerprinted and registered with the usage rule so automatically allowed or blocked in future 10 Too Many Deals Have to Be Done - Overwhelming for Both Sides Copyright Copyright Copyright Copyright Copyright Owner Owner Owner Owner Owner Online Online Online Online Online Online Platform Platform Platform Platform Platform Platform In addition, each online platform has to create a set of tools/systems 11 A Simplified Solution With Standard Licensing and Comprehensive Tools Copyright Copyright Copyright Copyright Copyright Owner Owner Owner Owner Owner Compliance and Licensing Platform Filter Claim License Track Report Collect Distribute Online Online Online Online Online Online Platform Platform Platform Platform Platform Platform 12 The Future of Content Identification • Technology is not static • Manipulations of content • Mashups and mixes • Musical compositions • Evolution from compliance tool to enabling licensing and monetization 13 Ref. Ares(2017)312988 - 20/01/2017 Doc. 2 Content Recognition Technology: Balancing the Needs of Online Platforms, Creators and Users Audible Magic Corporation www.audiblemagic.com 14 Audible Magic Overview Business Audible Magic is a leading provider of automated content recognition technology to online platforms and networks. The services identify Overview copyrighted music and television/film in user generated video uploads Online platforms including Dailymotion, Soundcloud, Spinnup, Twitch, Vimeo, Tumblr. Customers & Partners Content owners including Canal+, RTL, PIAS, Beggars Group, Merlin, Universal Music, Sony Music, Warner Music, Fox, NBCU, Disney/ABC • Industry leader and pioneer of content identification technology • In production processing over 11B transactions in 2016 (4B in 2015) Key • Accurate and robust technology with over 30 patents granted Strengths • Embedded in customer infrastructure/workflow • Services paid for by the social video platforms. No financial relationship with content owners Company • Founded in 1999 Background • Offices in Los Gatos, Berkeley, and London 15 ‘Effective’: A Range of Content Identification Technologies ID Tech Description Effectiveness Cost/file* Comment Not a good filter Matches words contained File Name <5% <0.001€ High false negative and in file name false positive rates Not a good filter File Hash Matches file bits exactly <10% <0.0001€ unless paired with other ID technology Good for identifying Matches embedded mark Watermark <10% <0.01-.05€ premium content. Little either visual or audible content is watermarked Standard today for Fingerprint Matches perceptual <0.005 - production. Range of (image, characteristics compared to Up to 99+% .05€ accuracy among audio, video) original technologies Not practical for primary Human Manual review <50% > 1€ identification process Review Good for counter-notice Effectiveness – How much of the targeted content is matched? *Cost per file is Audible Magic’s own estimate based on market knowledge, and may not reflect actual prices from individual vendors. 16 Content Identification Fingerprinting 1. Building a Reference Database Inherent Reference Database Characteristics of Registrations A1,B3,C4 A1,B3,C4 Fred Reference A2,B4,C8 Sally Database A6,B7,C4 John Fred S. 2. Matching Unknowns to the References A2,B4,C8 Match Reference Sally Database 17 Enabling Online Platforms to Identify Copyright Content Online Platform Ingest Post/Publish Uploader • Music, Film and TV content owners register Content Registration their works into AM databases Copyright • Online platforms integrate AM services to Compliance review at time of ingest (less than a day) Studios • AM notifies online platform of matches and Reference business rule Music Labels/ Database • Online platform allows, monetizes or blocks Publishers posting of videos • Pricing is based upon number of files analyzed per month. Affordable for small online platforms. 18 Trusted 3rd Party Mediating Between Content and Online Platforms 19 AM Customers Voluntarily Comply •Highest standards demanded by copyright owners •Good digital citizens •Maintain good relations with copyright owners •Pre-emption v takedown: • More efficient & less costly • Licensing opportunity • Better user experience 20 SME Case Studies • Swedish start-up • Vietnamese start-up • Distribute unsigned artists to paid • UGC video platform distribution platforms • Limited audience and ad revenue • Need: Prevent fraudulent uploads of potential (Vietnam-only) others’ creations • Around 5,000 uploads