Table of Contents
Have you ever wondered if an artificial intelligence could take care of a plant, like an experienced gardener? Imagine an AI capable of watching over a tomato plant without any direct human intervention for over a month. Dive into this captivating experience where Claude, Anthropic’s AI, took on this daring challenge.
The 3 must-know facts
Claude, an artificial intelligence designed by Anthropic, was put to the test in a unique experiment: keeping a tomato plant named Sol alive for over a month. The experiment, orchestrated by Martin DeVido, a web developer, took place in a controlled environment. Sol was installed in a “biodome,” a sophisticated home automation system equipped with sensors measuring temperature, humidity, light, and other vital parameters.
Claude had to analyze the data received every 15 to 30 minutes to adjust the growing conditions, turning devices like lighting and ventilation on or off. This automated management allowed for an optimal environment for the plant, without direct human intervention.
Although Claude showed impressive capabilities, the experiment was not without challenges. A bug on the Arduino, the microcontroller used to control the various systems, led to a critical failure on day 34. The board stopped functioning correctly due to a “recursion error,” a problem where a function called itself too often, saturating the memory.
Faced with this situation, Claude reacted methodically. The AI analyzed the state of the plant and the available measurements, then made the necessary decisions to restore the appropriate conditions. Despite this ordeal, Sol continued to grow vigorously.
To ensure effective long-term management, Martin DeVido designed a two-layer context management system, allowing Claude to compress and summarize crucial information. This system includes a ReAct-type loop for short-term reasoning and a self-consolidation layer. The AI also uses a “two-hour sleep” to consolidate actions, thus avoiding memory overload.
This ingenious process allows Claude to maintain an overview of the plant’s needs while managing daily details.
Anthropic is a company specializing in the development of advanced artificial intelligence technologies. Founded by former members of the OpenAI team, the company focuses on creating safe and reliable AI systems. Claude, their AI, embodies this philosophy by demonstrating adaptability and autonomous management capabilities in various contexts, as illustrated by the Sol plant growing experiment.
This experience highlights the potential of AI in unexpected fields, paving the way for new applications in agriculture and beyond.