Google uses 330-number fingerprints to predict populations better than night lights
anselm · x · 2026-08-22
Google developed a new method converting 46,000 locations into 330-number vector fingerprints for population prediction. Traditional methods relying on physical traces like night-time lights or road networks can be misleading (e.g., industrial zones or commuter roads). This embedding-based approach captures latent features and outperforms models based on visible geographic data, signaling a significant shift in population mapping.
More from Research
- U of T team mines 122,194 patient reviews across 45 Ontario hospitals with NLP to expose satisfaction and equity patterns — ishtiaqueSIA · 2026-10-16
- University of Toronto hiring AI postdoc: builders of interventions, not just benchmarkers, due Oct 30 — ishtiaqueSIA · 2026-10-16
- WorldMemArena Accepted at NeurIPS: Memory Storage ≠ Memory Use — xwang_lk · 2026-10-16
- Mech Interp Veteran Rethinks Alignment, Misgeneralization, CoT Oversight — akbirthko · 2026-10-16
- WEDGE paper generates training environments conditioned on navigation agent failures — _akhaliq · 2026-10-16
- RExBench: best coding agent completes only 33% of autonomous research extensions — najoungkim · 2026-10-16