Urgent.News

What's breaking now, across thousands of outlets.

AI

Bro Saw Two Photos and Made Up Lore

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass Apparently, asking a 2B model for evidence was like asking a guy in a group chat for sources. I showed a small AI on my laptop two photos from a walk and asked how they were connected. It almost always had an answer. In 6 of my 10 test pairs, it invented the answer. So I changed the rules of my game: the AI…

This is a story about a game called UNSEEN, which was created to encourage people to go outside and notice real things in the world. The game's creator showed a small AI model two photos from a walk and asked it how they were connected. Surprisingly, the AI invented the answers in 6 out of 10 test pairs. The creator decided to change the rules and let the AI describe what they saw, but not decide what was true.

The game, UNSEEN, requires players to walk a set distance, put their phone away, and take a photo of something they noticed. A local open-weight vision model then describes what they saw. A separate evidence gate checks if the observation has required evidence, is eligible, and has not already been claimed. If the observation passes this gate, the player earns materials for their virtual world.

UNSEEN was built using the open-weight vision model Qwen3-VL-2B-Instruct through Ollama, which runs locally on the player's device. This means that photos are not sent to a hosted AI service and each photo costs nothing. The game is open-source, allowing others to run their own experiments and build upon the concept.

The creator of UNSEEN wanted to create a game that had a real-world connection and encouraged players to explore their environment. The game's philosophy is to be grounded in reality, making the physical world the most intricate and surprising world ever made.

Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at dev.to →

More in AI

Sensing Comes Before the Model

Three published earth observation architectures decide what to watch from law, revisit and ownership, never from model accuracy. Two of the three ship with an empty model register.

  • Sensing decision precedes model selection in earth observation projects.
  • Sensing constraints include legal restrictions, sensor frequency, ownership.
  • Hardware choice influences sensing decision and model analysis objects.

More from Monday 12 October →