Give it a go

In the previous step, you heard from Rebecca Fiebrink who discussed her work with the Wekinator and how it is helping musicians and artists to explore this new technology.

The Wekinator website includes a section of example projects that lists some of the great work that has been done with the software. These example projects are a great source of inspiration for you to come up with your own ideas.

_{Example of a Wekinator model}

Have a look through the projects and familiarise yourself with the workflow. The IML workflow, demonstrated by the Wekinator examples, plays an important part in getting to grips with this form of creative AI. Once you have confidence in the procedure, you will find that you will be able to gain a better understanding of the different algorithms and what they can best be used for.

Another great resource for creating art using machine learning is the MIMIC Project. This is an online resource that works in your browser. The aim of the MIMIC project is to bring new technologies in AI and signal processing together and integrate them into a collaborative online platform.

MIMIC is mainly focused on using machine learning processes in a musical context. The platform also contains several examples of how to integrate systems for sound, music and art making. These examples can be used as inspiration for your own machine learning projects. In order to use the MIMIC platform, you will need to engage in some coding. However, try not to be too daunted by this as the resources available on the platform make it very easy to learn.

Given the examples provided by Rebecca and also the example projects shown on the Wekinator and MIMIC websites, you might be inspired to design your own machine learning system. Where would you begin? There are some fundamental beginning steps in creating your own machine learning system:

1. To start, you might ask yourself, “What form of data do I have access to?” In terms of IML, this would usually be data from a sensor that you can manipulate in real-time.
2. You would then ask, “What output do I want from my system?” Visual output? Audio output? Both?
3. Then, think back to what you learned about regression and classification. Do you want your system to output a continuous stream of data to smoothly control parameters? Or do you want your system to output discrete labels based on the input?
4. Next, you should choose an appropriate algorithm to achieve your desired output. The Wekinator gives you access to a range of algorithms that you can experiment with.
5. Finally, train your model and run it!

From here, you can integrate machine learning algorithms into any part of your practice and experiment with the results.

Have a go

Beginning with the examples on the Wekinator and MIMIC pages, try to find some creative AI projects that inspire you. For more inspiration, also check out the link to aiartonline below.

Use the Comments section to share a project that you particularly like.

• Why do you like it?
• What elements might you draw on in your own design?