Digital Creatures Learn To Walk | Two Minute Papers #8
Why Hacker News recommends it
HN commenters use it as a compact example of simulated creatures learning to walk, whether to question why robots still struggle, to rebut claims about training in virtual environments, or to illustrate evolved gaits.
Recommended by 0 people across 0 different months, 2019 to 2026.
What it is
A short Two Minute Papers episode by Károly Zsolnai-Fehér covering the paper "Flexible Muscle-Based Locomotion for Bipedal Creatures." It shows simulated bipedal creatures with varied body shapes walking using optimized muscle layouts and activations, robust to different speeds and uneven terrain.
What people said
Cites gridlockd · 2019-08 · Hacker News · on “The Computer Maverick Who Modeled the Evolution of Life (2014)”
Ah, so that's the other part of your misconceptions. As far as I can see, the point made in the (first part of) the book is that informed algorithms (on average) always outperform random searches and that genetic algorithms are akin to random searches. The first part is true, the second part isn't. Evolution isn't random, mutation is random. Selection is very much not random, it is informed by the environment. This drastically reduces the size so-called "search space". Using the term "search" for the outcomes of evolution is a curious choice anyway. Evolution doesn't search for anything in particular, it produces adaptation to an environment. It's a process without end. As for computer simul
Cites fxtentacle · 2022-06 · Hacker News · on “By Exploring Virtual Worlds, AI Learns in New Ways”
I feel like they are stretching the word "new". "New studies suggest that training in virtual environments can help robots with these and other skills" Here's a paper from 9 years ago where people train in a virtual environment to help bipedal robots learn to walk: [LINK] And here's a Two Minute papers about it (from 7 years ago): [this video]
Cites amelius · 2026-10 · Hacker News · on “A robot I didn't expect to build”
Can anyone explain that if we had this more than 10 years ago: [this video] we are still struggling with making a robot walk?