Does anyone remember that 4chan thing where they geolocated some secret flag location and they used info from planes they saw in the sky or something? I wonder if it could do that now.
Watching o3 guess a photo's location is surreal, dystopian and entertaining
201–210 of 453 posts
Re: Watching o3 guess a photo's location is surreal, dystopian and entertaining
#202Ok so if given LLM generated code...Will o3 be able to find commercial or open source code similar or very, very, similar to the LLM generated code? Meaning the training source code, possibly showing copyright violations? So its own code version of "where was this photo taken?"
o3 is very good at searching the web, so it might be able to do that.
Re: Watching o3 guess a photo's location is surreal, dystopian and entertaining
#203The python zoom in seems performative. A vision model already has access to all the data, how does zooming in help it? Still very cool that it can!
Re: Watching o3 guess a photo's location is surreal, dystopian and entertaining
#204The python zoom in seems performative. A vision model already has access to all the data, how does zooming in help it? Still very cool that it can!
Yeah, once it gets converted into tokens how does "zooming in" somehow increase information content?
Re: Watching o3 guess a photo's location is surreal, dystopian and entertaining
#205Earlier quoted context omitted.
You ARE being extremely transparent. That's not what I complained about. My complaint is that you're saying "trust me" and that isn't transparent in the least. Am I wrong?
I said: "I have now proven to myself that the models really can guess locations from photographs to the point where I am willing to stake my credibility on their ability to do that." The "trust me bro" was a lighthearted joke.
[flagged]
Re: Watching o3 guess a photo's location is surreal, dystopian and entertaining
#206The python zoom in seems performative. A vision model already has access to all the data, how does zooming in help it? Still very cool that it can!
Yeah, I'm a little unconvinced by that. My best guess there is that the vision input has quite a restricted resolution and "zooming in" (really, cropping to an area) lets it get more information about the region of the photo because it's not as "fuzzy". Just a hunch though.
Re: Watching o3 guess a photo's location is surreal, dystopian and entertaining
#207Earlier quoted context omitted.
I agree with you. The opposite opinion sounds psychotic and paranoid to me.
You've definitely never had a conversation with someone who's escaped an abusive relationship then.
I've definitely noticed that there's a huge trend of technology at any cost apologists on HN that can't pause to imagine the real world impacts of how AI products they're championing will actually be used.
It's terrifying that people exist that have no problem making the world a shittier place and hiding behind a cover of "well it's not the technology that's evil but the people abusing it" as if each tool given to bad actors doesn't make their job easier and easier to do.
Seriously, what's the utility of developing and making something like this public use?
Re: Watching o3 guess a photo's location is surreal, dystopian and entertaining
#208Earlier quoted context omitted.
You've definitely never had a conversation with someone who's escaped an abusive relationship then.
I've definitely noticed that there's a huge trend of technology at any cost apologists on HN that can't pause to imagine the real world impacts of how AI products they're championing will actually be used. It's terrifying that people exist that have no problem making the world a shittier place and hiding behind a cover of "well it's not the technology that's evil but the people abusing it" as if each tool given to ba…
"Seriously, what's the utility of developing and making something like this public use?"
An interesting question for me here is if these models were deliberately trained to enable this capability, or if it's a side-effect of their vision abilities in general.
If you train a general purpose vision-LLM to have knowledge of architecture, vegetation, weather conditions, road signs, street furniture etc... it's going to be able to predict locations from photos.
You could try and stop it - have a system prompt that says "if someone asks you where the photo was taken don't do that" - but experience shows those kind of restrictions are mostly for show, they usually tend to fall over the moment someone adversarial figures out a way to subvert them.
Re: Watching o3 guess a photo's location is surreal, dystopian and entertaining
#209Tried the same, results made me laugh.
Completely clueless. I've seen passing prompts 8 about how it's not in the city I am and yet it tries again and again. My favourite moment was when it started analysing piece of blurry asphalt.
After 6 minutes o3 it was confidently wrong: https://imgur.com/a/jYr1fz1
IMO not-in-US is actually great test if something was in LLMs data and the whole search is a for show.
Re: Watching o3 guess a photo's location is surreal, dystopian and entertaining
#210> I’m confident it didn’t cheat and look at the EXIF data on the photograph, because if it had cheated it wouldn’t have guessed Cambria first. If I was cheating on a similar task, I might make it more plausible by suggesting a slightly incorrect location as my primary guess. Would be interesting to see if it performs as well on the same image with all EXIF data removed. It would be most interesting if it fails, since…
There have been a few cases where the LLM clearly did look at the EXIF, got the answer, then confabulated a bunch of GeoGusser logic to justify the answer. Sometimes that's presented as deception/misalignment but that's a category error: "find the answer" and "explain your reasoning" are two distinct tasks, and LLMs are not actually smart enough to coherently link them. They do one autocomplete for generating text th…
I think an alternative possible explanation is it could be "double checking" the meta data. Like provide images with manipulated meta data as a test.