Composing Jazz Music with Artificial Intelligence - venkat k - Medium

Views:
 
     
 

Presentation Description

The human audience is laughing happily in their seats, and the musicians are undeniably flesh and blood, sweating and straining at their instruments.

Comments

Presentation Transcript

slide 1:

9/16/2019 Composing Jazz Music with Artificial Intelligence - venkat k - Medium https://medium.com/venkat34.k/composing-jazz-music-with-artificial-intelligence-23b64689885f 1/3 Composing Jazz Music with Articial Intelligence venkat k Sep 16 · 4 min read It was Wednesday night in North East London and upstairs machines at the Vortex Jazz Club were calling the shots. The human audience is laughing happily in their seats and the musicians are undeniably flesh and blood sweating and straining at their instruments. But music is the product of electronic brains — trained to soak up great artists ’music and filter new melodies. It was “the first concert with music composed by Artificial Intelligence” said Professor Gerent Wiggins of Queen Mary University at the beginning of the evening. In a few minutes we hear medieval hymns baroque chorals and jazz and pop — all made with artificial intelligence with the help of computer scientists programmed by “composers” in the evening. There was a tremendous buzzing in the room while reading the list of Wiggins helpers. The atmosphere is like school reading No wonder that most of the

slide 2:

9/16/2019 Composing Jazz Music with Artificial Intelligence - venkat k - Medium https://medium.com/venkat34.k/composing-jazz-music-with-artificial-intelligence-23b64689885f 2/3 audience members are computer scientists it is interesting to see how their offspring work — as well as predicting competition. To facilitate that verdict Wiggins said evening music can be in the style of familiar styles or composers. “If we produce computer-to-computer music who’s to say it’s any good” He asks as we settle into the tight rows of French bistro chairs. “We’ve been trying to emulate styles for ages so you can identify them and tell us what they are good at.” Another unspecified reason for this mimicry is that it fits well with the current capabilities of AI. Much of the music of the concert is the product of deep learning — a kind of artificial intelligence that tech companies like Google and Facebook have a profound influence on such as speech and facial recognition. Deep learning is great at separating samples from large libraries of data and then labeling that information for example it’s a cat it’s human and so on or creating completely new data that matches what we’ve seen before. In the case of these nightly concerts the researchers fed sheet music of their deep learning systems from specific composers and time periods. Machines analyze repetitive patterns — a sequence of notes harmonic sequences and so on — and create something similar. You may think of it as a photocopier with unruly nation. You feed in your latest report from work and at the other end it looks broadly identical but for some reason talking about a non-existent company and citing unattached sales figures. You could argue that this isn’t really considered composition and that computers just like humans. But what the scientists at the concert said Can you really say that any artist is without preference Are all creative actions the sum of their effects Shut up and let my kid play. At night I could see their point. The jazz combo is like the original thing. I won’t go so far as to say that things are heating up but there is also a slight tingling of toes drinks and a little bit of hooping from the assembled scientists. The first AI-written song to be played using software developed by Sony Computer Labs in Paris. The AI’s contribution was just a lead sheet and they started running through the theme before entering the range of the solo telling me that they were all human. AI is not a handle structure — it is perfect for jazz Jazz is a style well suited to an AI composer so performance is part of it. Although artificial intelligence systems are great at pattern recognition they are very short- sighted and only work on short sequences of notes. They happily produce some melodic bars but do not understand the elaborate structure of a piece — the way a symphony returns to the central theme for example by repeating the melody with the added

slide 3:

9/16/2019 Composing Jazz Music with Artificial Intelligence - venkat k - Medium https://medium.com/venkat34.k/composing-jazz-music-with-artificial-intelligence-23b64689885f 3/3 tweaks it develops into new territory. When it comes to jazz it’s less of a problem. People don’t expect construction from jazz. During the break he was a scientist for the Austrian Research Institute for Artificial Intelligence who stepped out for a cigarette. He is bullish on the future of AI composers and says they will soon be producing background music for things like advertising. But he also lamented their current inability to comprehend the form of music. “The problem is that these models only learn the statistics of music.” “We’ll give you a chance to be the next note C or F or E but we don’t look too far.” A pair of harmonies created by a deep learning system trained on piano sonatas collected by Mozart — certainly presented the challenge of a structure. The first part of the so- called “Mozart Unchained” show was not recognized as classical music. This is not disagreement but the melody is everywhere As written in the mid-20th century. Articial Intelligence AI Ai Services Ai Solution Machine Learning About Help Legal

authorStream Live Help