FellKlang Musik Technik
Facebook Google+ LinkedIn New MySpace Pinterest
YouTube Twitter SoundCloud ReverbNation Doktor Fell

27 October 2014


Last August, Tesla chief executive and PayPal payment system pioneer Elon Musk tweeted his view that artificial intelligence (AI) is “potentially more dangerous than nukes”. And, in a recent address to students at MIT, likened it to “summoning the demon”. Such doom-mongering is all very well, but we musicians have more than just songs of the apocalypse to write in the meantime. Could AI assist us in this, or will we all end up miffed with the machine?

FellKlang collaborator Roland Trimmel (left) of Re-Compose, developer of AI-esque songwriting assistant Liquid Notes, echoes Elon’s concern: “Finally, we see the rise of the machines and with it develops a fear that artificial intelligence will render humans useless.” However, having ruminated long on the potential for a robotic Rapture, Roland suggests answers to some of the Qs at large in Mass...
"Is the world ready for the perceptional battle that AI in music poses?" This question was posed at Boston’s A3E Conference last month by a team member at Landr. The company had received death threats from people in the mastering industry after it released a DIY drag-and-drop instant online mastering service powered by AI algorithms.

It illustrates the resistance that the world of AI has incited amongst us. Some fear that robots will take over à la Terminator 2. Some fear that the virtual and artificial will replace the visceral. Some cite religious views, and others? Frankly, others just seem ignorant. That sets the tone for our journey into artificial intelligence and the lessons we have learned from it.

Fast rewind to late 2011 when we launched our first product, Liquid Notes, a songwriting assistant driven by intelligent music algorithms. We had spent more than three years developing algorithms to enable software to read and interpret a composition like an expert does. Coming from a music and technology background, our team was hugely excited to have accomplished this.

Make no mistake, it’s really difficult to make a computer understand music . For us, this was an important first-step towards a new generation of intelligent music instruments that assist the user in the songwriting process for faster completion of complex tasks, resulting in no interruption of the creative flow and more creative output. When you spend so many years working on a technology product, you run the risk of losing sight of the market. And, this being our first product, we had absolutely no idea what to expect.

To find out, we had to bring Liquid Notes to the attention of the target group and eagerly awaited their reaction. It meant a lot of leg work for us, starting discussions on multiple forums and collecting users’ feedback. It takes time to cut through the noise, but it does create some great threads.

It was interesting to monitor was how the discussions about our product unfolded on those forums and how opinions were split between two camps: one that embraced what we do, and the other that was characterised by anger, fear or a complete misunderstanding of what our software does.

“This is like suggesting that the World Series could be won with a pitching machine,” sums up one attitude. At times we felt like being in the middle of the fight between machines and humans. We hadn't expected this; our aim was to make a cool product that shows what the technology is capable of.

Eventually, we spent lots of time clearing up misunderstandings, explaining our product better to win over those forum members’ hearts. And occasionally we had to calm down heated discussion between members who were insulting each other through fear that our product eliminates the craft in music composition.

We’ve made a lot of progress with the software, much of it down to communicating openly with our community so as to address immediate questions they may have, and to involve them deeply in product development. But has the tone in discussion about our technology changed?

Yes, certainly it has. But please don’t think it’s an easy journey. It’s still hard to convince music producers to rely on the help of a piece of software that, in some regards, replicates certain processes of the human brain. The efforts that go into being a pioneer and driving this perceptional battle take one close to insanity. It’s an endless stream of work and it requires endurance. Below are five things we learnt from our journey that I’d like to share so you can judge better before dismissing AI in music.

But first, let’s have a quick discussion of the first and second digital waves in music: Understanding the perceptional battle in music...

The first digital wave brought about such music technology as soft synths and DAWs. And with that, everything changed. Sound synthesis and sampling made entirely new forms of expressiveness possible. Sequencers, in combination with large databases of looping clips, laid the foundation for electronic dance music which led to a multifaceted artistic and cultural revolution.

The second digital wave has been rolling along for a few years now and is washing up intelligent algorithms for processing audio and MIDI. As an example, AIs can already help control the mastering process of music tracks, either as assistant tools, or as fully automated processes. In the not-too-distant future - and we’re talking years, not decades, from now - we will be accustomed to incredible music-making automatons controlling the most complex harmonic figures, flawlessly imitating the great artists. The output quality of such algorithms is unbelievable. Computer intelligence can merge styles of various artists and apply them to yet another piece without breaking a sweat.

We regard the main application of AIs for music composition and production as helper tools, not artists in their own regard. And this is not cheating. We have been utilising digital production tools for decades. It was just a matter of time before more complicated and intelligent code began to emerge.

However, rest assured that computers will not generate music all by themselves. The art and craft of composing will prevail.

There will always be human beings behind the actual output that's under the control of an artificial intelligence. Such technology will help, though, to create less complex, leaner user interfaces that are simpler to operate. And so to the answers I promised you earlier...

1. Will AI render human music makers useless?
Definitely not. The final decisions over creative output will always remain with the (human) artist because, quite simply, a computer is not a human with feelings and emotions. What makes us get to our knees in awe will leave machines clinically indifferent. Technical approximations, as deceptive as they may get, are simply not the real thing.

2. Is AI in music tech becoming a reality?
It already is. There is no stopping it. But then that is the course of a natural evolutionary process which can only push forward.

3. What impact will AI technology have on the music industry?
A huge one. This is a game changer! See the statement on AI's main applications above.

4. Why does this topic cause such strong emotional reactions?
We cling to our egos, having trotted down the same paths for decades. Many believe that laboriously acquired expertise is threatened by robot technology; by a new, ruthless generation. The truth is, if we embrace AIs as our helping friends and maybe even learn how to think a little more technically, think how much more colourful the world of music will become in the hands of talented musicians of all generations.

5. Will the industry embrace this change?
Yes, because it enables a completely new generation of products and startups like us pushing for innovation. The agreeable side effect: It will make people happy; musicians, consumers, and businessmen alike, full circle.

Most importantly, it's not only AI that is changing the music industry. Social changes are equally responsible for change, perhaps even a larger part of it. Here’s an excellent article by Fast Company on the topic, and more coverage on A3E in this article by TechRepublic.

It’s an interesting time for all of us in music and beyond, and there’s so much yet to come. Don’t be afraid — humans also prevailed in Terminator: “There are things machines will never do. They cannot possess faith, they cannot commune with God, they cannot appreciate beauty, they cannot create art. If they ever learn these things, they won’t have to destroy us. They’ll be us.” - Sarah Connor.
Roland Trimmel, FellKlang collaborator, Re-Compose
- - - - - - -