#LyX 1.3 created this file. For more info see http://www.lyx.org/ \lyxformat 221 \textclass article \language british \inputencoding auto \fontscheme bookman \graphics default \paperfontsize default \spacing single \papersize a4paper \paperpackage a4 \use_geometry 0 \use_amsmath 0 \use_natbib 0 \use_numerical_citations 0 \paperorientation portrait \secnumdepth -1 \tocdepth 0 \paragraph_separation indent \defskip medskip \quotes_language english \quotes_times 2 \papercolumns 1 \papersides 1 \paperpagestyle default \layout Title Whose future is it? \layout Author Philippe Aigrain \begin_inset Foot collapsed false \layout Standard The author is presently CEO of \begin_inset LatexCommand \htmlurl[Sopinspace, Society for Public Information Spaces]{www.sopinspace.com} \end_inset . His personal page is at \begin_inset LatexCommand \url{http://www.publicdebate.net/Members/paigrain/} \end_inset \end_inset \layout Date 12 October 2004, keynote speech at the ISMIR 2004 Conference, Barcelona \layout Address \begin_inset ERT status Open \layout Standard \backslash copyright \end_inset \SpecialChar ~ Philippe Aigrain, 2004. This text and the associated slides can be used under the terms of the \begin_inset LatexCommand \htmlurl[Creative Commons Attribution License 2.0]{http://creativecommons.org/licenses/by/2.0/} \end_inset \layout Standard Good morning. It is a great honour to deliver this talk to a research community of which I have been a member when we were not sure that it was a domain of its own. It has since then developed in an impressive manner, and today it offers all the signs of a maturing scientific community in a technical field: a yearly conference with growing attendance, some structuring paradigms, progress in modelling its challenges, attraction for bright graduate students, evaluation testbeds and benchmarks. That's exactly when a community generally puts itself on tracks that will lead it to deliver little or much progress for science, little or much usefulness for society. That's when it is worth for such a community to think a while about where it tries to go, to take \color red [Slide: one step aside] \color default . \layout Standard This honour is also a challenge. If I were totally ignorant about information retrieval and content processing of music, it would be much more comfortable. I would share with you some general thoughts about property and commons for information and information technology, about how research funding mechanisms and intellectual property rules influence the targets of research efforts. And from my blissful ignorance, I would leave you with the task of figuring out what it can mean for you and the ISMIR community. But I can't. I know just enough about music information retrieval to feel forced to connect my general thinking with concrete examples in this domain and neighbour ing fields. Inevitably, I will say a few stupidities in this process. I hope you won't stop there, at identifying these mistakes, that you will still try to understand the issues I am raising, and only then, decide if they are worth your consideration. \layout Subsection \emph on Little boxes \layout Standard In 1998, Andrew Odlyzko \begin_inset Foot collapsed false \layout Standard Andrew Odlyzko, Content is not king, First Monday, 6(2), February 2001, revised version on an original 1998 paper. \end_inset and myself \begin_inset Foot collapsed false \layout Standard Philippe Aigrain, \begin_inset Quotes eld \end_inset The Best of Both Worlds: Can Mesia Quality be Combined with Open Contents? \begin_inset Quotes erd \end_inset , invited talk af IFTA's world conference, Vienna, 1998, \begin_inset LatexCommand \htmlurl{http://www.debatpublic.net/Members/paigrain/texts/IFTA.pdf} \end_inset . \end_inset had the same outcry of intellectual revolt: \color red [Slide: \begin_inset Quotes eld \end_inset Content is not king! \begin_inset Quotes erd \end_inset ] \color default . We actually meant something different. I meant simply \begin_inset Quotes eld \end_inset people are \begin_inset Quotes erd \end_inset , while he meant \begin_inset Quotes eld \end_inset communication is \begin_inset Quotes erd \end_inset . However our message was consistent, and though it was not intended specifically for music, it applies to it. Music is a relationship between people, between gesture and sound, between mind and percept. Music is not contents. Of course, there have been \emph on music boxes \emph default (I mean tools that automatically produce music) for many centuries, and \emph on boxed music \emph default (I mean recorded music on carriers that is sold as a commodity) for a little more than a century. However, we should not let our minds be framed in the \emph on little boxes \emph default that the father of Anthony Seeger (last year's keynote speaker) has made famous by interpreting Malvina Reynolds' song \color red [Slide] \color default . Don't worry if you think my implicit statement is demagogic, there is also a version by Woody Guthrie (the \begin_inset Quotes eld \end_inset beatnik version \begin_inset Quotes erd \end_inset ) where he makes fun of the critics. Anyway, music is not boxed music. Music is so much a relationship that if someone knows to play any of it, he or she can earn some money in the streets of many cities in the world, which is sort of a miracle since at the same time, boxed music is forced upon us in the stores and restaurants of the same cities. Of course this miracle is possible only because and when street musicians don't pay performance rights. \layout Standard So what has happened around us in the past 5 years? What has happened to people, the way they access music and use it ? What has happened to people, the way they do research and use it? \layout Subsection \emph on The ground beneath our feet \layout Standard \color red [Slide: Societal peer-to-peer file sharing networks have become a more efficient source of access to published and broadcasted material than libraries and archives] \color default . Raise your hand if you don't believe me. Of course this is a very general statement. It is absolutely not true for unpublished primary source material. But for material that has been published in any form, if you take a real benchmark, not an hypothetical researcher that would instantly travel to all libraries and archives of the world and have credentials there, but a real person, who suddenly wants access to one performance of \begin_inset Quotes eld \end_inset La Ultima Noche \begin_inset Quotes erd \end_inset , or a documentary movie that was broadcasted 2 years ago on some channel, the chances he or she will succeed in getting it are much higher on P2P networks. Of course this is only recently developing, and it is under threat. Legal and police threats, but also threats from the private war gangs who inject fakes under the name of P2P warfare. I will come back later to how one can regain the benefits of this wonderful societal mechanism of P2P file sharing, while appeasing some of the pains it is claimed to cause. \layout Standard \color red [Slide: Remix and hiphop scenes are competing with \begin_inset Quotes eld \end_inset official \begin_inset Quotes erd \end_inset contemporary music composition as promising users for content-based synchronisa tion and interaction] \color default I guess Paull Miller's quote speaks for itself. This is not a new development, but research and creation organisations are slow to react to it. Many people in your community are involved as musicians or technicians in these communities, and some are even deriving inspiration from them, such as the idea of \emph on query by beat boxing \emph default . \layout Standard \color red [Slide: Creative commons have gone from idea to project to reality in many domains] \color default . Many people in this community are active supporters or would-be supporter of musical commons. However, very few are actively supported by their organisations in building them. Creative commons licenses are moving us out of the false dilemna between respecting music by restricting the relationship that people build can build to it, and making it a commodity that it is freely lootable for the boxed music business. \layout Standard In WIPO and UNESCO, there is presently a debate about the so-called \begin_inset Quotes eld \end_inset protection of traditional knowledge, folklore and genetics resources \begin_inset Quotes erd \end_inset . This is supposed to be an answer to the well-grounded complaints by developing countries that powerful businesses are pirating (in the real sense) their traditional culture and natural resources. What WIPO -and at a lesser extent UNESCO- propose to them is to participate into the loot. Enclose them into property, and become little looters of the common goods. Of course the little looters will be locally powerful guys, and they will have to make deals with the big looters for deriving true income, and you can be certain that this income will flow to about anywhere except into the pockets of the poor people who created and cared for these treasures. Creative commons, and fair global redistribution are not an easy path, but at least it is a path in the right direction. The same is true in the developed world. In the era of information, inequality of power just as inequality of wealth is fractal: you find it between countries, between regions in a country, between neighbourhoods in a city, and even between companies and musicians. \layout Standard \color red [Slide: Software and information patentability are the object of fierce political debates in Europe and globally] \color default . ISMIR is a community of open exchange, where all proceedings are accessible free of charge, and usable freely for some usage (even if it would be better defined in my opinion through a Creative Commons license than by the present clause). But this open exchange is stopped at the border of becoming fact. Technology transfer departments of public research centres are patenting algorithms like hell, and intellectual property divisions of companies are patenting software systems components and application features. \layout Standard \color red After 20 years, reviews of the impact of the Bayh-Dole act conclude to an unsignificant contribution of licensing as a funding source for public labs \color default . However, the same technology transfer departments conclude from their failure that you should help them do more of the same. The funny thing is that while they carefully enclose the public domain of research, innovation disseminates from other sources, even if they are initially much less sophisticated technically, simply because they have chosen cooperation and sharing instead of enclosures. \layout Subsection \emph on Inventory of fixtures \layout Standard While preparing this talk, I read through the summaries of communications at this conference, and in a number of cases through the proceedings paper version. I produced a taxonomy \color red [Slide] \color default , which is slightly more refined than the one the program committee has done to organise the sessions. Which insight can we derived from it? \layout Standard \begin_inset Float table wide false collapsed false \layout Standard \align right \begin_inset Tabular \begin_inset Text \layout Standard \series bold Topic \end_inset \begin_inset Text \layout Standard \series bold Number \end_inset \begin_inset Text \layout Standard \series bold Topic \end_inset \begin_inset Text \layout Standard Number \end_inset \begin_inset Text \layout Standard \series bold \emph on Feature extraction, classification in one feature space, similarity measures, structure extraction \end_inset \begin_inset Text \layout Standard 45 \end_inset \begin_inset Text \layout Standard \series bold \emph on Systems \end_inset \begin_inset Text \layout Standard \series bold \emph on 12 \end_inset \begin_inset Text \layout Standard Tempo and beats (including tracking) \end_inset \begin_inset Text \layout Standard 7 \end_inset \begin_inset Text \layout Standard Distributed music processing \end_inset \begin_inset Text \layout Standard 1 \end_inset \begin_inset Text \layout Standard Timbre incl. drum sounds and gestures \end_inset \begin_inset Text \layout Standard 5 \end_inset \begin_inset Text \layout Standard Metadata databases and browsers \end_inset \begin_inset Text \layout Standard 2 \end_inset \begin_inset Text \layout Standard Instrument \end_inset \begin_inset Text \layout Standard 2 \end_inset \begin_inset Text \layout Standard Digital libraries and Web services MIR systems \end_inset \begin_inset Text \layout Standard 3 \end_inset \begin_inset Text \layout Standard Drum patterns, rhythm \end_inset \begin_inset Text \layout Standard 4 \end_inset \begin_inset Text \layout Standard Fingerprinting for DRMs \end_inset \begin_inset Text \layout Standard 3 \end_inset \begin_inset Text \layout Standard Voice (singing detection), singing language \end_inset \begin_inset Text \layout Standard 2 \end_inset \begin_inset Text \layout Standard Real-time synchro (w/contextual info, w/ accompaniment) \end_inset \begin_inset Text \layout Standard 2 \end_inset \begin_inset Text \layout Standard Tonality, key \end_inset \begin_inset Text \layout Standard 3 \end_inset \begin_inset Text \layout Standard Toolboxes (Matlab) \end_inset \begin_inset Text \layout Standard 2 \end_inset \begin_inset Text \layout Standard Chords \end_inset \begin_inset Text \layout Standard 2 \end_inset \begin_inset Text \layout Standard \series bold \emph on Evaluation, testbeds, and benchmarking \end_inset \begin_inset Text \layout Standard \series bold \emph on 9 \end_inset \begin_inset Text \layout Standard Motivic and melodic lines \end_inset \begin_inset Text \layout Standard 10 \end_inset \begin_inset Text \layout Standard Feature extraction and similarity \end_inset \begin_inset Text \layout Standard 2 \end_inset \begin_inset Text \layout Standard Ornementation \end_inset \begin_inset Text \layout Standard 1 \end_inset \begin_inset Text \layout Standard Query-by-humming based systems \end_inset \begin_inset Text \layout Standard 2 \end_inset \begin_inset Text \layout Standard Other segmentation \end_inset \begin_inset Text \layout Standard 2 \end_inset \begin_inset Text \layout Standard MIR systems \end_inset \begin_inset Text \layout Standard 2 \end_inset \begin_inset Text \layout Standard Structure extraction \end_inset \begin_inset Text \layout Standard 5 \end_inset \begin_inset Text \layout Standard MIR systems with DRMs aspects \end_inset \begin_inset Text \layout Standard 3 \end_inset \begin_inset Text \layout Standard Sound synthesis meta-language \end_inset \begin_inset Text \layout Standard 1 \end_inset \begin_inset Text \layout Standard \series bold \emph on Classification (artists, songs, genres) \end_inset \begin_inset Text \layout Standard \series bold \emph on 7 \end_inset \begin_inset Text \layout Standard Optical music recognition \end_inset \begin_inset Text \layout Standard 1 \end_inset \begin_inset Text \layout Standard Clustering pieces and collections \end_inset \begin_inset Text \layout Standard 2 \end_inset \begin_inset Text \layout Standard Multi-feature and sequence matching \end_inset \begin_inset Text \layout Standard 5 \end_inset \begin_inset Text \layout Standard Artists \end_inset \begin_inset Text \layout Standard 1 \end_inset \begin_inset Text \layout Standard \series bold \emph on Usage paradigms and their support \end_inset \begin_inset Text \layout Standard \series bold \emph on 23 \end_inset \begin_inset Text \layout Standard Genre \end_inset \begin_inset Text \layout Standard 4 \end_inset \begin_inset Text \layout Standard Sampling to MIDI instrument \end_inset \begin_inset Text \layout Standard 1 \end_inset \begin_inset Text \layout Standard \series bold \emph on Empirical studies of usage \end_inset \begin_inset Text \layout Standard \series bold \emph on 6 \end_inset \begin_inset Text \layout Standard Query by humming \end_inset \begin_inset Text \layout Standard 1 \end_inset \begin_inset Text \layout Standard Personal music collections \end_inset \begin_inset Text \layout Standard 2 \end_inset \begin_inset Text \layout Standard Query by voice percussion \end_inset \begin_inset Text \layout Standard 2 \end_inset \begin_inset Text \layout Standard Digital libraries MIR systems \end_inset \begin_inset Text \layout Standard 2 \end_inset \begin_inset Text \layout Standard Annotation \end_inset \begin_inset Text \layout Standard 3 \end_inset \begin_inset Text \layout Standard P2P MIR systems \end_inset \begin_inset Text \layout Standard 1 \end_inset \begin_inset Text \layout Standard Notation, pitch spelling \end_inset \begin_inset Text \layout Standard 2 \end_inset \begin_inset Text \layout Standard Social networks (artists) \end_inset \begin_inset Text \layout Standard 1 \end_inset \begin_inset Text \layout Standard Transcription, music to score, mapping to patterns \end_inset \begin_inset Text \layout Standard 4 \end_inset \begin_inset Text \layout Standard \end_inset \begin_inset Text \layout Standard \end_inset \begin_inset Text \layout Standard Alignment to score \end_inset \begin_inset Text \layout Standard 4 \end_inset \begin_inset Text \layout Standard \end_inset \begin_inset Text \layout Standard \end_inset \begin_inset Text \layout Standard Production and usage of textual contextual info \end_inset \begin_inset Text \layout Standard 3 \end_inset \begin_inset Text \layout Standard \end_inset \begin_inset Text \layout Standard \end_inset \begin_inset Text \layout Standard Document space representation \end_inset \begin_inset Text \layout Standard 3 \end_inset \begin_inset Text \layout Standard \end_inset \begin_inset Text \layout Standard \end_inset \end_inset \layout Caption Topics at ISMIR'20204 \end_inset \layout Standard Don't worry, you do not have to read all this stuff. The table is in the written version of my talk. Let's look at the top-level categories. First, good news, the ISMIR community is doing lots of basic technical work, more than a third of all papers, more if you count the evaluation benchmarks. However, the mass of accumulated know-how is far from being available as components for practical music applications. This is where the phenomena highlighted in my slide on software patents and technology transfer hits stronger. Nothing specific of the music information field: as soon as something seems to have some real potential, the motto seems to be \begin_inset Quotes eld \end_inset let's make sure than nobody can become rich with it other than us \begin_inset Quotes erd \end_inset . Why is this so detrimental? Because then nobody does. Because the idea that in a field like music information any single organisation , even with its industry partners, will succeed in disseminating for usage a global platform is ridiculous. Nobody can have a clue of what people will really do with a full chain of music information processing, retrieval, re-creation, exchange. The last who can know, or more precisely the last who can accept to know are the major companies. You don't believe me, well why do you think that for years Sony and Philips have carefully buried all the nice ideas and technology that we all know exist in their labs? Only when we have a set of freely re-usable and adaptable music information technology modules, that are more or less combinable, and with which people can explore new ways of usage, will we start to see some real usage. I will come back to that later. \layout Standard At the other end of the technology spectrum, the systems and testbeds end, there is also some very nice work, but it suffers from a related disease: restrictive rights and their implementation by DRMs. Don't mistake the dreams of majors and their trusted technology suppliers for reality. DRMs have no future, though they can wreck ours. Only creative commons and digital rights information are worthwhile, because only them enable the relationship which is music. Don't let yourselves be boxed by boxed music. What's wrong with DRMs? Access and usage control by technology. The user as dangerous enemy. And their complexity. No idea that is that complex, and that interacts with human behaviour ever worked. The Internet would not work if it had been designed with that much determinism, that much a-priori rules about what people can and can't do. And decreasing returns. DRMs and fingerprinting surveillance have a cost per item, and the returns are decreasing. DRMs may be sold under the name of fair reward, but all they will do is augment the already huge concentration of reward on bestsellers. I can hear you thinking \begin_inset Quotes eld \end_inset as we were the ones deciding \begin_inset Quotes erd \end_inset . Yes, you are. Many of your research centres do produce music, all of it, from composition to publishing. Many of your organisations care for public archives, and have accepted for years without enough reaction that the various extensions of copyright duration would stop anything to enter the public domain. Many of your organisations are producing music education software that is used in schools on some predefined music corpora. Is it acceptable that this is often proprietary software (paid by public money), and music corpora with terms of use that forbid any real-world usage? Do you think music learning stops at the walls of the class-room? \layout Standard In between, we have what I have called \begin_inset Quotes eld \end_inset usage paradigms and their support \begin_inset Quotes erd \end_inset . This is often the most important item in a technical community: the structuring meeting point. The ISMIR community has gone a long way there, from the dominance of the initial Query by Humming paradigm to things like alignment to score (often a great enabler), and document space representation, that has proved a powerful approach in other fields of information retrieval. There is clearly still a long way to go, many other paradigms to discover, so I can address you with a less contentious statement: just let's keep our minds open. \layout Standard Finally, there is a new category: real world usage studies and the connected music classification. This is a very welcome category, at least when work there starts with an open mind. For instance, this is the only domain of MIR where people consider P2P file sharing as an interesting element of reality, and not a compromising devil. I have been surprised however about how many people consider automatic genre classification. It seems to me that \begin_inset Quotes eld \end_inset genre \begin_inset Quotes erd \end_inset is a concept that music information technology is destroying quickly. It is a concept that was invented for formatting (for instance in radio) and marketing purposes (for instance in record stores), because there was no better way to point people to what they might like. So there is a risk that the target of recognition will evaporate faster than recognition technology will develop. \layout Subsection \emph on A selection of music information dreams \layout Standard I will conclude on a \color red [slide] \color default of dreams. User dreams, I am no longer trying to build these things (in case you wonder, that's not because I am retired, that's because I am trying to build other things). \layout Standard I would like to see a set of free software modules, for various music content processing, indexing & retrieval, and interaction functionality, with a basic framework for combining them into practical chains of usage. I know various people are working on it, both in this community and outside it, so I have at least one realistic dream. One challenge here is that we need a framework for combining them, but this framework must be minimalistic. If the complexity of understanding the framework, of adapting a module to it is too high, we will lose the benefits of wide cooperation, of explorator y usage. Another challenge is that we need to get rid of any notion that the algorithms or the functionality in these modules could be patented. This may be obtained in the European legal framework, though it is a hard fight. \layout Standard Considering the limited resources that are available to build large-scale MIR systems and experiment with them, it is necessary to focus on music that be accessed, exchanged and re-used without high transaction costs. Public domain (when there is one), voluntarily created musical commons, and music that can be reused under legal licensing schemes are the only candidates. And the way to start is by constructing music commons. Once again, there are people in your community who work on it, but they are faced with many obstacles. So I dream that there would be more of a community-wide approach at it. \layout Standard The tragedy of the present repressive approach to P2P file sharing is that it prevents it from maturing by putting it under siege, and then uses this immaturity to further justify warfare. Installing P2P file sharing as a respected social and technical paradigm for archiving and reusing music is a very valuable aim, to which the MIR community can, it it wishes, contribute key elements. This is closely linked to the previous point (musical commons). If you love fingerprinting, use it to control injection instead of controlling usage, for instance control that no fakes are injected. Develop new schemes of connecting contextual metadata with P2P access to raw music (by the way don't forget to look at what the P2P users are already doing in that respect). Stand for attribution, for making available information about usage rights without trying to force or monitor their respect by individual users by technical means. Work with people who construct P2P services with added-value information for media commons, such as the Réseau National d'Echanges that Tariq Krim is trying to create in France. \layout Standard Finally, the retrieval paradigm however valuable it is may be too narrow. There is more to music information than retrieval, as can be seen from this conference's topics. \layout Standard \color red [Slide] \color default In case you share any of these dreams, don't believe somebody else will remove the obstacles. \layout Subsection \emph on Thank you \layout Standard After many potentially contentious statements, to install back some serenity, or to finish destroying it, I propose you a very short passage in music before questions, easy listening music that illustrates how the genre category is suffering a bit. Thank you for your attention. \the_end