Making Music and Art Through Machine Learning, How Google is making music with artificial intelligence


About Magenta

Magenta is a project of Google Brain Project, which asks questions and answers them: “Can we use machine learning to create attractive art and music? If so, how? If not, why not? “Our TensorFlow work has been completed, and we regularly release our models and equipment in open source. These demonstrations with the educational blog and technical documentation. To follow our progress, check out our Github and join our discussion group.

Magenta includes two goals: it is the first research project to promote the art of the state in the field of creating music, video, images, and text. Much has been done with studying the machine to understand the content – for example, speech recognition and translation; In this project, we discover content creation and creativity. Secondly, Magenta artists symbolically create a community of researchers studying the machine. To make things easier, the Core Magenta team creates an open source infrastructure around TensorFlow to create art and music, it already includes tools for working with data formats, such as MIDI, and expanding these platforms. Helps artists join the machine learning model.

Whatever you now find weird, ugly, uncomfortable, and nasty about a new medium will surely become its signature. CD distortion, the jitteriness of digital video, the crap sound of 8-bit, all of these will be cherished and emulated as soon as they can be avoided. It’s the sound of failure. So much modern art is the sound of things going out of control. Out of a medium, pushing to its limits and breaking apart.

Doug Eck is Research scientist at Google

How does Magenta compose music?

Learning Keys We are not spending any efforts on the classical Ai approach, which makes intelligence using the rules. We have tried various machine-learning techniques, including recurrent neural networks, mating nervous network, diversified methods, anti-training methods and learning reinforcement. All these buzzwords understood is too much for a small answer, what can I say that they have all the different techniques to learn from the example to create some new things.

What examples does Magenta learn from?

We have prepared the algorithm NSynth, which uses neural networks to synthesize new sounds, on notes created from different devices. The SketchRnn algorithm has been trained in millions of images from our Quick, Draw! Game Our newest musical algorithm, the RNN performance was trained on a modern piano playing on a classical piano performance [listen]. I would like the musicians to easily teach the models their musical compositions, Then enjoy the music and improve it

How has computer composition changed over the years?

Currently, the focus is on a focused algorithm that learns by example, for example, by learning a machine, instead of using hard-coded rules. I also believe that interchangeable technologies such as our work and computers as helpers for human creativity, and not a “dad’s car” from Sony (a computer-written composition inspired by The Beatles and created by a man-producer) Computers have focused on using it. ,

Do the results of computer-generated music ever surprise you?

Yes. All the time I was very surprised that the recent performances by Ian Simon and Sagev Ore were so much from the algorithms of RNN. As they were trained in demo demonstrations shot in Midi on Disclive Piano, their model was able to produce sequences with real-time and mobility.

What else is Magenta doing?

Previously, we spent summer internships on jokes, but we did not cause ridiculous jokes. We are also working on creating images and creating drawings. In the future, I want to see more in areas related to design. Can we provide tools for architects or web developers?

How do you respond to art that you know comes from a computer?

When I was at the Computer Science Department at the University of Montreal (in Canada), I heard some kind of computer music from the music teacher Jean Pitch. He wrote a program that could create some kind of music, like jazz pianist Keith Jarrett. It was not as attractive as the real Kate Gerrat! But I still enjoyed it, because programming algorithms is just a creative task, I think that knowing the genes and credit for this quiet program made me more responsive, because I would otherwise be.

If once the recognizable abilities can be backed up by an algorithm, do we have to think about them differently?

I think differently about chess, that these machines can play well. But I do not think that a computer playing chess devalues. People still love to play! And computer chess has become an excellent tool for learning. In addition, I think that it is relatively interesting and how chess owners make games, unlike how computers find problems – visualization and experience against brute force, for example.

How can people and machines be more creative?

I think this is a process of repetition, every new technology that made a difference in art took some time to take some time. I like Magenta as an electric guitar, with the goal of being sufficient enough to compete with the guitar of Rickenbacker and Gibson with other devices. Jimmy Hendrix and Johnny Mitchell, Mark Ribot and St. Vincent and one thousand other guitars pushed the envelope to how you can play the game, these instruments were used incorrectly, some said – rebuilding, Distorted, bending threads, lower, paddle paddles, etc. No matter how quickly the machine learns from the point of view of producing models, the artist will work faster to advance the boundaries of what is possible.

- Advertisement -

- Advertisement -