Amazon Web Services (AWS) has announced a keyboard that can be used by developers as a way to get started with generative AI. In other words, this means that developers can use the keyboard to create AI-generated pieces of music.

With the generative AI technique used by the keyboard, developers can teach machines to perform creative tasks. This is done using generative adversarial networks, a technique in which two neural networks are juxtaposed, after which they then build a new piece of music by using large sets of (musical) samples.

The keyboard has a range of two octaves and a total of 32 keys. In addition to the physical device, which can simply be connected to a PC, it is also possible to use a virtual keyboard in the AWS cloud, in the DeepComposer console. Developers can train their own models for the device, but the device also works with models provided by AWS.

Creative process

The process starts with the creation of a piece of music by means of a model, after which the user can customise the piece in the console to her/his own preferences. For example, it is possible to choose rock, pop, jazz, classical or a genre adapted to your own preferences. Then the final piece is rolled out. Afterwards it is possible to share musical pieces on Soundcloud.

The intention is to make DeepComposer a pleasant way for users to learn more about machine learning, on the one hand because of the creative approach and on the other hand because the pieces of music invite users to share their results with others. This not only creates a creative process, but also an opportunity to receive feedback on a practical application of machine learning.