ChatGPT Can Now Help Humans Speak to Trees. But Why?
Commentary: The ePlant TreeTag gives your eucalypt an AI voice. I'm not sure it needed one.
Jackson RyanFormer Science Editor
Jackson Ryan was CNET's science editor, and a multiple award-winning one at that. Earlier, he'd been a scientist, but he realized he wasn't very happy sitting at a lab bench all day. Science writing, he realized, was the best job in the world -- it let him tell stories about space, the planet, climate change and the people working at the frontiers of human knowledge. He also owns a lot of ugly Christmas sweaters.
For the past six months, I have stared into the AI abyss with the eyes of a man unable to fully wake from a nightmare. Every morning my inbox is cluttered with PR firms pitching stories about how Startup X or Business Y has just used ChatGPT to revolutionize industry Z.
ChatGPT, a chatbot developed by OpenAI to respond to human queries about anything from cake recipes to math problems and drafting emails, seems like it can do it all. It's based on a large language model, or LLM, built to understand human language using machine learning algorithms. It can find patterns in text and spit out confident (but sometimes completely wrong) answers in response to a user prompt. It's the hottest commodity in tech right now.
As a result, my Deleted Items folder is bursting at its digital seams with everyone riding the hype. Most pitches end up in that folder, but last week there was one that survived the daily purge. It presented the possibility of revolution.
"How Will ChatGPT Play a Role in Tree Health?" it asked.
The headline led into a pitch about a device built by Californian tech firm ePlant: a solar-powered device known as a "TreeTag."
The TreeTag, a tiny box about the size of a Roku, is affixed to the trunk of a tree and has a suite of five sensors that enable it to capture information about the tree's health. The sensors record light, moisture and temperature. It can tell you how much the tree is moving or leaning via an accelerometer. There's also a roundabout way to understand the water and nutrient flow, which is determined by dendrometry -- measuring the size of the plant's inner structures.
The data is uploaded to a cloud and accessible via the ePlant app. For any budding arborists or forestry folk, it sounds like a great addition to the toolkit, one that helps monitor a tree's vital signs in near real time. As someone living in a Sydney apartment that has no garden or yard, it doesn't really appeal to me, but it has obvious benefits for others. All pretty cool stuff, really.
So, where does ChatGPT come in? The TreeTag app allows the tree to have a conversation with the user about itself. About what it needs.
In other words, yes, ePlant wants your tree to "speak" to you.
This is a sample conversation provided in the press release:
Tree: "Hello there, It's me Gumdrop! Your Eucalyptus Friend. I am feeling a little bit concerned today because I have been leaning more than usual."
Homeowner: "How much are you leaning?"
Tree: "I'm leaning 2 degrees more to the North versus a week ago and I'm worried about toppling over."
Homeowner: "How can I help you?"
Tree: "It's time to call an arborist to take a look at me. I'm concerned about falling on my surroundings."
Put aside the fact Gumdrop is the most cliche name for a eucalypt, and you're left with an obvious question: Why on Earth would I even want to have a conversation with a tree? How does this conversation improve the experience of caring for it? The conversation feels like an unnecessary complication. It takes a single piece of data from the accelerometer -- Gumdrop is leaning 2 degrees more to the north -- and wraps it in a conversation that appears unnecessary. An app could present this data, visually, in less than a second. It might even flash a warning for a user. A notification, perhaps.
An ePlant spokesperson said that "there's a logical reason to use the generative language models" because they've been "pre-trained on published horticultural guidance that can be targeted to the specific tree in question." OK, maybe, but I still have to talk to the tree via a mobile app.
The TreeTag might be a useful product, but its ChatGPT integration is emblematic of a fundamental problem of the generative AI and LLM boom: Companies are shoehorning these new technologies into places they just aren't required.
Perhaps not everyone will be like that. ePlant, according to its spokesperson, sees it as a way to build a relationship with a tree. "It goes from something you care about, the same way you care about a nice pair of shoes, to someone you care about, elevating the tree into an entity that deserves care," they noted. "The team at ePlant hopes that actually changes how you think about all trees."
I'm not so sure.
The TreeTag is reminiscent of the way Web3 and cryptocurrency, during the 2021 boom, started appearing in every product and pitch even when it really didn't need to. For instance, some claimed the blockchain would revolutionize the way we play video games. But very few games are building themselves around blockchain and Web3 technologies. Those who've attempted to push a bitcoin-shaped-peg into a Super Mario-shaped-hole have not seen any long-lasting, mainstream success.
If I've learned anything from the rapid rise of ChatGPT, it's that people are obsessed with the ways to break the tech. Yes, there are plenty of productivity hacks, but there are just as many folks trying to peek behind the curtain and find some hidden, twisted personality trait. Remember when a reporter found Microsoft's early version of Bing might try to break up a marriage after lengthy conversations and the internet went into meltdown for a few hours? You can see that a homeowner is going to M. Night Shyamalan their Gumdrop and try to kickstart a global tree apocalypse, a la The Happening.
The Shyamalan twist here is that it might not even be beneficial to give plants a coat of human personality paint. Kathyrn Flinn, a plant ecologist at Baldwin Wallace University, wrote in 2021 that "plants are fundamentally unlike us: mute, rooted and inscrutable." Rather than turning toward anthropomorphism to understand them, she suggested, we should try something entirely different.
"We need to meet the challenge of cultivating respect for organisms that are different from us -- in their separate and complex bodies, in their sophisticated interactions, in their unfathomable lives," she said.
We can pretend that we're giving "voices" to the organisms around us and building relationships with non-sentient beings with LLMs, but these voices are just our own voices speaking back to us. Sometimes, those voices even tell us the wrong thing, as we've seen time and again with ChatGPT hallucinating (as well as other large language models, like Galactica). The billions of human words across the internet upon which ChatGPT has been trained are the basis for Gumdrop's responses. It isn't the tree speaking to us. It's us, pretending to be a tree, speaking to us. That's weird. That's strange. That feels a little insane.