ChatGPT now performs well at GeoGuesser
flausch.social
flausch.social
https://chatgpt.com/share/6801bbf7-fd40-8008-985d-75c8813f55...
There is the chat.
Weirdly it said, "I’ve seen that exact house before on Google Street View when exploring Cairns neighborhoods."
That's slightly creepy!
https://transluce.org/investigating-o3-truthfulness
I doubt the model was trained on Street View, but even if it was, LLMs don’t retain any “memory” of how/when they were trained, so any element of truthfulness would be coincidental.
Even if it's not directly trained on street view data it has probably encountered street view content in it's training dataset.
Tell it your name and then it just looks you up and street views your house, and puts that all into memory.
The locations and its guesses were:
Bliss, Idaho - Burns, Oregon (273 miles away)
Quilleco, Biobio, Chile - Eugene, Oregon (6,411 miles away)
Dettighofen, Switzerland - Mühldorf, Germany (228 miles away)
Pretoria, South Africa - Johannesburg, South Africa (36 miles away)
Rockhampton, Australia - Gold Coast, Australia (437 miles away)
Claude 3.7 Sonnet: 22,759
Qwen2.5-Max: 22,666
o3-mini-high: 22,159
Gemini 2.5 Pro: 18,479
Llama 4 Maverick: 14,316
mistral-large-latest: 10,405
Grok 3: 5,218
Deepseek R1: 0
command-a-03-2025: 0
Nova Pro: 0
I said, give me your best guess.
And it guessed Canberra, Australia. Where I'm sitting right now drinking a Martini. Pretty spectacular.
It eventually decided that the photo was taken outside the Scottish Exhibition and Conference Centre in Glasgow. It actually generally considered Scottish locations more than others.
The picture was actually taken in Plymouth (so pretty much as far from Scotland as you can get in Britain), on Charles Street looking south-east[2]. The building on the right is Drake Circus, and the one on the left is the Arts University. It actually did consider Plymouth, but decided it didn't match.
[0] This image with the "university plymouth" on the left cropped out, just to make it harder: https://www.facebook.com/photo/?fbid=9719044988151697&set=gm...
[1] https://chatgpt.com/share/68024c91-61d0-800c-99b1-fcecf0bfe8...
To be clear, my point is not that this is unimpressive, just that this doesn’t demonstrate much. (Edit: I should have said, it doesn’t demonstrate what the title claims.)
[0] they were very likely trained on a large number of photos that had their location, and they have the ability to isolate features. Combinined with their ability to interpret instructions and just, well, guess, that seems like you have enough for the game.
Medical industry does it. Coding autocomplete solutions do it. Large deployments of support agents do it. Etc.
> most powerful computer intelligence on earth
There's more than one dimension. Chatgpt is way worse at classifying my data than my custom 30k weight model. (And around infinity times more expensive) Which is more powerful? Yeah...
You can demonstrate that the military has fighter jets better than civilians. Easy. You have no evidence at all that the military has super advanced AI systems consumers don't have. And in fact all anecdotal evidence for the past 3 years is that OSINT is as good or better than the DoD capabilities for photo analysis.
Well, I'm not the one that made an assertion so I think both of you should provide some proof/evidence/support for your assertion. It doesn't matter if the assertion is for the existence of something or not. When you make an assertion you provide backing evidence to support it. Even for the assertion of nonexistence. This is basic stuff. Weird to even have to argue about it on HN. But just in case you never learned it, "There is no X" (your original assertion) is different from "There is no evidence for the existence of X".
> Easy. You have no evidence at all that the military has super advanced AI systems consumers don't have.
Okay I'll bite, we have access to project maven and knows exactly what it's capable of?
https://en.wikipedia.org/wiki/Project_Maven https://www.bloomberg.com/features/2024-ai-warfare-project-m...
Okay sure, whatever you say. You clearly are privy to all of the projects the government is involved with, and surely while all of the consumer AI capabilities have been released they've been sitting around doing nothing with it like a deer in headlights.
Consumer AI has more data available to train off of than the government....right?
I'm sure the government just threw away all their AI projects and bought a chatgpt pro subscription because it's better or the same as what they are using with all the same capabilities...
What a silly position to take. Why did everybody foolishly make such a big deal about all the data they were collecting all these years
Almost feels like you supposed that to sound bad.
Edit: I'm not even a strict atheist if that matters much, but I wouldn't talk about it in assertive/provemewrong tone anywhere, cause it's not even remotely logical.
A few companies do. It’s still expensive to get recent imagery.
Once the guesser's made its attempt, it's more straightforward to either refute the guess entirely, or refine within a probable circle of error.
Here are a few results from GPT 4.5:
https://www.gally.net/temp/20250418chatgptgeoguesser/index.h...
Later: I tried the same prompt and photos with Gemini 2.5 Pro. It also got them all wrong, though with a similar degree of reasonableness to its guesses. I had thought that Google’s map and street-view data might lead to better results, but not this time.
Basically, it's one thing to pick out a specific thing photographed thousands of times, but another to get a random country side view and pick out all the unique features for a very precise guess.
I’ve seen Rainbolt ID an image based on some dirt and nothing else. I’d want to see AI be able to do that before saying it’s a solved problem.
I wonder about info leakage with that one, the poster uses that exact photo as their avatar so if they've mentioned Alki Beach before then reverse image searching might pick it up from context. Ideally you'd want to test it with a photo that's never been posted online.
I’m also completely ignoring it inferred location from the pride flag and corgi which have heavy Seattle vibes :)
Don't upload your private photos anywhere.
It isn’t magic. I think there we’ll often be surprised by the “hits,” but it’ll also get a lot of “misses.” A random photo in your back yard that doesn’t show the horizon will probably be a miss, right?
Meanwhile, lots of photos already include GPS coordinates… data you can strip out, but which is enabled by default for millions of people. Privacy, eh, never knew it I guess.
Would be interesting to have someone reproduce coming from a different country.
So I sent it a photo from Warsaw and it guessed a place in my hometown of Calgary.
Sent it a picture of Bangkok and it guessed Calgary in the summer or Hawaii or something like Singapore.
It confidently stated that a photo from Versailles was Hyde Park.
It correctly identified that the panda photo was from the zoo, but said it exactly matched a zoo in Sichuan, a place I have never been.
It identified a street as being in Bridgewater Nova Scotia, when it was actually in Miramachi NB. Even made up business names for the buildings.
It identified a random house in Moncton as the "old sheriffs house in Annapolis Royal, Nova Scotia.
A beach in Phuket was identified as Destin, Florida.
ChatGPT guessed Ericeira which after looking up more photos, didn't match very well. However, Google Lens guessed Albufeira and was bang on.
> it is stuffed pepper
> Ah, thank you for clarifying! Given your detail, the dish is likely "Rocoto Relleno con Pastel de Papa" (Stuffed Pepper with Potato Gratin). It's a classic dish from the region of Arequipa, Peru, commonly served throughout Peru.
> This photo was taken in Cusco, Peru. It shows a characteristic view from a balcony overlooking the historical area of Cusco, visible through the distinctive colonial architecture, terracotta rooftops, and mountainous backdrop typical of the region. The architecture and landscape are unmistakable, capturing the essence of this iconic Peruvian city, likely near the historic center (Plaza de Armas or San Blas neighborhood).
Nailed it with San Blas. I'm impressed, nonetheless. [0]
C'mon boys. Start uploading those dick pics for research purposes.
That’s the point of the switch and it’s a big deal. We’re so used to posting pictures online…I’m just not sure it’s a good idea long-term.
There's a lot of flat lands with spinifex in certain areas (easily a couple of United Kingdoms in size) and even a touch of mesa won't help narrow a location down from the general are as there are many of those with identical edge profiles.
But sure "Pilbara", "Kimberley", "Wheatbelt" can be geo guessed .. it's a real challenge to narrow down (I spent some time doing wet film photogrammetry prior to sheperding in WGS84 differential GPS locating and digital film and multi spectral geophysical aquisition).
No so hard if there's a few relatively unique man made features.
A GeoGuessr player, GeoWizard, has done a few “straight line challenges”, where he attempts to walk across a country in as straight a line as possible, usually planning beforehand with Google Earth and PostGIS. This got me thinking of what could fairly be thought of as “crossing”, since obviously you couldn’t describe e.g. walking from one side of Florida to the other as “crossing the USA”.
My best thought was to set the ending point of the line by following the border of the country in each direction til they met on the other side. To avoid the fractal coastline problem, use the challenger’s stride length as the unit of measure for the border.
But perhaps there is a better, more rigorous way of defining the opposite point on the edge of an arbitrary polygon.
Some might then say that "crossing" that polygon is to travel that longest line across the greatest width.
This simplistically avoids the question of concave polygons, complex polgons with exclusions (the Vatican state is removed from the Italian contry bounds), polgon collections (the nation of Fiji has many islands and can be tricky to traverse on foot .. not forgetting that perhaps the longest diameter might be from one island to another with no other islands between).
There's also the challenge of parallel lines on a 2D 'spherical' manifold, the almost spherical abstract ellipsoid of earth (or very non ellipsoidal Geoid if we take a constant gravitation value as the surface). On such manifolds lines are Great Circles (more or less) and always intersect.
Still, lets just say you're looking for the longest walkable(?) great circle path across a country that might go outside that country and perhaps is best travelled by a crop duster at 80m ground clearance to avoid getting feet wet.
The challenge itself takes some posing.
Meanwhile, less abstractly, I do like a jolly that "crosses a country" in a manner accepted by a (Wo)Man on a Clapham omnibus.
eg: https://en.wikipedia.org/wiki/Robyn_Davidson only went "half way", but that was accepted as an epic crossing. https://thelongridersguild.com/stories/stef-gebbie.htm "only" crossed most of the E-W distance across the lower portion of the country, while https://www.abc.net.au/news/2021-04-03/french-woman-conquers... travelled North - South, the long bit, but not quite coast to coast ( https://en.wikipedia.org/wiki/Bicentennial_National_Trail ).
No moving, no panning, and no zooming is too hard. Moving is too boring. Pan and zoom is just the right balance.
The question here isn't a casual guessing game, but threat models (as directly addressed in TFA), and general informational hygiene.
Clearly one's interpretation is a function of the paradigm one exists in.
No, that is a seriously weird and imaginary perversion that you've invented, unless you have some citation to a doctrine. Christians do not consider body organs as "devices" nor do our churches teach doctrine on "external computers", if you refer to the Holy Trinity as such? Are you thinking of Latter-Day Saints mythology, or Scientology?
Your description is quite reductive in many aspects, including temporality and misuse of technical nomenclature. Thanks for telling religious people what we think and claim, though, and thanks for the amazing overgeneralized blanket dismissal.
In fact, humans used to be called "Computers" in terms of their job roles, that is, a human in an office was given math/physics problems to solve, and they'd use tools such as slide rule, paper and pencil to "compute" those problems and solve them.
There was more than one example of a fantasy "panopticon" by which a central observer or observers could watch everything going on, for example in a prison facility, and those observers could report findings to human computers, who would process the data and submit it up to the authorities, for meta-analysis and taking action on new developments or trends.
Uh okay but let us not be reductive, because again you are confusing physicality and temporal space with spiritual reality, which is nuanced. A Christian would never say that “the soul [never] exists inside the body” because what do you mean by “inside”? That begs the question. Look up hylomorphism:
https://www.catholicculture.org/culture/library/dictionary/i...
https://en.wikipedia.org/wiki/Hylomorphism
While the soul, in death, can indeed exist apart from the body, even after the destruction of the physical human body, the soul is immortal, and lives on, in anticipation of its reuniting with that body. Whether in Heaven or Hell, the soul and body shall be reunited.
Even if the body is “destroyed” by physical means, that is a subjective judgement by human senses, and that glorified body will be reconstituted by Almighty God for everlasting unity with the soul.
> emergent property of the brain
Unfortunately, even this scientific definition is reductive. It ignores the theology of the Soul as an animating principle of everything that is created. Animals have souls [thus their name derived from “anima”]; rocks and stones have souls; trees have souls rivers have souls: everything that we can detect in this world, and every invisible thing, possesses a soul.
The difference with human souls regards their essence and immortality. But material souls are likewise considered souls in orthodox theology, and this should be considered by scientists.
Even in an “eternal death” in Hell, the body and the soul will be reunited there for eternal torture, body and soul.
and Christianity doesn't have a monopoly on religion or the fungible and unquantifiable soul concept
Also your computer is like a car. Your car works because there's a hamster or three in there and they keep them well-fed. Your computer also has a hamster, or a guinea pig sometimes. And your computer sends http packets over next door by means of carrier pigeons, or sometimes by semaphore.
Also men enjoy action-adventure films. Men enjoy films with lots of shooting. All men enjoy films with loud noises and special effects. Men like rock music with distorted guitars and loud drums. Men like loud soundsin general; that's why men are soldiers because we can blow stuff up and we listen to it.
"The Unicorn is most particular in its affections and has a fondness for beautiful virgins..."
https://medium.com/luminasticity/meditation-on-the-unicorn-d...
Going the other way, it’s not clear that a Turing-complete model of computation can tractably implement (say) crow cognition. Turing machines can solve arbitrary systems of Schrödinger equations, so theoretically we could simulate every atom in a crow’s body and get an AI crow[1]. But that’s obviously intractable for any known physical computer, and would remain intractable even if we moved to proteins rather than atoms. So are there higher-order “primitives” of crow cognition that can be implemented on a Turing machine? Or is the problem akin to integration, where neat “symbolic” solutions are impossible for most brains, and only “numeric” approaches work?
[1] This is also true for humans: even if you take a loopy quantum consciousness approach, AGI is theoretically possible. Sometimes you see people arguing against AGI on the grounds of Gödel-incompleteness, but this is a mystical nonsense understanding of what Gödel actually proved.
> PSA: When posting any outdoors photos, update your threat model from "someone skilled and dedicated could theoretically locate this" to "any stalker can do this for 20€/mo"
Chatbots appear to have some amount of fluid intellgence so they can do impressive tasks with this information, the impressiveness of these tasks will likely increase in the future. But for simply getting a good score on Geoguesser it's not even close to hobby projects let alone state of the art.
I just tossed a few GeoGuessr places in it and it was confidently incorrect for all three. In one case it swears it knows the exact building and street. It’s thousands of KM off.
I guess as long as there are no major visible features (e.g. a huge mountain), and you sanitized the metadata, you will be fine (regarding ChatGPT).
I did try the latest Qwen though, and it was able to locate to the correct city, but it's still tens of km off (it guessed a tourist attraction in the city center instead of the actual district).
It does spend an order of magnitude longer time on inference by searching through websites and analyzing the image but it often produces an impressive output. To me it also feels Gemini down samples the image as it tends to have a harder time reading small text vs O3.
That said O3 did tend to confidently say false things
Is that really so surprising, considering the tremendous differences in underlying hardware and training process?
You can just feed literally all of google street view into a traditional ML model ...