2/28/2026 at 2:21:49 PM
> Werld drops 30 agents onto a graph with NEAT neural networks that evolve their own topology, 64 sensory channels, continuous motor effectors, and 29 heritable genome traits. communication bandwidth, memory decay, aggression vs cooperation — all evolvable. No hardcoded behaviours, no reward functions. - they could evolve in any direction.In the days when Sussman was a novice, Minsky once came to him as he sat hacking at the PDP-6.
"What are you doing?", asked Minsky.
"I am training a randomly wired neural net to play Tic-tac-toe", Sussman replied.
"Why is the net wired randomly?", asked Minsky.
"I do not want it to have any preconceptions of how to play", Sussman said.
Minsky then shut his eyes.
"Why do you close your eyes?" Sussman asked his teacher.
"So that the room will be empty."
At that moment, Sussman was enlightened.
by gyomu
2/28/2026 at 3:12:59 PM
Love the MIT AI Koans. Minsky's actual words to Sussman were "well, it has them, it's just that you don't know what they are." And he's right, the room isn't empty.Werld's room has walls. The graph topology, energy mechanics, metabolic costs, seasons, those are all design choices. But those are the physics, not the behavior. I chose the laws of nature, not what agents do with them.
Whether they cooperate or attack, broadcast or stay silent, grow complex brains or prune them down, that's selection, not me.
The agents also aren't randomly wired like Sussman's net — they start with minimal NEAT networks and evolve structure through survival. So the preconceptions are there, I just tried to make them physics rather than policy.
Curious how you would approach removing those from an artificial sim like this?
by urav
2/28/2026 at 9:05:09 PM
some hand-wavy thoughts..> it has them (preconceptions), it's just that you don't know what they are.
further in this direction... the "thing" might evolve into some cyclic (or not) system. a bit like that LIFE game, emerging a tv-tennis-like ping-ponging, or whatever. How would you know there is such thing? just stats/counts do not tell. (Which pulls another freaky question - how would u notice a different intelligence/world-order/culture/resemblance-of?)
maybe feature: some stop-gap animation over world-map in time? Then, some pattern analysis over that.. History of the world, part one..
btw check these "interactive simulations", maybe some ideas about "loading" the agents with preconceptions :)
by svilen_dobrev
3/1/2026 at 11:25:36 AM
thing you've caught onto something there, after running it for a while (maybe not long enough) it can feel cyclical.I'll definitely add in the time-lapse, and some sort of a pattern detection over the agent positions/actions - shouldn't be too tough given the graph structure of the world.
for the ncase.me - super interesting way to visualise it - the polyworld example someone gave before also had a 'view' into their worlds. on the preconceptions, maybe running two parallel experiments and comparing the outputs might work best? thanks for the pointers - let me know if you've got any ideas on how to approach.
by urav