assignment 4

How can machine learning support people's existing creative practices? Expand people's creative capabilities? Personally, I believe creativity fundamentally relies on utilizing chaos (as perceived by us), and machine learning generally in terms of scale is chaotic. More literally, like what Rebecca Fiebrink talks about, machine learning has a lot of practical uses. It connects people and their ideas by synthesizing massive amounts of information. It can be used to generate new, complex ideas based entirely in existing data.

Dream up and design the inputs and outputs of a real-time machine learning system for interaction and audio/visual performance. This could be an idea well beyond the scope of what you can do in a weekly exercise: I'd like to see a public installation that records people dancing. The input is everyone who has danced in front of it. The first user will be the first input. Then, a large display shows the user a generated figure dancing based on all previous inputs, layered over their own video. Obviously, initially there would probably be major issues due to sample size, but as the art piece ages, each new person who interacts with the piece simultaneously dances with a synthesized version all previous users, while also training it to dance a little more like them.

In my p5 sketch, I improved upon our class example by using lerp to give the circles some ease. I only eased the draw of the circles though, not the input into the machine, since I figured it is optimized to be trained in a certain way. It took a lot longer than anticipated to implement the easing property though. Usually, it is very straight forward, but after many attempts I realized that the thumbX, thumbY, etc. variables had to be given a value of 0 at the beginning of the sketch, or else it won't work. Also declaring the value of two variables (let ex, ey = 0;) doesn't work.

I then implemented a new stage of the sketch where after it is trained, users can draw when their fingers are open. There are definetely still small things to be fixed in this, and it is not a very useful drawing tool, but with better input data and processing power, I could see something like this being useful in any situation where a user is relying on hand gestures to trigger events.

training drawing