I Turned My Phone Into a Robot | Art of the Problem

The Gist

Old smartphones make surprisingly capable robot brains because they contain high-end processors, batteries, and sensors that outperform custom builds. By connecting an Android phone to a simple carrier board with two servos, anyone can build a fully autonomous AI creature for under one hundred dollars.

Quick Overview

Old smartphones serve as the brains for autonomous AI robots that walk, talk, and interact with the physical world. After a previous build using custom sensors sparked massive demand, the creator pivoted to recycling discarded phones inside pizza boxes and custom 3D printed stands. This approach gives everyday users open source access to advanced neural networks and onboard sensor logging.

Key Points: Old smartphones contain better chips and sensors than most custom-built hobbyist robots can achieve. The creator spent over a year building the first prototype from scratch before realizing phones solve the hardware bottleneck. A simple setup requires an Android phone, a carrier board with a microcontroller, and two servo motors for the legs. The Growbot web application allows users to run open source language models and store experience logs directly on the device. Over 10,000 people requested kits or finished robots after the initial viral video launch. The system costs under one hundred dollars and can be built in less than an hour using readily available parts. The robot operates completely autonomously, sensing its environment through the phone and executing movement policies.

Context: Following the viral success of a project that gave ChatGPT a physical body, the creator faced overwhelming demand for robot kits during a global chip shortage. Instead of manufacturing expensive hardware from scratch, the project evolved into an accessible open source initiative that transforms discarded electronics into living AI creatures.

Detailed Analysis

The project begins with the realization that old smartphones sitting in drawers possess superior computing power and sensor arrays compared to expensive custom robotics kits. The creator devises a low-cost method using a microcontroller, a carrier board, and two hobby servos mounted to a 3D printed frame or even a pizza box. The phone runs a custom web application that streams sensor data to a large language model backend, translating environmental inputs into physical movements. This setup democratizes robotics by allowing anyone to download the software, inspect the neural network code, and retrain the robot without corporate restrictions. The video details the evolution from a painstaking custom build to a universal smartphone platform designed for widespread home use.