This is cool and reminds me of Dr. Adrian Thompson's pioneering experiment with evolutionary techniques on an FPGA [1][2].<p>He evolved a tone discriminator on real world hardware, and after a few thousand generations the resulting circuit was one no engineer would ever imagine - it made use of transistors operating outside their saturation region, and subtle secondary magnetic or PSU-line effects from nearby gates not even connected to the logic pathways. But it was effective and amazingly space-efficient.<p>Physics in the real world offers an incredibly rich and vast set of variables for evolution to play with, and I feel like our attempts to simulate it in software constructs may be too limiting to yield results approaching AGI.<p>[1] Article: <a href="http://www.damninteresting.com/on-the-origin-of-circuits/" rel="nofollow">http://www.damninteresting.com/on-the-origin-of-circuits/</a><p>[2] Paper: <a href="https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.50.9691&rep=rep1&type=pdf" rel="nofollow">https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.50...</a>
That reminds me of the Golem project - <a href="http://demo.cs.brandeis.edu/golem/" rel="nofollow">http://demo.cs.brandeis.edu/golem/</a><p>I remember running that as a screen saver back in the day.
Y’all might also enjoy the GOLEM project: <a href="http://www.demo.cs.brandeis.edu/golem/" rel="nofollow">http://www.demo.cs.brandeis.edu/golem/</a><p>It’s of a similar age and goals, but evolved physical creatures with a screensaver.<p>I ran that screensaver as a middle schooler and happened to work with Prof. Lipson at the Computational Synthesis Lab in college.<p>Like other posters here, I feel genetic algorithms got a bit overshadowed by neural networks. Circa 2010, GA’s were capable of some real feats that still seem cutting edge today: deriving the full set of differential equations of metabolism for a bacteria, self-modeling through exploration, finding fundamental laws of physics by watching a double pendulum video, and more.<p>A ton of good came out of that lab, including a big part of modern open-source 3d printing (which was originally pursued to print the multi-material GOLEM robots!)
“results from a research project involving simulated Darwinian evolutions of virtual block creatures”<p>It would be interesting to run this software today… probably would run on a phone, now.
Jesus Christ. Just think of what those people back in 1994 must have expected from us here in 2022. They would be utterly disappointed, if only they remembered.
How were the creatures actually 'evolved' or even 'born' in the first place?
What biological rules were enforced over their development?
How did they learn to overcome obstacles and opponents?<p>This is very interesting, and especially intriguing since this is a 1994 video and I haven't seen any modern examples of this, which would be equally interesting.