Our “Member Of The Week” this week is Stephan Baumann, an active musician and sound artist, passionate about modular techno and interactive sound installations and Germany’s best informed and probably most experienced researcher on AI for music.
For over 17 years now you are working as a researcher at German Research Center for Artificial Intelligence(DFKI) to identify and forecast tech and cultural trends.
For our readers: You have scientifically explored reviews for fitness tracking devices, social data for tourist mapping, worked on a trend radar, developed a hybrid recommendation systems for streaming services and much, much more. All of them in touch with the latest trends and thrilling themes. But what was your personal highlight, your most exciting project so far and why?
I did my Ph.D on the subject of music recommendation systems. Between 2002-2004 the ISMIR community on music retrieval achieved amazing breakthrus regarding content-based analysis of music, collaborative approaches which is the basis for todays methods used by Spotify, etc. I met amazing researchers (the guys of Echonest, Francois Pachet of SONY CSL) and had the chance to do a research in residence at IRCAM. My recommender MPEER based on sound similarity, cultural metadata and similarity of lyrics was tested with user studies and I had to learn that “similarity” is not alone the basic ingredient for a good recommendation, but novelty and surprise and a bit of anomaly have to be taken into account, too.
If you are allowed to communicate, what’s your current challenge?
Over the years I expanded my expertise and started to work with sensor data, biophysical stuff, applications in sports, etc. We use Deep Learning networks for time series analysis and will try to apply these methods while observing creative people and in parallel to teach computers to produce arty artefacts. The most interesting question is if there is creativity “in the black box” and where we can locate such effects.
One of your area of expertise is AI. Recently, AI has helped write pop ballads or imitate styles of great painters. And there is a debate out there about creativity and the potential of AI, about will it match one day with human creativity and become a true creative partner or even the creator of solo works of art. What do you mean as expert, can AI learn to be creative? Will it ever be able to work without a human?
As I mentioned above this is really the most exciting field of research in AI these days. If the training of deep networks (GANs,VAEs, CANs) is undertaken with the right amount of data and structure these networks can “copy” art in certain styles and genres. Even human beings have hard times to distinguish between man and machine-made art this way. But what does this mean? As human beings are kean on really new stuff, blends of genres, the real genuine act behind the crafting of art we have to consider these processes. As an artist and musician I have a certain skepsis because I see no desire and strong will in the machine to create for such reasons and to cope with life in general. The desire to communicate about love and pain, the need to share experience with other human beings thru art is something which is not “in the machine per-se”. For this reason humans are needed to create data about such issues for the machine or at least to code such mechanisms into the future creative algorithms.
What’s your opinion: will AI be a collaborator and an accelerant or a disruptor for human creativity?
AI will be a powerful collaborator and for the creation of music it is already used by artists such as Holly Herndon. She and other artists are using the AI to create new variations of their own voice, synthetic sounds, etc. And some other artists use AI to support the creation of basic song structure, melodies, chord progressions as ingredients. Therefore AI is for these artists accelerating the creative process. It’s a tool, very similar to other music technology such as drum computer, sampler, DAW, etc.It has the power to create MUZAK and toy music but I believe that the greatest artists will “disrupt AI” and not vice versa. There will be very creative uses of AI by the most talented artists – if you look at music history: TR808 ergo house, sampling ergo hip hop.
Who or what influences your music?
I grew up with disco, funk, soul, Jean Michel Jarre, Giorgio Moroder, … In parallel I had organ and piano lessons, my first synth was a Korg MS10 at the age of 16, over the years I had them all: Juno, DX7, D50, Ensoniq, … But I still love to play the Fender Rhodes, a grand piano is unbeatable, I felt in love with modular systems 4 years ago since they are meanwhile affordable. I did a little project (MODSICH) with my personal all time fav keyboard player Daniel Nentwig of THE WHITEST BOY ALIVE, and I have to say he was a huge impact, catalysator and is meanwhile a friend.
Would you have any music recommendation for our community?
MEZERG – This one is really fresh! I found him yesterday over facebook via Mixmag, he is SO COOL! French do it better! He is funky and he seems to be nerdy … love those guys.
In addition to all these areas of expertise above, you are also involved in several spin-offs from the institutional environment and advises technology-intensive startups across Europe. What’s your tip for our community to motivate them to found their own venture? And which advice would you share with young startup founders?
I founded 3 times and it was always a complete different experience. To sum it up: it is an unbeatable experience if you wake up in the morning full of energy to check into YOUR office, open the door and feeling that this work place and topics exist only because YOU CREATED IT! This is really such a great feeling.
For the more rational and pragmatic tips:
Avoid fake experts, people consuming your time and money – which you don’t really have a lot of in the beginning- . Instead of trusting in such experts try to work with people who went thru the process of founding on their own. And for the service, product, experience whatevery your business might be .. prototype early!!!
What 3 tools, resources, other apps etc. do you use daily?
HORST – this is our own semantic music recommender
FACEBOOK – sorry for this ☺ but it is indeed a way to find events, musicians, things going on .. a bit lame I know.
MODULARGRID – the place/gearporn where I configure several times a week my own modular system virtually and if budget allows go for new gear.