ChatGPT Plans Custom Running Routes Using OpenStreetMap Data
AI model analyzes local infrastructure to generate precise 5K/10K routes with safety notes.
1AI-powered route generation
ChatGPT successfully created precise 5K and 10K running routes when given a specific address, demonstrating unexpected utility in fitness planning. The model processed OpenStreetMap data for 27 minutes to generate accurate paths, complete with visualizations and downloadable GPX/GeoJSON files. This application showcases how large language models can transcend text generation to solve concrete spatial problems.
2Multi-source data integration
Beyond basic map data, ChatGPT cross-referenced municipal websites for road construction updates and sidewalk availability. It also accessed county greenway maps not included in standard road systems. This comprehensive approach resulted in routes accurate within 100 meters, with specific safety notes about pathway conditions and road crossings that require extra caution.
3Unanticipated use case
The experiment reveals how users may overlook practical LLM applications even when all necessary data exists. While ChatGPT is typically associated with text generation, this implementation demonstrates its capacity for spatial reasoning and local infrastructure analysis. The successful outcome suggests many undiscovered real-world problem-solving applications for generative AI.
4Technical execution details
The model's 27-minute processing time indicates substantial computational work analyzing geographic data. Output formats included both visual embeddings for immediate review and standardized GPS exchange formats for practical use. This dual-output approach bridges the gap between AI demonstration and real-world utility, providing tangible assets for runners.
5Safety considerations
ChatGPT proactively identified potential hazards by analyzing sidewalk continuity and construction zones. The system generated specific warnings about areas requiring verification before running, demonstrating an understanding of pedestrian safety beyond basic route calculation. This level of detail suggests AI can account for practical human factors in spatial solutions.
6Community website utilization
The model's ability to parse and interpret municipal and county websites shows advanced web interaction capabilities. Unlike static dataset queries, this required understanding unstructured government site layouts to extract relevant construction notices and greenway maps. Such functionality could be adapted for other civic information retrieval tasks.
7Precision benchmarking
The final routes achieved remarkable accuracy, staying within 100 meters of exact 5K/10K distances. This precision level indicates sophisticated distance calculation algorithms working with imperfect real-world data. The system compensated for mapping inconsistencies while maintaining target distance requirements.
8Future application potential
This experiment demonstrates how LLMs can synthesize disparate public data sources into personalized solutions. Similar approaches could be applied to cycling routes, walking tours, or accessibility mapping. The success suggests untapped potential for AI in hyperlocal problem-solving beyond traditional chatbot applications.
Original article: MacStories — ChatGPT Plans Custom Running Routes Using OpenStreetMap Data