PIGÉON GO
A social, AI-powered bird-spotting game.
HOW THE AI WORKS
Just like your phone can recognise your face, our AI will learn to recognise individual pigeons by their unique features—their markings, colours, the way they move, and where they tend to hang out.
To spot a pigeon:
-
You point your camera at a pigeon
-
The AI analyses their unique features (wing patterns, colour variations, posture, location)
-
The system checks if it's seen this specific bird before
-
You get a match—or discover a "new" pigeon to name and collect
The AI gets smarter over time.
Every time someone spots a pigeon, the AI learns more. The more encounters you have, the easier it gets. This approach, called machine learning, means the game gets better at recognising individual birds as more people play.
​
Built on proven science.
We're building on research from existing studies which have demonstrated that AI can successfully identify and track individual birds over time. We'll also be working with AI researcher Daria Kern, who specialises in animal re-identification (she built the first open source chicken re-ID data set!), and training our models using thousands of named pigeons from sanctuary and rescue networks.
​
Pigéon GO applies this technology in a playful, accessible way—turning cutting-edge research into something anyone can use on their daily walk.
AI concern: Idea theft
You may have heard concerns about AI companies scraping the internet and using people's creative work without permission. That's not what we'll be doing.
​
Pigéon GO won't take anyone's art, writing, or creative output. Instead, players actively choose to contribute imagery of wildlife they encounter in public spaces—pigeons that already exist in the world, not creative works owned by individuals.
​
Where we do use generative AI:
In our early stages, we'll use AI to help generate playful personality descriptions for pigeons—things like "loves hanging out near the bakery" or "always first to investigate new visitors." This is a safety measure: it prevents abuse and keeps the experience positive while we build our moderation capacity. As we grow, players will have more input into these descriptions, with community moderation ensuring the space stays respectful and kind.​
AI concern: The environment
AI can have significant environmental costs. Data centres consume water and energy, contributing to climate change. However, we're committed to minimising our footprint and helping you more than offset it.
​​
​How Pigéon GO is different:
We're not training massive generative AI systems. Our computer vision models are focussed on one specific task (recognising pigeons) and require far less computing power than tools like ChatGPT or image generators.
We also plan to use:
-
Edge/device-based compute (processing happens on your device when possible)
-
Cloud processing only when necessary for accuracy
-
Green-powered data centres (Google Cloud, AWS, and Azure all offer carbon-neutral or renewable energy options)
But there is more. ​​
Animal agriculture has a staggering environmental footprint.
Just a handful of facts:
-
While ChatGPT uses about 18.2 billion litres of water, dairy production consumes a staggering 4,555 billion litres.
-
Producing one kilogram of beef requires approximately 15,400 litres of water.
-
One chicken breast requires about 735 litres.
-
Animal agriculture accounts for ~14.5% of global greenhouse gas emissions.
-
Agricultural runoff from livestock farming is the primary cause of the Gulf of Mexico's massive dead zone—an 8,776 square mile area where marine life cannot survive.​
​​
Why this matters:
We believe the water saved by skipping even one animal-based meal far exceeds the water cost of using Pigéon GO. And the mission of Pigéon GO is to help reduce animal product consumption.
So if Pigéon GO helps even a small percentage of players reduce the animals on their plate through gentle reminders or building empathy, we could help turn back the clock on climate change.
​​