Why Humans Love to Build Stories from Pictures
Human cognition is inherently narrative. When presented with visual cues, the brain searches for patterns and connections, turning isolated images into vivid scenarios. This tendency is visible in online memes such as the “Bro Saw Two Photos and Made Up Lore” phenomenon, where minimal visual evidence sparks elaborate, often humorous backstories. The meme reflects a broader cultural practice of generating fan lore, creepypasta narratives, and internet folklore from simple stimuli.
AI and the “Two Photos” Experiment
In a test scenario, a 2B AI was shown two photographs taken during a walk. The model was asked to explain a connection between them. Out of ten pairs, the AI invented an answer for six, illustrating its propensity to fabricate links when evidence is sparse.
This outcome prompted a re‑evaluation of the role of AI in storytelling: the model can help observe details but should not decide the truth of a narrative. The result was a new interactive concept that blends observation, verification, and playful speculation.
The UNSEEN Walking Game
UNSEEN is a mobile‑first walking game that turns real‑world exploration into a virtual quest. Players begin an expedition, set their phone aside, and return with a photograph of a noteworthy object or scene. A locally hosted vision model describes the image, but game rules determine whether the capture earns in‑game rewards. The process encourages deliberate observation and critical thinking about what counts as evidence.
How It Works
- Walk. Set a duration, silence the phone, and stroll.
- Notice. Spot an element in the environment that feels significant.
- Return. Submit a photo; the on‑device model generates a textual description.
- Discover. An evidence gate evaluates the description against predefined criteria, accepting or rejecting the entry and explaining the decision.
Benefits for Outdoor Exploration
- Promotes mindful walking by encouraging attention to surroundings.
- Provides a low‑screen‑time activity with a tangible reward system.
- Introduces players to the intersection of AI, vision, and human judgment.
Technical Foundations
- Node.js runtime for server logic.
- Three.js for immersive 3D presentation.
- Vite for streamlined development.
- Ollama as the local inference engine.
- Hugging Face Qwen model for image description.
Getting Started
Source code, documentation, and quickstart instructions are available on GitHub: Unseen Repository. The project welcomes contributors interested in expanding the evidence rules, adding new open‑weight vision models, or improving the mobile experience.
By integrating real‑world observation with AI assistance and gamified feedback, UNSEEN offers a unique platform that respects the authenticity of physical experience while engaging the imagination. It demonstrates that technology can amplify human curiosity rather than dictate narratives, echoing the playful spirit of the original meme and the deeper psychological drive to find meaning in everyday moments.
Leave a Reply