A nano-factory in a box! 🥡
A system called MicroFactory has been shown autonomously assembling a photo frame, an actual product sold on Amazon.
Instead of relying on a massive general AI model, the team broke the job into 27 discrete actions: 23 recorded trajectories and 4 small, fine-tuned AI models for steps involving uncertainty.
When motion is predictable, like flipping a frame in a holder, the robot simply replays a recorded path.
When variability appears, like peeling off a protective film, the AI model takes over.
It’s a simple but powerful approach: AI for randomness, recorded motion for repetition. Together, they handle the full assembly reliably and at low cost.
Hybrid methods like this can close today’s capability gap, collect high-quality data, and make automation economically viable.