AI camera with no lens
theprompt.io
theprompt.io
Fun to the see a modern reincarnation of that idea.
(While digging around to find the above, I did find yet another camera project that does the opposite: "Matt Richardson's "Descriptive Camera" sends your pictures to Amazon's Mechanical Turk and jobs out the task of writing a brief description of each image, then outputs the text on a thermal printer. It's a camera that captures descriptions, not pictures." (https://boingboing.net/2012/04/25/descriptive-camera-prints-...)
https://philippschmitt.com/archive/2018/work/camera-restrict...
At a hilarious intersection of the present concerns about AI, bias, and prejudice (especially as voiced in the current front-page article^1 about a different blind camera project, from 2022), and the state of internet search engines, given the query [blind camera art project], Bing and Google find only articles about visually impaired photographers, but Yandex finds an entire results page of articles about the work we're looking for! What gives?
I've found often that the gut feel that makes you take the shot doesn't necessarily know when you have the right composition or what the subject of the shot actually is, you just hit he shutter, you know that this was the shot.
Then, when looking at the shots, you have all the time and the world to analyse and find meaning and beauty in this sliver of an instant.
By replacing this by a random seed, a 20 word prompt and gps localisation, I doubt that anyone would have a personal connection to the image, or to the instant it was taken. It become a "clean", "sanitized" image, that's only esthetic (or arguably memetic depending on your prompt), and is wholly separate from the person that took it.
You also lose all of the information that you can not consciously perceive while taking the shot/writing the prompt, since you filter what you see through the lens of language, and then back into visual.
It's neat !
For example it's easy to make a photo look like it was taken from a non-existing tower if you crop the upper part of a regular photo. Or you can focus on details that you always skip over because there's something more eye-grabbing nearby.
This is also why I love to see photos made by people who visited my city for the first time. They don't know which parts are pretty so they capture stuff I wouldn't think to photograph.
This always been my idea of what makes a great photograph.
I assume I could develop it with practice. I just never did. I rarely had a camera growing up, and now that I have one with me all of the time, I treat it like the Instamatics I used to have. The pictures are terrible.
At best, they're a kind of bookmark that I was at that place and saw that thing. It won't have an emotional resonance for anybody but me, and for me it's just bringing up the much better picture in my head. If they want a good picture of the thing, I'll go find one that somebody else took.
All of which is to say... I really admire good photographers. I respect their work, and the diligence it took to develop their eye.
This is, as you note, an interesting art piece on that same subject. I'm afraid I'm better with words, so this is mine.
You need to stay there, move around, explore the space, the light, the interaction with the living.
Eventually you find a good way to tell the story of the place. You get lucky. Sometimes it's fast, sometimes it isn't.
I didn't take pictures I was proud of before going to protests to take pictures of people there, and trying to frame them as "badass" as possible.
Now I mostly shoot animals and landscapes, but I've been doing it for years at every oportunity I get, so my "gut feel" is strong there. I'm not great at portraits, for example, and have 0 interest in street shooting.
3D printed case. Resting on a table. Witness describes the suspect. 10 seconds later it prints out an AI generated ink sketch.
I mean why not? Sell it through to every police station in your state. You can even put a cute little police badge emblem on the case.
So what happens if you now generate a photorealistic "sketch" based on a description? Are officers going to be sufficiently aware to know that's not a actual photo of the guy they are looking for, and act accordingly? Or is it going to heavily bias a manhunt effort? moreover, what happens when the photo randomly ends up close to someone present in the dataset?
Driven by a little knob selector.
Featuring 3 art styles: 1. Police sketch classic 2. Realistic photo 3. Manga character
would be crazy, probably a movie plot somewhere
"White Male, Curly hair, mole on face"
Generate.
"Good, but he had a larger nose, and blue eyes."
Generate.
"He was a bit more gaunt, and had some stubble."
Generate.
"Nearly there. More pronounced check bones, and make the jaw a bit softer"
Generate.
In 5 minutes or less, you could get a near exact picture of the potential criminal; something that might take up to an hour or more normally with a professional police sketch artist, and it could easily be in 3D too. There's tremendous value in that.
Even victims themselves are famously bad at identifying criminals.
These types of problems are very widespread - it's not rare that people misremember details because of the stress and trauma, and it's also well known that the process of describing/asking questions can cause bias into the victim, as seen in the many cases of people admitting crimes they didn't commit after long interogations.
I've also heard that the quality of police sketches was highly related to the person making the sketch, some have high correlation rate but that is not the norm i.e.: the average sketch artist might not be reliable on average.
JCS Psychology on Youtube is a great channel showing the processes happening during interrogations, if you're interested.
People have bad memories and bad perception in stressful situations. They don't actually know what the person looked like; they don't have a strong model in their brains. Police sketchers use clever questioning techniques to get details about features that people wouldn't otherwise think to describe or even realize they have knowledge about. The truth is that there is an absolute limitation to the effectiveness of any facial image reconstruction, which is the limits of human memory. Adding AI to the mix can't change that, but it's extremely likely to influence the witness to describing a less accurate face with higher confidence. In other words, a disaster.
In fairness, with the ubiquity of cameras, sketches are much less required...
Please draw your attention to the discussion about the witness in the process of image generation. For example:
Officer: "Could you describe the man who attacked you, miss."
Witness: "Well, he had ...eyes, a ... forehead, and ..."
<here's the impotent part for you, _lady>
Officer grabs the first rendering from the machine and shows it to the witness: "Did he look like this?"
Witness: "No, his eyes were set further apart."
Whir, whir, the machine prints another image.
Officer: "More like this, then?"
And so on...
In the scenario I described, I'm not sure where a new source of racism is introduced.
Help me see this differently.
It can't think, or form opinions. It's not "intelligent" in any real sense.
It's just Eliza with a really, *really* big array of canned responses to interpolate between.
So, just like people, then.
> It can't think, or form opinions. It's not "intelligent" in any real sense.
Honest question, what is the purpose of this comment? What is the change you want to see coming out of this semantic argument?
The argument is that the output is racially discriminatory for a variety of reasons and it's easier to just say "it's racist" than "Many of the datasets that AI is trained on under- or over-represent many ethnic groups" and then dive into the details there.
Imagine that a police officer is looking for someone matching the image but doesn't know that it's hallucinated from a vague description, they could let the real suspect go or incorrectly arrest someone who happens to look like the AI generated image but otherwise doesn't have any reason to be a suspect.
Police are already greatly overestimating the accuracy of their own facial recognition tools because they don't realize the limits of the technology, and this would just be worse.
That's not a necessary property of AI image generation. You could just add a 'output as a sketch' system prompt.
the progression of technological "AI" has just been the automation and acceleration of their logic and operations
what paperclips are the police maximizing?
everything the alarmists are afraid of has already happened
Might make a neat like coin op charicature thing though.
Somewhat tangential: the "part of the brain" that is responsible for recognizing faces is incredibly well developed. That "peek-a-boo" game that you play with children? Every time you uncover your face millions of neurons in the childs brain suddenly fire giving them a jolt of "joy". The face recognition is so developed that we tend to see faces were there are none (face pareidolia).
... the point being that the brain does a lot of unconscious work recognizing individuals. Describing those individuals later consciously is pretty error prone.
Jokes aside, I think this demonstrates that AI generation isn't too great if you have something very specific in mind, at least it looks like the generated picture deviates from the real one, though it's impressive that it's still so extremely similar.
The virtual one doesn't load for me unfortunately
The source linked from the tweet is https://bjoernkarmann.dk/project/paragraphica, but it seems to be not responding.
Neat if at all possible, probably not. Perhaps make a cheap lens work as a good one, with some calibration on known images?
I guess it's a "camera" as much as 3D software has "cameras" for controlling what the viewport is pointing towards.
So is the midjourney discord bot a camera? Microsoft paint? Your printer?
You could say its kinda a camera because it takes pictures are your location. But it is in no way seeing anything (obviously, because there is no lense). It's just sliding some preset parameters around based on location.
Again, not that its not impressive, but camera seems like the wrong word for it.
https://bjoernkarmann.dk/project/paragraphica
The author's site is currently struggling to load, so here is an archive link:
This isn't really a camera, it's a GPS hooked up to Stable Diffusion.
“ The star-nosed mole, which lives and hunts underground, finds light useless. Consequently, it has evolved to perceive the world through its finger-like antennae, granting it an unusual and intelligent way of "seeing." This amazing animal became the perfect metaphor and inspiration for how empathizing with other intelligences and the way they perceive the world can be nearly impossible to imagine from a human perspective.”
When Stable Diffusion was released, and I saw the whole model was ~4 GB.. I instantly thought how insane it would be if it somehow was possible to take the model and a compiled binary of the inference code for x86_64 (without any modern extensions) on a DVD back in time, say around 2005-2006ish, and the implications that would have, psychologically, on the world.
You could load that model on a moderate desktop with a 64 bit Core 2 Duo and 8GB of ram and let it chug .. without GPU acceleration, on CPU only it would take ~2 hours to make an image. But it would do it without an internet connection, without any inspectable code or heuristics... just... numbers, spitting out an image from text of whatever_people_want.
It would be called a hoax. (In fact, I came across people on reddit when Dalle2 came out claiming it was somehow a trick or a hoax, and that all the images it produced must be existing beforehand somehow and prerendered).
Scientists who dissected the weights file and the machine code for the inference engine would eventually figure out it was a neural net, but how such a net was trained would be a complete mystery. Theories involving aliens would likely appear.
I wonder if it would be allowed to be made public, just the knowledge that such a thing was working. It would scare people, I think. Having it make these images without anyone knowing how.
Hell, it is kinda scary now, ever knowing how it all works.