I'm amazed that such stupid stuff takes up attention, and more importantly, takes up my attention. Thankfully it's just a minute or two, but clearly I need to be more careful about what I click on (and let into my head...I'm sure this type of thing degrades bayesian priors for everyone reading it).
> I'm sure this type of thing degrades bayesian priors for everyone reading it
What kind of thing do you mean by "this type of thing"? Why do you think it degrades bayesian priors?
Life is, among other things, a constant quest to better understand the world around us. For every "real discovery" there is a ton of noise. Most of that noise can be easily discarded, some look more convincing. This particular instance looked more convincing.
We know that modern imaging equipment is very complex. You shoot a picture and complicated processes happen which were optimised by thousands of engineers to produce a good looking photo. In the past we have seen that such complex pipelines can introduce plausible looking artefacts in a different context: https://www.dkriesel.com/en/blog/2013/0802_xerox-workcentres...
This is why it could be possible that maybe something similar happens with the camera app. People investigated, and it turned out that that is not the case.
For every breakthrough in our understanding there were thousands of perhaps and maybes. For every eureka moment, there is a bin full of superluminal neutrino experiments. This is normal and part of the process of reasoning about the world.
If your thinking processes get damaged by encountering as of yet unexplained phenomenon then your brain is broken and you should ask for a refund.
> They just said wrong data makes them (and others) more likely to make mistakes in the future.
First of all, there is no wrong data here. Twitter user X said their iphone did a funny thing. Twitter user X attached a snippet of the image and said they think the iphone can overlay leafs in certain circumstances over faces. That is it. If your conclusion from that is that "iphones has AI magic which overlays leafs on faces for sure" that is your hasty conclusion and it is on you. The data is not wrong, your conclusion is wrong.
> None of this weird strawman.
:( sombremesa drawn a hasty conclusion. When they realised that it is unsupported they said (direct quote) "I need to be more careful about what I click on (and let into my head..."
Which basically says: I should be less curious because look here, I ended up believing the wrong thing.
The problem is not being curious, but drawing hasty conclusions and believing the wrong thing.
They should stay curious, let a lot of information in their head, but engage in critical thinking. Read the weird twitter thread purporting to show the weird bug and think: What if the twitter user is mistaken? Is it possible that the picture doesn't show what they think it shows? What other possible explanations there are? And so on and so on.
That strategy is much more widely applicable, more robust against even intentional deceptions and leads to ultimately better understanding of the world.
I think we bought into it because it looks plausible. In the original image[0], the boundary between the leaf and the person's face looks legitimately blended in a way that only AI auto-filling does. From that, it's not a leap to extrapolate that the leaf itself was auto-filled from the surroundings.
Yeah, it has that very distinct blurry mess look that you get with all the AI stuff that is always conveniently skipped over.
And frankly looking at the original picture again, focus is no excuse for that. Yes, there is a leaf, no, the AI whizbang still blurred it all to crap.
Original image was using digital zoom, so working at the limit (or beyond) what the sensor could provide. Image processing is just doing the best it can with very little input.
The blurriness and "oil painting effect" likely has more to do with aggressive noise reduction and the fact that small lenses usually aren't super sharp to begin with.
>but that couldn't more clearly be a leaf connected to a branch
Yes, the leaf itself is "clear" but what threw people off was the blurry and dark halo surrounding that leaf which looks like manipulation artifacts.
When you use Photoshop tools like healing brush, clone/stamp tool, or A.I. Content Aware Fill, you often get strange visual artifacts like that. Here's a quick example of erasing a person that leaves behind blurry artifacts:
https://www.youtube.com/watch?v=eC4TsTRTHiY&t=1m42s
> but what threw people off was the blurry and dark halo surrounding that leaf which looks like manipulation artifacts.
That halo is pretty clearly what's left of the subject's head after the phone has tried to get rid of it. It's difficult for me to understand how this wouldn't qualify as "replacing someone's head with a leaf". Sure, the leaf is there. But the head used to be there too, and in the photo it's been removed.
If it’s not clear to you, the original claim probably sets critical context. The first thread was spawned by the photographer claiming there were no leaves in the foreground, and that the post processing software had picked some random background content and used that in place of a human face.
Clearly, what you say is in some sense true; digital zoom and post processing have left some artifacts. To me it sort of looks like there’s another leaf on the same branch as the first.
But when you are digitally zooming a tiny iPhone camera to photograph someone 50-100ft away and there’s multiple depths captured, it seems hard to imagine how the photograph would be crystal clear in that circumstance, and even further it’s crystal clear the post processing only failed to add clarity, and it certainly did not introduce a nonexistent leaf / twig / branch connected realistically as some software bug.
I fail to see how the similarity of those “artifacts” would imply that the camera software added a leaf… and a branch… and connected them realistically… in the foreground… in place of the face it apparently “replaced”
A simple application of Occam’s razor could’ve saved a lot of furor
Well, that's the crux of the issue that you don't seem to see. For many observers, it does not look realistic. The unnaturalness of the pixels is what makes it plausible for Apple software to have a bug. If it looked 100% realistic to everyone, we wouldn't be in this thread discussing it.
EDIT reply to: >, especially when the image so clearly depicts that.
You still seem to be missing why this discussion of AI error was even possible. In this very thread, there are commenters who still don't see "clear depictions" of a realistic foreground leaf even after being made aware of parallax demonstration in the update... because the blurriness keeps hijacking the brain to make it look like a fake artifact. Consider that the actual person who took the original photo didn't realize it was a real foreground leaf. (Or so he claims.)
In other words, saying "Am I crazy? I don't see why there's debate about the dress being blue or white when ... it is clearly and realistically BLUE" -- doesn't actually add to the discussion. Hey, it's great that the foreground leaf looks realistic to you. But your personal perception is not
relevant to why others perceived it as an artifact.
I don’t think you read what I wrote, so I’ll say it again: claiming some image processing AI introduced an entire branch, twig, leaf construct in place of a face is comical, especially when the image so clearly depicts that.
I’m much more likely to interpret this thing as anti-capitalist sentiment being expressed in a ridiculous fashion than an actual good faith argument there could be a bug so rare but so profound that people’s faces are mistaken for the background and the AI’s post hoc justification is that there is a branch and leaf there so it’ll output that. It doesn’t even make sense what is being claimed here, your links don’t really change that
Edit: I see a leaf in the foreground, and the situation was a highly (digitally) zoomed picture with a tree in the foreground. This has nothing to do with color perception like you imply; the fact that a bunch of people - whose perception was previously influenced to believe in AI error - don’t think it shows the exact literal situation it correctly shows hardly makes it a relevant discussion
Edit 2: what I see might not be relevant, but the camera software showed the situation correctly, and you’re still arguing because the digital zoom produced artifacts that the entire branch / twig / leaf that was literally there might NOT have literally been there and could have been added by some magic unexplained AI replacement.
It doesn’t even make sense. There is no room for human perception here. If the software was rendering leaves instead of faces, it wouldn’t be a single guy finding that scenario, and especially not when the argument the leaves didn’t exist in the first place was based on compressed video. Comical stuff. This is literally conspiracy theory argumentation from you
You keep repeating "I". Again, the emphasis on "I" in your sentence is not relevant to the commenter right above you perceiving something else: "The blurry plus the tiny twigs connected to those leaves still make my brain think it's some GAN artifacts."
EDIT reply to: >This is honestly blowing my mind that you think theory of mind applies
Let me try to put it this way... nawgz sees a clear and realistic leaf in the foreground. Case closed. Therefore, nawgz would not even submit a Twitter post or a new thread to HN wonder why the head is missing. And yet, such a thread exists with (some) commenters wondering if the software had a glitch:
https://qht.co/item?id=29739235
Stepping outside of yourself and considering Theory Of Mind might help explain to you why such a thread exists with comments even though you yourself see no issue with the photo.
EDIT2: >There is no room for human perception here.
There is room for human perception because the person who took the original iPhone photo and posted it on Twitter is a senior iOS programmer with a degree in computer science and yet he overlooked the possibility of a real leaf. If we take his story at face value, his particular mind and perception initially chose to believe that iOS photo software replaced the head with a leaf rather than consider a scenario that a tree branch in front of his own house occluded the walker's head. Only when another programmer on Twitter theorized that a real leaf is in the foreground, he then went back outside again to notice that there's a low hanging branch on his tree. (That's his later video showing parallax demonstration.) The original photographer did not believe the pixels in his own photo. That's a human perception issue.
That faulty perception led to software theories from some programmers in that Twitter thread including this comment:
>When you open iPhone camera, it keeps recording. When you take a shot, it takes some frames from before and after you pressed the shutter and then merges those to create a detailed photo. One of this frame had this leaf on her head. -- from https://twitter.com/masnun/status/1476457354128281600?s=20
There is a massive difference between a perception debate about color - infamously tricky - vs whether or not a leaf existed. This is honestly blowing my mind that you think theory of mind applies to digital zoom not being able to resolve details at multiple distances flawlessly. The AI replacing a face with a leaf is also a comically circumvent construct to explain those artifacts
Edit: I explained how theory of mind applies - a bunch of anti-capitalist or anti-Apple folks who think that magical AI is everywhere broke this photo spectacularly. I don’t really care why people think it’s something that it’s clearly not; people think covid vaccines are meant to shorten lifespans, or that climate change is a hoax, or that god exists.
None of the justifications posted are even close to relevant or believable at a glance, let alone after seeing the reality of the thing
Yes, well said, it’s a human perception issue; even smart guys can do really dumb things, like accuse image processing software of adding a leaf, even though the leaf is shown to attach to a tree that is indeed right in front of this person’s house.
What this isn’t is a software issue, or something that was reasonable to point to as one; after all, the software was exactly correct, and to an unbiased observer in an immediately obvious way.
In other words our brains took our knowledge about the background of the situation (new phone tech rolling out recently, ML features, etc.) and auto-filled a scenario that covered up the underlying reality here. So, our brains were the ones doing the reality augmentation.
It looks like two images—one with the face obscured, one with the face visible—were combined. Notice that both the leaf and the person are sharply in focus, and there’s a distinct circular, blurry outline around the face where a transition occurs between the two images.
No, the leaf in the photo isn’t fabricated by iOS. No, iOS didn’t copy a leaf from elsewhere in the image to continue the pattern. But the conclusion, “iPhone Camera app did not replace person’s head with a leaf” is probably incorrect: it looks to me like it did, in fact, replace the person’s head with a leaf. It just so happens that the leaf was really there from the perspective of one lens or photo.
Wasn’t this documented as intended behavior on newer iPhones? Images from different lenses will be combined and faces replaced with clearer versions in some scenarios. In this particular instance, the “clearer” version was a leaf, as there was a leaf in the way.
This isn’t really bad thing; it’s an understandable bug in a piece of technology that usually works amazingly well. However, it does mean there’s room for improvement.
> No, the leaf in the photo isn’t fabricated by iOS. No, iOS didn’t copy a leaf from elsewhere in the image to continue the pattern.
Those two things are exactly what everyone was assuming happened. So while you're right about what is going on with the camera behavior, I think you're disagreeing about a completely separate conclusion than everyone else arrived at.
> It just so happens that the leaf was really there from the perspective of one lens or photo.
In the original thread the photographer was fairly certain there weren't leaves in front, which is what started all of this.
>> No, the leaf in the photo isn’t fabricated by iOS. No, iOS didn’t copy a leaf from elsewhere in the image to continue the pattern.
> Those two things are exactly what everyone was assuming happened.
What? Look at the other discussion thread:
> The cause of this is image-stacking.
> The phone takes ~20 frames, over 0.2 seconds. In that time, lots of people and things in the frame move.
> Optical flow is used to track all moving parts of the image, and then 'undo' any movement, aligning all parts of the image.
> Then the frames are combined, usually by, for each pixel, taking something like the median or throwing out outliers and using the average.
> When the optical flow fails to track an object in more than half the frames, the 'outliers' that are thrown out can in fact be the image content you wanted.
> It happens with leaves a lot because they can flutter fast from one frame to the next, so tracking each individual leaf is hard. A few bad tracking results on more than half the frames, and all you end up seeing is leaves where there should be a face..
There was nobody arguing that the phone hallucinated the leaf. The problem is that the phone removed the person's head so that it wouldn't pull focus from the leaf.
I meant the original Twitter thread. But even here everyone was wondering where the leaf came from because the original photographer was insistent there weren't any in the foreground.
The general consensus was that yes, it was image stacking, but perhaps it was moving a leaf that was in the background over her face. That turned out not to be the case after the updated tweet mentioned there was a tree in the foreground.
Never underestimate the power of starting your root cause analysis with the assumption that it's pilot error! At least, when doing my own root cause analysis, it has saved me a lot of embarrassment to assume that I did something stupid and to rigorously rule out all possibilities before I blame someone else.
Or really starting your root cause analysis without being open enough to a wide enough set of possibilities.
As a fairly trivial example, years ago now I wrote a little article observing that a Sony ad boasting about how responsive their camera was (at a time when many digital cameras had a distinct delay) was using a rather famous wildlife photo shot on film. I also noticed that it wasn't quite the famous frame though it was still an excellent photo. Odd. Maybe something about rights.
You probably see where this is going. Subsequently I saw other ads in the series which had clearly mistimed versions of other famous photos. At which point the thing I had never considered clicked--to my somewhat embarrassment.
(In my defense, the wildlife photo used was certainly not an obvious example of bad timing unless you were familiar with the original and even then it was subtle.)
This makes some of the "informed opinions" in the recent discussion hilarious to read in retrospect, considering that the event reasoned about in those comments didn't actually happen to begin with.
Happens all the time, what's fascinating is that in the original discussion there are people who came to the conclusion that it's just a leaf that is in front of her face, either one that is in the process of falling down or one that was pushed right in place by the wind...
The people who made such comments were never replied to, and while I can't see upvotes my suspicion given how far down you have to scroll to read them is that no one cared to give those possibilities much credibility.
I feel like it would have saved everyone a lot of time if he had included the full photo in the original tweet. You can see it here in this thread by @sdw [1], and it's much more apparent that there's a tree in front with leaves hanging off it.
Yeah I don't understand everyone going "oh, that explains it". That is not a photo of a leaf in front of a face. A foreground leaf is involved, but it's not large enough to obscure a face. There's also a leaf-textured halo of brown which looks like it was once someone's hair.
Yes, it's clearly still a photo that has been messed up by the phone in terms of some kind of artificial image replacement, smoothing and blurring. We just have slightly more context for where the big leaf in front of the face came from.
You mean people haven't been holding the leaning tower of Pisa all this time. I am shocked!
I've seen some strange things from stitching photos for panoramic, it isn't a huge leap of logic to think some compression or interpolation algorithm went wrong.
I wouldn’t. We do know, indeed boast, that AI is used to boost the apparent performance of consumer grade cameras and especially iPhones. There is that example of a photocopier changing numbers in the images it copies. Illusions are reasonably likely to be either in the human or the machine vision systems.
I think a better question is how many people would independently develop the accusation if they were presented with the original photo without comment?
Yes. In an area with lots of leaves, not unreasonable to think motion or parallax unexpectedly included a leaf - vs some wild convergence of leading edge technological bugs perfectly rendering content in a system which shouldn’t.
Definitely. I'm into photography, although I'm not by any means an expert, and when I first saw this news I just ignored it thinking there were so much going on, and the cause might be anything simple, it didn't really spark my curiosity.
I thought maybe it was just some processing glitch, maybe it was computational photography messing up, maybe it was just a question of focus, or maybe a mix of things.
I mean, I've seen dumber reactions to a leaf. I've seen dozens of cryptid enthusiasts argue about the nature of the bright orange Sasquatch that was living in the woods near a nature camera because of the blurred image of a falling leaf.
isn't it more likely that they're optimizing for their users? The goal is to have a product where people are scrolling past a multitude of media content quickly. Having very small versions of that media immediately display is the UX their users benefit from.
Definitely this. I still wish they’d have a “show original” button or even detect when you’re staying on an image/zooming in and download the full res then.
Could someone give an ELI5 please? As I understand it, the explanation is that there was a leaf close to him that obscured the face. But what's he saying about parallax? Isn't the preview image that he would have seen on the camera display coming from the same lens that captures the image?
The parallax in the video is just a tool since photographs don't have any real depth perception like our eyes do. The small leaves on the foreground and the big leaves in the background enhance the illusion, since they're about the same size in screenspace. Maybe the preview showed a leaf, but it's close enough to skintone I doubt it stuck out to his brain on the little phone display.
In meatspace, his eyes are at a different vantage point focusing on the face, while cellphone cameras try to keep everything visible as sharp as possible.
It wasn't the AI that did the Auto-filling this past day, it a large mass of humanity that auto-filled an incident report into a plausible outline of a story. I think this will happen more and more once AI reaches a point of higher cognition.
Out of the hundreds of thousands of people who saw the original accusation, only a few will see the follow-up. It will forever be known that iPhones replace people's heads with leaves. This is how urban legends are born.
This is one of the few times I'll ever feel sorry for Apple! I can only imagine the calls they're going to get from Karen's who didn't take good pictures to begin with and blame Apple for it because they saw an internet that one time that somebody's iPhone replaced a head with some leaves.
It's quite amusing though that so many brain cycles were wasted trying to figure out how ML image processing was to blame. I don't recall anyone in the previous thread saying: maybe there's a leaf between the camera and the woman's head. It certainly didn't occur to me that it was something so obvious.
(btw, is the "Karen" reference really necessary? It's a sexist, racist insult that detracts from an otherwise interesting comment)
Nobody dare to insult saying that the leaves must have been in front of a face, just like nobody would dare to say to "google search doesn't work on my computer!" - "did you turn it on?".
Still interesting to see how people can rush to blame ML post-processing over a trivially overlooked parallax effect even when the former argument/theory may still potentially exist.
> btw, is the "Karen" reference really necessary
No, it wasn't necessary but at this point, what isn't considered a slur anymore?
Karen is a pejorative term for a white woman perceived as entitled or demanding beyond the scope of what is normal [1] so it is just as "racist" as many other "racist" things. If "Karen" isn't racist then neither are many - if not most - things which are deemed racist. I'm fine with that so I hope you're sincere and are willing to follow this old-new definition of racism. If the term goes back to where it belongs - the assumption of people's traits based on their physical appearance with a heavy emphasis on skin tone - and it ceases to be bandied around as a label for anything which happens to displease the one doing the labelling the world will become a better place.
I sincerely appreciate your comment! With that definition in mind then I can understand why people downvoted my comment so much.
To anyone I offended, I sincerely apologize.
The truth of the matter is that I'll call anybody a Karen who is dumb enough to deserve the label. I don't care about color they are I don't care where they're from I don't care what they look like or how their hairstyle is done or what they're wearing or how many kids they have or how rich they are or how poor they are. I could have just as easily said a***e.
But I chose Karen because people who act like a Karen are idiots.
White people, particularly white women. I'm fairly sure you knew this? Here's the definition of the pejorative "Karen" [1]:
Karen is a pejorative term for a white woman perceived as entitled or demanding beyond the scope of what is normal. The term is a meme depicting white women who use their privilege to demand their own way.
> is the "Karen" reference really necessary? It's a sexist, racist insult
And also misused quite frequently. Its origin was a white woman who called the police on a group of black people who were doing nothing wrong other than being black in the wrong place.
I don't know where you heard that, but for a long time it was just a fill in name for a middle aged woman with a short hairstyle who would usually cause issues at customer service in retail. The "I'd like to see your manager" and the Karen stereotype go way further back. It's not usually about race with a Karen it's about power and getting their way even if that's not how the rules are set up. (This coupon is a year expired but I want it, this TV was marked for a dollar so you have to sell it to me, etc.)
Incorrect. Karen was originally someone who tried to wield power over retail workers, and always demanded a manager. The white lady calling police on black people usage came wayyy later.
Now it's been abused and overused to where I don't even know what it means anymore. Lady comes outside to ask TikTokers to quit doing burnouts in her yard and it's 'shutup kawen, stupid kawen LOL amirite'
I know exactly what the origin is (and as somebody commented below, it predates the incident you refer to). I also used it because it aptly describes people whose exaggerated sense of entitlement and privilege overshadows their desire to think critically.
My daughter saw this in person, where a woman was mad at people at the butcher section of the grocery store, calling woman who bought a whole chicken an 'animal murderer'. Plot twist: the very same woman had chicken breasts and thighs in her cart, but because they were not a whole chicken, she disassociated them from the animal they came from. I rest my case.
you can’t very well say a meme is misused, a meme is whatever it is.
Also, central park karen is a recent addition to the long history of using karen as a punchline, anytime you think you know the origin story, check know your meme first, you might be surprised: https://knowyourmeme.com/memes/karen
So to make it meme it gets abstracted to "some self absorbed busybody stirring up unnecessary trouble for other people" or something similar. Nothing really racist or sexist about it at that point is there?
Its origins are "racist" and "sexist" in that the stereotype is specifically a white woman.
I don't think there's anything wrong with a little stereotyping: the only reason it's "bad" is that actual bigots ruin the fun, so to speak. If anything, I feel bad for people named Karen.
No it wasn’t- there was just a heavy intersection between middle aged women complaining and BLM + other wokeness. “Karen” is used correctly here and your origin story is completely false.
Where do you see men okay the victimization card at the drop of a pin? As far as I know there are as rare as hen's teeth, I have not seen a single example of either. If you have some examples I'd be interested in learning about them.
This is not just urban legends. In Hungary there's a soft dictorship where Viktor Orban is controlling the mainstream media with lots of lies. After independent organizations go to court, they usually win against Orban, and the media organizations publish corrections, but nobody reads them.
OMG, i was reading Quora today thinking exactly about what you wrote in the first paragraph. but in a different context.
there is that false story that gathered tons of likes and comments, even though a buried comment debunked the whole thing. the worst part is that it's since been shared in many Quora spaces. where it is gathering more likes and comments.
i stopped for a minute and wondered if the situation could have been prevented.
It's also an example of selection effects: this dude mentions he's taken 44k photos. You might think that such an experienced photographer wouldn't be fooled by this - "what, it was just a leaf in the way when he was zoomed in? that's all? surely he wouldn't make such a mistake, and it's surely machine learning at fault!" - but he's only human, and that's 44k+ chances to be wrong, for just a single photographer, where social media will amplify the oddest anecdote it can find across every photographer in the world.
It's not, "he took one photo and it came out wrong", it's, "humans everywhere took billions upon billions of photos and you only were told about the weirdest ones which fit a Narrative".
Looking at the original, yes, something leafy is in the foreground; however, the person’s visage and locks are made into cubist/polygonal-like leafiness in the background.
The (a of course, but the one originally involved) responded 'it's a very cool lesson in focal length compression' though; so it seems to me from that & still from looking at the image that the iPhone is doing something - making choices, engineering trade-offs - that makes this accidental edge case worse.
This case actually seems even simpler though. There was a leaf in the foreground obscuring the face from the camera's perspective. Distortion doesn't quite enter into it.
They're not talking about the obscuring itself, but the focal compression effect which makes it look a lot like the leaf is where the head should be, rather than clearly being in front of said head.
Though the aggressive denoising and small lens probably doesn't help.
Oh I didn't realise that was called 'compression', makes sense though, I was just thinking too software-y! Don't modern iPhones have multiple cameras for depth of field though, it could (or intends to) do something clever to make it clearer couldn't (resp. ) it?
I think it also speaks greatly to people's paranoia about technology they don't fully understand.
I think a lot of the HN crowd are susceptible to that because they know just enough to know it's possible but are not perhaps familiar with a given domain (I certainly know that applies to me).
It's perhaps some spin-off of the dunning-kruger observation... There must be a spot on the curve where a person knows just enough to be freaked out by what they're seeing, but not enough to explain it fully. Whereas here the Halide devs know the domain well enough to quickly diagnose it.
> I think a lot of the HN crowd are susceptible to that because they know just enough to know it's possible
Absolutely think this describes what happened
I'd have probably been more skeptical myself had I not seen an ad earlier in the day of an Android phone removing people from the background of an image as easily as you'd crop it. Whoops!
Eh, I mean you’re right that more people will hear the accusation than the retraction, but I think everyone already knows and will continue to know that iPhones have excellent cameras. I wouldn’t be worried about the reputation of iPhone cameras.
It's less about the reputation of the cameras but there was an unhealthy level of conspiracy in the original thread with regards to trust of digital images in general.
That post was already being widely used to justify the fairly ludite views on image upscaling by the judge in the Rittenhouse trial for example.
The judge didn't merely ask whether any such thing was used. He continuously pressed his own uninformed opinion over the evidence rather than trying to educate himself on the matter, especially given "expert" testimony
Interpolation algorithms (such as bicubic) used for zooming are not “AI” anymore than a decompression algorithm is. There’s no “training” involved; it’s an entirely “pure” algorithm. Same input in gets the exact same output out.
There was no AI upscaling. If you enlarge an image between two white pixels, the inserted pixel is white, you will not get an interpolated black pixel.
And in the opposite corner, we have valid concerns about the "just fill in the gaps" approach to photography getting shouted down by "have you missed the update? That was an actual leaf, not an iPhone dreaming!"
Dangerous precedent works both ways, I don't know which effect will be stronger.
But the iPhone has replaced her face with leaves. If you look at the photo. The place where her face and neck would be is replace with a blurred out brownness even on the parts not directly obstructed by the leaf.
> This is one of the few times I'll ever feel sorry for Apple!
They're a trillion dollar company who don't give even the slightest little fuck about you. Stop feeling sorry for them, that is an insanely unhealthy way of thinking.
It’s the race stuff that now has me laughing. The machine learning had a data set that had too many white people and too few people of other colours (leaf colours?). Somehow this was relevant.
Imagine what a person's head would look like in relation to the size of the body. Where has the head gone? What's the nebulous dark stuff textured like leaves where the person's hair should be?
I am happy that the iPhone didn't invent the leaf, but it has pretty obviously done something weird with it.
