I remember in 2009 being part of an online community where I preferred to be anonymous and while reading a thread where people were sharing photos of sunsets, had the thought that eventually, there would be enough information out there that my precise location could automatically be determined based on me taking a picture looking out at the city from my balcony. That's a pretty creepy thought, and it seems that google is approaching it, given that they can have you point your camera "across the street" to determine your location from their glorgabytes of street view imagery.
Geoguessing has been going on for a while, and some people are very good at it.
One of the situations where it’s deployed for the good of all is real estate: it’s very common for brokers to hide the exact location for various reasons, but there are forums and individuals who from just a few pictures pointing outdoors and a rough location (e..g closest city) can very quickly find the exact location the picture was taken from.
Super useful so you don’t have to waste time visiting in order to find out that the house which looked good over the internet is sandwiched between a 6-lane highway and a pig farm, or is in a gated community, or can only be accessed through an easement.
Yes, I intentionally used the word "automatically" to my description, since it's very obvious to me that it's already often doable if a human is willing to engage their brain and do research. Whenever I use airbnb, I make a game out of finding the actual addresses of the places I'm considering based on clues I can see in the photos and approximate maps on the site.
I also find locating images an interesting skill to develop. I originally only wanted to mention https://old.reddit.com/r/picturegame,* which I find more "organic" than Geoguessr: players submit single images, which are identified using whatever means are available. It has a quirk in that often people submit images found online, which must be obfuscated to prevent trivial reverse image searching. A randomly transformed translucent mask is recommended, but those are still relatively easily reversed! I haven't tried, though--maybe there's greater difficult than at first apparent.
* It turns out, after an HN search, that this is a more common organized activity than I'd thought; this thread lists several other communities:
I've been doing something like this privately for my own picture library – because my backlog of pictures-to-be-geotagged is about 15 years, I don't necessarily remember all the details where exactly a picture was taken. Of course I could just drop the location marker in roughly the right location and call it a day, but pinpointing the location as accurately as possible is usually too tempting.
> […] there would be enough information out there that my precise location could automatically be determined based on me taking a picture looking out at the city from my balcony.
In Japan someone doxed a singer by zooming in on the reflections that were in her eyes from a selfie she took:
> A Japanese man accused of stalking and sexually assaulting a young pop star told police he located her through the reflection in her eyes in a picture, according to local media reports.
>...there would be enough information out there that my precise location could automatically be determined based on me taking a picture looking out at the city from my balcony.
Not yet automatic, but the folk at Bellingcat are impressive at figuring out location from pictures and videos (manually, I assume).
There's a fascinating technolibertarian metaverse dynamic to Minecraft anarchy servers, where gameplay is driven by an arms race of normalized cheating. Players are careful about what screenshots they share online, as this concern is the reality there. There's kind of a similarity to a future real reality where the entire Earth has been autonomously scanned: the terrain of the entire map of a Minecraft server is known to all, since the seed for procedural world generation can be reversed from nearby terrain, and reverse engineering of the world generation has resulted in several methods of determining in in-game coordinates of a player taking a screenshot (terms to search: trees, texture rotation, bedrock).
Related are the impressive endeavours that determined the seed of the screenshot used in the game's splash screen and the seed of a 128px (tiny) screenshot used in the UI. The top comment of the first link explains it better than I could here:
It's interesting how the anarchy server hacking scene is sort of like science, except with the god-like abilities of reverse-engineering the game's binaries.
Inspired by 1) an HN title claiming that Minecraft servers are the/a metaverse[0] and 2) the old Net philosophy of minimal regulation of technological advancement. Treating a server as a society, political and economic concepts apply quite well, including the advancement of the "technology" of gameplay: strategies, in-game inventions, client-side modding, etc. This applies to every game, but Minecraft in particular is known for its extensive unintentional emergent gameplay.
And server owners--the government--typically draw some line before major glitches and definitely before blatant the cheating of server-side exploits, banning players who do not obey the rules. But the owner of an anarchy server explicitly does not do this, and instead treats any advancements in even cheating technology as legitimate in-game technology. Hence, a technolibertarian metaverse.
People have triangulated down to the exact apartment balcony from a picture taken from it, multiple times. Google maps helps in that it allows someone not from the area to do it, but it's always been possible if there's enough information available in the picture (building skyline is the main one).
The key thing I was talking about was the difference between it being possible to deduce the location with enough effort and it simply being automatic. Google has automated it in one specific scenario, which is that you are looking at something which is substantially covered by street view imagery taken from a similar vantage point.
