Fascinating stuff! I'm not familiar with these new reasoning based models, I wonder how much of the reasoning ability comes from the weights vs the architecture of the model itself vs the system prompt.
Watching o3 guess a photo's location is surreal, dystopian and entertaining
441–450 of 453 posts
Re: Watching o3 guess a photo's location is surreal, dystopian and entertaining
#442Earlier quoted context omitted.
I'd be fascinated to read more about this. I'd love to see a sample screenshot of a few of your cards too.
Sure, I'll write something up later. I'll give you two samples now. One reason I love the Obsidian + Markdown + Spaced Repetition plugin combo is how simple it is to make a card. This is all it takes: https://cdn.jsdelivr.net/gh/sampatt/media@main/posts/2025-04... The top image is a screenshot from a game, and the bottom image is another screenshot from the game when it showed me the proper location. All I need to do…
Fascinating - I just got a little lost thinking about how and why you selected this exact image. If you were able to select a single image containing the maximum aggregate count of unique features for any given location and describe them textually, that would make a very useful training data set... for competitive geoguessr and neural networks alike!
Re: Watching o3 guess a photo's location is surreal, dystopian and entertaining
#443Earlier quoted context omitted.
I made it myself. I use Obsidian and the Spaced Repetition plugin, which I highly recommend if you want a super simple markdown format for flashcards and use Obsidian: https://www.stephenmwangi.com/obsidian-spaced-repetition/ There are pre-made Geoguessr decks for Anki. However, I wouldn't recommend using them. In my experience, a fundamental part of spaced repetition's efficacy is in creating the flashcards yourself…
> In my experience, a fundamental part of spaced repetition's efficacy is in creating the flashcards yourself. +1 to this, have found the same when going through the Genki Japanese-language textbook. I'm assuming you're finding your workflow is just a little too annoying with Anki? I haven't yet strayed from it, but may check out your Obsidian setup.
I do everything from Obsidian now. Anki was very much outside my main flow, and the plugin I linked above works so well for me that I've never looked back.
I did write a tool to convert Anki decks to my own format. I haven’t used it much but it's nice to have.
Re: Watching o3 guess a photo's location is surreal, dystopian and entertaining
#444Earlier quoted context omitted.
Sure, I'll write something up later. I'll give you two samples now. One reason I love the Obsidian + Markdown + Spaced Repetition plugin combo is how simple it is to make a card. This is all it takes: https://cdn.jsdelivr.net/gh/sampatt/media@main/posts/2025-04... The top image is a screenshot from a game, and the bottom image is another screenshot from the game when it showed me the proper location. All I need to do…
Fascinating - I just got a little lost thinking about how and why you selected this exact image. If you were able to select a single image containing the maximum aggregate count of unique features for any given location and describe them textually, that would make a very useful training data set... for competitive geoguessr and neural networks alike!
I have had so many ideas related to this. Particularly around automation.
Because the way the community arrives at meta knowledge and its geographical distribution is very informal. There are quite a few false metas out there, or the commonly understood range of the meta is inaccurate.
What needs to happen is automated image recognition and mapping of the metas. I started building an open source tool that allows for people to do this manually, but it's quite a difficult project to do myself and I'm not pursuing it anymore. But I would like to see a less manual process emerge.
Re: Watching o3 guess a photo's location is surreal, dystopian and entertaining
#445Earlier quoted context omitted.
That is cool, but we cant be guaranteed that will always be the case, nor could we make a statement about all phones, it would be a phone by phone basis. Esp on Android where someone could have an alternative screenshot application. Depending on your threat model, I'd argue that it would be impossible to prove that metadata is not included within the image itself (alpha channel, noise, pushed pixels, colorspace skew,…
Take a photo of the image displayed on your laptop screen with your phone. Ultimate EXIF removal!
Screen dust and smudges now form a fingerprint to cross correlate images.
Re: Watching o3 guess a photo's location is surreal, dystopian and entertaining
#446Earlier quoted context omitted.
I don’t know if most software engineers build toy CRUD apps all day? I have found the state of the art models to be almost completely useless in a real large codebase. Tried Claude and Gemini latest since the company provides them but they couldn’t even write tests that pass after over a day of trying
Our current architectures are complex, mostly because of DRY and a natural human tendency to abstract things. But that's a decision, not a fundamental property of code. At core, most web stuff is "take it out of the database, put it on the screen. Accept it from the user, put it in the database." If everything was written PHP3 style (add_item.php, delete_item.php, etc), with minimal includes, a chatbot might be rathe…
LOL how does the AI keep track of all the places it needs to update once you make a logic change ? I have no idea what you're doing but almost nothing I do is basic CRUD - there's always logic/constraints around data flow and processes built on top.
People didn't move away from PHP3 style of code because it's a natural human tendency - they moved away because it was impossible to maintain that kind of code at scale. AI does nothing to address that and is in fact incredibly bad at it because it's super inconsistent, it's not even copy paste at that point - it's the "whatever flavor of solution LLM chose in this instance".
I don't understand what's your background to think that this kind of thing scales ?
Re: Watching o3 guess a photo's location is surreal, dystopian and entertaining
#447Earlier quoted context omitted.
Our current architectures are complex, mostly because of DRY and a natural human tendency to abstract things. But that's a decision, not a fundamental property of code. At core, most web stuff is "take it out of the database, put it on the screen. Accept it from the user, put it in the database." If everything was written PHP3 style (add_item.php, delete_item.php, etc), with minimal includes, a chatbot might be rathe…
LOL how does the AI keep track of all the places it needs to update once you make a logic change ? I have no idea what you're doing but almost nothing I do is basic CRUD - there's always logic/constraints around data flow and processes built on top. People didn't move away from PHP3 style of code because it's a natural human tendency - they moved away because it was impossible to maintain that kind of code at scale.…
I can see how it would be confusing.
Re: Watching o3 guess a photo's location is surreal, dystopian and entertaining
#448Earlier quoted context omitted.
This is pure luddism. A human could have done the exact same thing. I’ll also point out that in this case the most confident guess was 200 miles off and the second correct guess was only down to the city level. Not remotely what anyone would consider precise.
The fact that it got it wrong was one of the reasons I picked that example: it's much more interesting that way. If you feed it a photo with a clear landmark it will get the location exactly right. If you feed it a photo that's a close up of a brick wall it won't have a chance. What's interesting is how well it can do on this range of tasks. If you don't think that's at least interesting I'm not sure what I can do fo…
I think you may have misunderstood my comment. I think it’s quite interesting. I was responding to parent who is horrified that AI can violate privacy like this:
> This tool just makes it easier for weirdos to achieve their goals at stalking women and kids.
> Crazy that this is even allowed.
Re: Watching o3 guess a photo's location is surreal, dystopian and entertaining
#449Earlier quoted context omitted.
The fact that it got it wrong was one of the reasons I picked that example: it's much more interesting that way. If you feed it a photo with a clear landmark it will get the location exactly right. If you feed it a photo that's a close up of a brick wall it won't have a chance. What's interesting is how well it can do on this range of tasks. If you don't think that's at least interesting I'm not sure what I can do fo…
I think you may have misunderstood my comment. I think it’s quite interesting. I was responding to parent who is horrified that AI can violate privacy like this: > This tool just makes it easier for weirdos to achieve their goals at stalking women and kids. > Crazy that this is even allowed. I was pointing out that to not “allow” using context clues to interpret an image location, something any well-trained human can…
Oops sorry about that, I read your comment out of context.
Re: Watching o3 guess a photo's location is surreal, dystopian and entertaining
#450Earlier quoted context omitted.
And yet many will continue to confidently declare that this is nothing more than fancy autocomplete stochastic parrots.
I will. I am willing to bet that most of this geolocation success is based on overfitting for Google streetview peculiarities. I.e., feed it images from a military drone and success rate will plummet. You're anthropomorphizing a likelihood maximization calculator. And no, human brains are not likelihood maximization calculators.
This overconfidence without making one attempt kills me.
Gemini 2.5 pro is free, try it!
I gave it 2 pictures I took on a film camera. I then screenshotted the pictures (to remove exif, even though there isn't any). It nailed them.
You lost your bet, now change your mind