Comparing ‘AI’ for Basic Plant Care With Human Brown Thumbs
1 min read
Summary
The experiment to train an AI chatbot, using Google’s Gemma 3, to look after a houseplant had initially appeared to be successful.
However, the plant, a chilli, was subsequently drowned due to erratic behaviour from the AI, which incorrectly interpreted data about the plant’s soil conditions and prompted the pump to flood the soil.
The AI had also turned on the grow light when it should have been switched off.
The experiment, which is detailed on Hackaday, appears to show that AI chatbots can be a long way from being able to perform complex, real-world tasks, despite their continual infiltration into people’s homes and workplaces.
