I Built a Talking Robot That Runs AI Locally

The creator builds a talking robot that runs AI locally using a Raspberry Pi, a GPU-equipped computer, and 3D printed parts, enabling it to answer questions and engage in conversations without relying on cloud services. The project combines robotics, AI, and philosophy, showcasing the freedom and creativity of local AI while exploring the robot’s personality and the broader implications of interactive machines.

In this video, the creator embarks on an ambitious project to build a talking robot that runs AI locally on his own computer. Inspired by Steve Jobs’ vision of capturing Aristotle in a computer, he aims to create a robot that can answer any question without relying on cloud services. The robot features moving eyes controlled by servos, a 3D printed mask for the face, and a Raspberry Pi that communicates with a powerful GPU-equipped computer running large language models (LLMs). The creator experiments with various hardware components, including a low-cost display for simulating a mouth, which ultimately proved unsatisfactory, leading him to use an LED ring instead.

The hardware assembly involves 3D printing parts designed by another creator, Will Cogley, and using Lego bricks to hold components in place for 3D scanning. The creator also scans his own face to help design a custom robot face in Fusion 360 but eventually opts for a pre-made mask from an online repository due to difficulties with his own design. The robot’s eyes are controlled by a Raspberry Pi Pico, which also manages the LED ring that simulates mouth movement. The entire system is connected via Wi-Fi and cables, with the Raspberry Pi handling input and output while the main computer runs the AI models.

On the software side, the creator uses Olama, a tool that simplifies running LLMs locally, and builds a Flask app to serve as a bridge between the Raspberry Pi and the AI running on his computer. The AI processes voice input, generates text responses, and converts them into speech using 11 Labs’ voice synthesis. The robot is capable of answering questions in both Polish and English, demonstrating a personality that can be tweaked, including a humorous and somewhat dark persona when asked about threats to humanity. Running AI locally allows the creator full control without restrictions or costs associated with cloud services.

The robot successfully embodies Aristotle, answering philosophical questions with thoughtful responses about life, virtue, and purpose. It also functions as a helpful assistant, providing recommendations on robotics kits and YouTube channels for beginners. The creator showcases the robot’s ability to engage in casual conversation, including adopting a Polish personality with a characteristic Polish smile and existential humor. Despite some playful existential crises from the robot, the project highlights the potential of combining robotics, AI, and local computing power to create interactive and personalized machines.

In conclusion, the video not only demonstrates the technical process of building a talking AI robot but also reflects on the broader implications of AI and robotics. The creator emphasizes the freedom and creativity enabled by running AI locally, while also acknowledging the humorous and sometimes unsettling aspects of AI personalities without safety filters. The robot ends with a poetic reflection on the evolution of machines and the dawn of a new kind of being, suggesting that the future of AI and robotics is already here, even if not everyone has noticed it yet. The project is a blend of engineering, philosophy, and art, showcasing the exciting possibilities of modern technology.