A VerySpatial Podcast | Discussions on Geography and Geospatial Technologies · VerySpatial

A VerySpatial Podcast - Episode 751

·35 min·2 clips

CarPlay starts routing Sue to Follies before she asks for directions.

1. A Very Spatial Podcast episode 751, dated December 1, 2024, centers on Microsoft Places, Foursquare Open Source Places, Niantic's visual positioning system, and Google's Location History changes. 2. Jesse, Sue, Barb, and Frank host the roundtable and move the conversation from enterprise software to geospatial data products. 3. The episode asks what location software knows about offices, phones, places, and routines when systems start predicting where people go. 4. Microsoft Places is introduced as a repeat from 2022, tied to Teams Premium or Exchange, and the hosts call "Places" a terrible name. 5. Sue's CarPlay example shows her phone routing her to Follies, the local music store where she takes guitar lessons on Monday nights. 6. The group uses that route prediction to talk about the "creepiness factor" of software that knows more about a person's day than they want it to know. 7. Foursquare Open Source Places is described as a base layer of 100 million-plus global places of interest with 22 core attributes. 8. The hosts also point to a schema page, a future placemaker contribution path, and a separate Places Pro premium data set. 9. Frank compares the release with broader platform work from Amazon, Meta, Microsoft, TomTom, Esri, and Virtual Maps, while noting that Foursquare is not part of that consortium. 10. The Niantic discussion frames scanning and visual positioning as a geospatial foundation model built from large-scale user participation in Pokémon Go. 11. Frank pauses on the phrase "visual positioning system" and jokes that corporate people do not need another positioning system. 12. The article's scale numbers stand out: 50 million neural networks and 150 trillion parameters. 13. Sue and Barb connect Niantic's approach to Microsoft Photosynth, where scraped point clouds missed building surfaces that were never photographed. 14. The hosts say humans make the same kind of spatial inference when they see a church or town square from one angle and assume the rest. 15. Jesse and Frank suggest that a "large language model" and a large geography model could combine textual context with fuzzy spatial inference. 16. Google Location History is moving to on-device storage, and the hosts tell listeners to use Google Takeout if they still want a web download. 17. The tone stays conversational and lightly skeptical, with the hosts interrupting, correcting, and riffing on each other's examples. 18. The format mixes news reading, a short web corner about You Are Here Podcast, and end-of-year event updates and Blue Sky mentions. 19. Geospatial technologists following maps, datasets, and location privacy. 20. Listeners wanting a single focused story arc.

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from A VerySpatial Podcast | Discussions on Geography and Geospatial Technologies