Home
Warning: there is text further to the right, so keep scrolling!
I'm going to flesh this section out with the things one can actually do with floatworld.
In the mean time, the README on the home page has about as much information as I have time for right now.
However, here are four screenshots to tantalize, showing the floatworld environment as it currently stands. Videos to come later.
First we have the dialog that creates a new world, specifying what mixture of different energy sources and objects to use, what various global parameters should be, and what the initial Creat neural network looks like.
Here is the main environment, which loading, saving, running the simulation at various speeds, adjusting the parameters of the entire world of various Creats or objects within it, and so on:
This second example demonstrates some curious 'frogs' that have evolved to occupy small circular portions of territory and occasionally 'hop' over each other when they want to reproduce.
The third example shows one of the many types of object in action. These are Skinner boxes, which reward Creats with a burst of energy if they are 'tapped on' a certain number of times. Here Creats have rapidly evolved to supplement the meagre supply of energy that I provided by triggering these boxes. Creats must alternate between tapping the box and collecting the reward, so the behavior is not merely a hardwiring between a vision neuron and the tap neuron.
There are many more such examples of fascinating behavior that can emerge, but it is not my priority to document them now.
I did wrote a short, amateurish paper on floatworld for my undergraduate degree a few years ago. It's totally out of date, but if you're don't mind the typos and weird pacing (I had to pretend it was a Pure Mathematics project) you can find it in the wiki repository (first page shown below) or here.