Little Shop of Conversations: Talk to anything using AI

Have you ever wished you could talk to things that can't talk? Like ask your plants how their day is going? Find out if they need more light or if they're thirsty? Just imagine the conversations you could have if only they could talk back. Well, with the power of (drumroll please, you’ll never guess) AI, now they can!

‍

Plants might not have mouths, voice boxes, lungs, or brains that would allow them to carry on a conversation, but they do generate electrical impulses in response to stress, touch, and environmental factors. We can detect these signals, along with objective environmental features such as light, humidity, growth, and nutrient levels. We can also store that data in a local database, essentially giving the plant a memory. Questions like, "How many days since I was last watered?" or "Should I be worried about the dry soil?" can now be reasoned.

‍

To collect this data, we use off-the-shelf and DIY sensors for light and soil moisture. These sensors connect to analog-to-digital converters that relay the raw data to a Raspberry Pi. This mini-computer decodes the data into actionable insights like soil moisture levels—whether the soil is moist, balanced, or dry.

Before interacting with the AI, we set up a few parameters to give our digital tomato plant a personality. Utilizing OpenAI's GPT-4, the system interprets and responds as though it were a tomato plant with a deep knowledge of its own state and environment.

‍

When someone speaks to it, their voice is detected by the onboard mic and is converted to text using the OpenAI Whisper API. This text is then fed straight to GPT-4, which has been primed with the personality and knowledge base of the tomato plant. GPT-4 crafts responses in the voice of the tomato, making it seem as though you're actually hearing the plant's thoughts and feelings. The text response from GPT-4 is then converted into human-sounding audio using the OpenAI Text-to-Speech API. This audio is played back to the loudspeaker, allowing for a natural, ongoing conversation.

‍

Future Enhancements

While this prototype uses relatively simplistic data to shape the personality of the tomato plant, you can imagine integrating more complex sensors for a richer representation of the plant's experiences. Things like live forecast data could inform the plant about potential weather concerns, adding another layer to our interactive photosynthetic friend. Throw in some bioacoustic sensors, air quality monitors, pheromone sensors, soil nutrient sensors to enrich the personality of the plant to the point that you might as well be talking to a human.

What in Green Goddess’s Name Does This Have to Do with Marketing?

We’ve recently been building interactive personas that are basically backed by real user feedback data with an LLM AI personality overlaid. It lets you ask personas questions about marketing tactics, research, and overall feedback. Instead of referencing a one-sheeter static marketing persona from time to time, you can keep the persona in the loop during the whole process, getting feedback on everything along the project timeline. Basically, instead of talking to a tomato plant backed by real-time data and a unique personality, you’re interacting with a persona backed by actual data and a unique personality. And that, folks, is how we got to talking tomato plants.

‍

So there you have it—you can now converse with your plant friends in a natural, engaging way using cutting-edge alien technology known as Artificial Intelligence. What other mouthless objects might you talk to next? With sensors and a bit of hackery, almost any object can be endowed with a personality. So next time you step on the grass, maybe ask how it feels about that. And remember, unlike in Little Shop of Horrors, these plants won't demand anything more than a bit of water and some sunlight.

‍