Hackaday · Creativity & design

Digital Pocket Fish Tank Uses LLM for Fish AI

A maker has developed a pocket-sized digital fish tank that simulates fish behavior using a custom-trained large language model (LLM) running on an ESP32 microcontroller.

The device, called Pocket Tank, features a virtual aquarium on a 1.8-inch AMOLED screen. It utilizes a Waveshare ESP32-S3-Touch-AMOLED-1.8 board for its processing power.

Fish movement and animations are rendered at 25-30 frames per second by standard code. The AI controlling the fish's decisions is a condensed LLM that runs directly on the ESP32.

The LLM observes the virtual environment and adjusts the fish's goals based on in-tank events. Users must perform maintenance tasks like feeding and cleaning to keep the virtual fish alive, similar to a Tamagotchi.

The project's creator has shared details on a blog and YouTube, with files available on GitHub for those interested in replicating or modifying the design.

AI-samenvatting op basis van de bron.