The Rise and Fall of AI-Powered Plant Care
Many people struggle to keep their houseplants alive, despite following care guides. [Liam], no exception, decided to turn to AI for help. The project, dubbed ‘PlantMom,’ aimed to leverage a Large Language Model (LLM) to care for a chili plant by monitoring parameters like light, temperature, and soil moisture, and controlling a grow light and water pump accordingly.
To make this work, [Liam] had to combine the LLM with hardware sensors and actuators. The LLM was given extensive prompts, including Python code, detailing its tasks and the available methods to perform them. The system was then left to Google’s Gemma 3 to manage.
The results were disastrous. The LLM made critical errors, such as turning on the grow light when it should be off and displaying erratic watering behavior based on incorrect interpretations of sensor data. This led to the chili plant’s soil becoming saturated and the grow light operating more frequently than expected, even with an empty water tank.
The experiment highlights the challenges of relying solely on AI for tasks that require precision and understanding of physical systems. It appears that, for now, the humble state machine remains safe from being replaced by AI, and even those with a ‘brown thumb’ can’t kill plants as efficiently as this AI-powered system.
The failure of ‘PlantMom’ serves as a cautionary tale about the limitations of current AI technology in handling real-world tasks that involve complex interactions between software and hardware components.
