LessWrong (Curated & Popular) “How Does A Blind Model See The Earth?” by henry
Aug 12, 2025
Explore how the perception of Earth's geography intertwines with AI and mapping techniques. Discover the charm of early cartography and its personal interpretations of the world. Delve into land probability maps to understand global distributions and the complexities of model architectures like Coda and QEN3. Highlights include advancements in AI, comparing GPT-4 with predecessors and their impact on geographical predictions. The discussion navigates the technical challenges of achieving accuracy in AI visualizations of land.
AI Snips
Chapters
Transcript
Episode notes
Mourning The Age Of Imperfect Maps
- Henry laments that satellite-era certainty erased the intimate, egocentric charm of early maps.
- He values imperfect maps for showing a single mind's knowledge and blind spots.
Seeing The AI's Private Map
- Henry wants to see the Earth as a model internally represents it, not as we do from space.
- He frames the experiment as peering into an AI's intimate, egocentric map of the world.
Probe Models Pixel By Pixel
- Sample evenly-spaced latitude/longitude pairs across the globe.
- Query the model per coordinate and convert land/water logprobs into probabilities to build a map.
