Google-owned defense contractor Boston Dynamics built a new, freaky robot dog called "Spot". It's no doubt a pinnacle of autonomous robotics thus far...


There's the obvious immediate thoughts of "whoa, cool!" and "OMG we're gonna die at the hands (or paws) of robots", but for me, right now, the interesting piece of this technology is the realization of machine learning and genetic algorithms that surely went into the software, and where that connects to the evolving hardware side. Thanks to the open source and maker movements, it's easier than ever to create our own smart software and hardware inventions.

Google-owned defense contractor - cacheflowe | ello

Just last week I was running a biped walker demo (made by the same person who built the genetic car algorithm a while back), that uses a 2d physics library to evolve a simple humanoid figure to learn how to walk. The above gif is the result of thousands of iterations (I let it run in my browser for about 24 hours). Each new generation of walkers will inherit traits from the most successful walkers of the previous generation - a simple but effective act of evolution.


I'm especially interested in where this technology overlaps into the world of art. Take Karl Sims' "Evolved Virtual Creatures" - an evolution simulation from 1994. Karl was a pioneer in modern computer graphics algorithms and created mind-bending, trippy animations with a lot of the same physical simulation/modeling code that went into these genetic algorithm studies. The many ways that computer code can model nature in such an authentic fashion is a constant inspiration to me. The universe abides by the same rules that our machines do, just at a different resolution for now...

Related Posts