I am not sure it’s the same software, but it’s a fairly good guess I think. Same software capabilities and same lab, with the same area of research.
Geoguesser is a subset of the skills used for general image geo location for open source intelligence.
In the specific cases of only using the data present in the image and relying on geographic information, it certainly does better.
Humans still do better, and can reach decent skill with minimal training, at placing images that require spatial reasoning or referencing multiple data sources.
AI tools will likely be able to learn those extra skills, but it doesn’t change that it’s the photo that’s the data leak, and not the tool. The tool just makes it vastly more accessible, and part of the task easier for curious human.
Some blues are reversible, and some aren’t. Some of them do a statistical rearrangement of the data in the area being blurred that’s effectively reversible.
Think shredding a document. It’s a pain and it might take a minute, but it’s feasible to get the original document back, give or take some overlapping edges and tape.
Other blurs combine, distort, and alter the image contents such that there’s nothing there to recombine to get the original.
A motion blur or the typical “fuzzy” blur can be directly reversed for the former, and statistical techniques and AI tools can be used on the later to reconstruct, because the original data is still there, or there enough that you can make guesses based on what’s there and context.
Pixelating the area does a better job because it actually deletes information as opposed to just smearing it around, but tools can still pick out lines and shapes well enough to make informed guesses.
Some blurs however create a random noise over the area being blurred, which is then tweaked to fit the context of whatever was being blurred.
Something like that is impossible to reverse because the information simply is not there.
It’s like using generative AI to “recover” data cropped from an image. At that point it’s no longer recovery, but creation of possible data that would fit there.
The tools aren’t magical, they’re still ultimately bound by the rules of information storage.
I mean, yes, but that’s not what they’re doing.
https://arxiv.org/abs/2307.05845 https://github.com/LukasHaas/PIGEON
It’s a Stanford project that does what it looks like is happening in the screenshot.
https://github.com/LukasHaas/PIGEON
https://arxiv.org/abs/2307.05845
Basically a combination of what the game geoguesser does, and public geotagged images to be able to get a decent shot at approximate location for previously unseen areas.
It’s more ominous when automated, but with only a little practice it’s easy enough for a human to get significantly better.
EDIT: yup, looks like this is the guy from the Twitter: https://andrewgao.dev/ and he’s Stanford affiliated with the same department that made the above paper and system.
Geo guessing is related to open source intelligence techniques, and it’s pretty easy to get surprisingly good at it.
People who are good at it can take a picture of someone’s room and deduce enough about them (sometimes) to be able to get their name, address and phone number.
It being automatic is pretty cool, but you were already leaking the information to anyone interested.
https://www.sans.org/blog/geolocation-resources-for-osint-investigations/
So I actually have one that does.
I get notifications when laundry is done.
I get a notification when I need to do routine maintenance like change filters, or refill the detergent. (It has a built-in jug and dispenser)
I can send it settings via the app, which is easier than via the built in controls. (It has things like extra rinse, wash times for different rinses, and steaming and stuff). It’s not impossible to do via the interface, but it’s a bit easier via the phone.
They did try that for a bit, but it didn’t take off as well.
https://www.polygon.com/a/pc-buying-guide/steam-box
The set of people who are interested in PC gaming but don’t have a gaming PC, or the people interested in couch/console gaming who don’t have one is pretty small.
To be fair, they didn’t offer that level of granular control for a while.
If you’re a company with development prioritization that makes it difficult to say “we need to take a few weeks of not working of things that make money to reimplement something we already have that works, because of best practices that don’t make us any money” then it can be really difficult to make changes like that.
There were some reports from game devs who said that the big reports from Linux users was worth it just for that.
https://playingtux.com/articles/developers-dv-rings-saturn-very-satisfied-bug-reports-linux-users?lang=en
He actually pulled together stats for it all, and it was 5.8% sales making 38% of the big reports, which tended to be high quality.
So from his experience as an independent game dev, he said it was worth it just for the QA you get out of it.
I think a lot of the libraries and tooling being updated to be more platform agnostic helps too. It’s not “press button to support Linux”, but it’s getting a lot easier than needing to rewrite your engine for every platform.