Scientists Taught an AI to ‘Evolve’ Robots, Then Built the Weirdest Ones in Real Life

2026-03-19 · Source: AI Archives - VICE · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems · Depth: Advanced, quick

Summary

Researchers at Northwestern University developed an AI system, modeled on natural selection, to autonomously design novel robots, detailed in *Proceedings of the National Academy of Sciences* and on Northwestern's website. The system simulated thousands of simple modular machines, each composed of basic components like limbs, motors, and sensors, competing to move effectively. Through a Darwinian rule set, the AI iteratively refined designs, prioritizing functional performance over human-centric aesthetics, resulting in "metamachines." These AI-evolved robots demonstrated remarkable resilience in real-world tests, capable of self-righting, continuing movement after limb loss, and functioning when split into independent modules, each sensing, computing, and navigating autonomously.

Key takeaway

For AI scientists exploring novel robotic architectures, this research suggests that relinquishing human design biases to an AI-driven evolutionary process can yield highly resilient and functionally optimized robots. You should consider integrating natural selection-inspired algorithms into your design workflows to discover solutions that human intuition might overlook, particularly for applications requiring extreme durability and adaptability in unpredictable environments.

Key insights

AI-driven evolutionary design can create resilient, unconventional robots prioritizing function over human-expected forms.

Principles

• Evolutionary algorithms optimize for function.
• Modularity enhances robotic resilience.

Method

An AI simulation uses a Darwinian rule set to evolve modular robot designs, testing performance, keeping successful configurations, and discarding failures, then repeating the process to create "metamachines."

In practice

• Explore AI for unconventional robot designs.
• Prioritize modularity for robust systems.

Topics

• Evolutionary Robotics
• AI Simulation
• Modular Robotics
• Natural Selection Algorithms
• Robot Resilience

Best for: AI Scientist, AI Researcher, Robotics Engineer, Research Scientist

Related on AIssential

• Robots that refuse to fail: AI evolves 'legged metamachines' that reassemble and withstand injury — Modular robots with athletic intelligence can reconfigure, recover from injury, and maintain movement in dynamic environments.
• Scientists Trapped 1000 AIs in Minecraft. And they Created A Civilization without being told to do so. — AI agents, given minimal prompts, can autonomously develop complex social structures and emergent behaviors.
• How robots learn: A brief, contemporary history — AI models and large datasets are revolutionizing robot learning, enabling adaptation and complex task execution.
• A neural blueprint for human-like intelligence in soft robots — A neural-inspired AI control system enables soft robots to learn once and adapt instantly to diverse tasks and disturbances.
• Video Friday: Humanoid Learns Tennis Skills Playing Humans — Robotics research is advancing dexterous manipulation, sustainable design, and complex locomotion through AI and novel hardware.
• Learning of Robot Safety Policies via Adversarial Synthetic Scenarios — An adversarial gamification framework enables robust robot safety policy learning by iteratively discovering high-risk edge cases.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI Archives - VICE.