The space of possible bodies was encoded into a latent design “genome”.
Evolution was inspired by Bayesian rather than Darwinian theory.
08.03.2026 17:48
👍 3
🔁 0
💬 0
📌 0
The space of possible bodies was encoded into a latent design “genome”.
Evolution was inspired by Bayesian rather than Darwinian theory.
Here’s another favorite of mine. This one has a less familiar form.
It evolved to move on land but really wants to try swimming someday.
Here is one of my favorite creatures that evolved.
We call it “the seal” (aka Lucille).
Here’s a peak inside the module:
Here they are showing off some of their acrobatic behaviors, resisting adversarial perturbations, and withstanding body-altering damage.
(Control policies use internal sensing only; motion capture was not supplied to the policy, it was used only for behavioral analyses.)
These are legged metamachines, legged machines made of legged machines…
Website: modularlegs.github.io
Paper: www.pnas.org/doi/10.1073/...
Always a blast introducing new folks to the history + future of #biohybrid #robotics - today at the University of Bristol SoftLab Seminar! Great to get feedback on what we've done so far + an exciting reminder of how much we have left to keep us busy in the coming decades 😊
Thanks Josh! It truly would not have been possible without you.
The Genomic Code: the genome instantiates a generative model of the organism www.cell.com/trends/genet... - really delighted to see this in print in @cp-trendsgenetics.bsky.social! 😊
Our new neuroscience paper - hot off the press! Check it out! 🧠🧠🧠💻💻💻
Today's evolutionary #robotics lecture: bipedal locomotion.
In which John Cleese and yours truly demonstrate several silly walks.
youtu.be/4o9CZK8Azcw
@drmichaellevin.bsky.social presents to the United Nations on how xenobots (Doug Blackiston & @samkriegman.com), anthrobots (Gizem Gumuskaya), and other discoveries from his lab can serve as "agential materials" that AI can employ to interface with the body.
www.youtube.com/live/T40utCZ...
New paper w/ Luke Strgar
sites.google.com/view/co-desi...
...in a nutshell:
⛰️ differentiable physics
🏋️ morphological pretraining
🧠 a universal controller
🧬 zero shot evolution
💡 adaptive phototaxis
☠️ diversity collapse
🔧 generational finetuning
🦘 few shot evolution
Muhan Li will present his freeform endoskeletal robots at ICLR in April:
endoskeletal.github.io
Evolve yours at home today!
