I like watching computers figuring things out.
For example, the Travelling Salesman Problem (TSP) has fascinated me since I was a teenager and first heard about it. At the time, massively parallel computing was going to be the solution. But they fell apart (late 1980s) and gave way to connectionism / neural networking – many of which can’t be parallelized (some can, “deep learning”), which then transformed into genetic algorithms, many of which can be parallelized.
[in short] “Genetic algorithms usually perform well on discrete data, whereas neural networks usually perform efficiently on continuous data.”
So, I wanted to find a visual travelling salesman solver that _could_ run in parallel which means a genetic algorithm and I found one here: https://github.com/bezzad/TSP – and deep in the src data is a compiled Windows version under debug, which I’m running here using 125 cities I made by doing a loose spiral that bounces off the walls.
The music is copyright-free from Scott Buckley, who I just discovered when looking for a 6 minute piece to fit the length of my run here, and has a copyright-free music channel on Youtube, this piece is called Signal to Noise and captures the feeling.