That fake photo of Maduro getting capturedThat fake photo of Maduro getting captured
I woke up this morning and saw this photo in the news:
Obviously fake, right?
Also I saw this photo:
Still fake. I felt pretty good that the denizens of the internet weren't going to pull one over on me.
And then I saw this photo:
Mostly, I assumed it was fake because all the others were, but I didn't really have anything else to judge by.
But then there was also this one:
And this one:
And this variation:
And then I saw this photo:
And I figured this one was real because Trump posted it from his Truthsocial account.
All of which makes me think: are sources more important than facts?All of which makes me think: are sources more important than facts?
Before going there though, let me tell you how I spent my afternoon. In their excellent post on what it's like in Venezuela right now, @bief57 noted that the most popular image of Maduro circulating this morning was actually AI generated (#1402478).
This gave me the idea of finding out where that fake image came from. I did want to get a look at the person who made the image and posted it, but mostly I wanted to see why it changed in so many ways. It's not like one person just made an image and everyone else reposted it. It was everywhere: on a news report, on a laptop, fuzzy, with dark lighting, or blown out, with a date and without a date, even with a weird green dot. Some of what gave it credibility -- the feeling of authenticity -- was that it seemed to be showing up in a lot of contexts.
Where the image came from (skip this part if you just want to get to the troubling question at the end)Where the image came from (skip this part if you just want to get to the troubling question at the end)
Of course, by the time I looked in to this at 1400 UTC, someone had already done a deep dive on where the image came from. Sadly, I didn't discover this until I had spent two hours trying to track down the exact time it emerged on X.
I started my search with this image posted by Yan_infor at 11:02 UTC:
What I found interesting about it was that it seemed to be a picture of the image on a laptop. There are some meta buttons at the bottom and the lighting is very different.
Working back in time through X archives, it wasn't hard to establish that image may have come from this one posted by Alex_RobertsJ at 10:43 UTC:
That looks a lot like this image from Categrico2 posted at 10:42 UTC:
Except in this shot you can tell there is a microphone in the foreground at bottom left and an person in the foreground of the bottom right, but the writing across the bottom of the screen is in front of the person. So I figured that all the images with the "official news" banners at the bottom must have come from this.
I was able to find the YouTuber. His channel is Bowery Newsroom - Guarimba Digital, and he was doing a live show about the events in Venezuela and started talking about this image at roughly 10:17 UTC. He says, "No es inteligencia artificial, es una foto de verdad. Me lo acaban de mandar de Washington, señores." ...which is a bold claim.
If you look back in the X archives right around 10:17 UTC, there are bunch of screenshots of Bowery Newsroom's livefeed, showing the fake Maduro capture picture with newsy headlines across the bottom. Apparently, it made quite a splash.
But it still didn't explain where Bowery Newsroom got it from, nor why there were so many pictures of a laptop screen showing the image.
Going further back, you find that these "laptop pictures" seem to come from this guy named Anton Gerashchenko who posted a picture of a laptop (his?) displaying the Maduro image at 09:53 UTC.
Looking earlier in the X archive, the image really doesn't show up, although there are a number of other AI contenders such as these:
And I probably never would have found the original, if Tal Hagin hadn't pointed out where the image seems to have come from:
At 9:41 UTC a user named Ian Weber put up this post:
If you click on the image, you'll see this:
I wasn't able to find any version of the image earlier than this.
The moral of this little adventure is that if you want people to be fooled by your AI images, it helps if you take a picture of them on some screen and then post that.
So, back to sources (and the troubling question):So, back to sources (and the troubling question):
Clearly, we are in a world where you can't just trust the images you see. Especially when you don't know how many people they've been filtered through. Probably, the easy answer is to just upload every image to some LLM agent and ask it to tell you if it was created by a bot -- or even better build this natively into apps (something I think is already happening).
But how long do we think it will be before this is not a reliable way to detect AI generated images?
Or even worse: who has tested to see how many false positives this technique will create?
The only picture I really trusted was the one posted by Donald Trump on his Truthsocial account because I figure it came from Trump.
Of course, social media accounts can be hacked.
And what if Trump wanted to trick us? In this case it doesn't seem like it matters much, but what about a scenario where it does matter?
It feels to me like we are being trained to rely on the heuristic of source over evidence -- that it is more important where the info comes from than what it says. And I wonder when this heuristic will get taken advantage of.
Is there a way around this dilemma
Hmm.. this used to be the the only heuristic available to the general population before the internet. And it was taken advantage of since forever. We've had some good years from the mid 90s to the mid 2010s or so when there was no answer to the loss of obscurity because CGI was expensive af, but the answer was being formed through massive false flag operations, which can now be done by anyone with a phone and an hour or so.
This is why many of us that were very active on the internet in the 90s are dismissing a lot as
false-flag, for everything. Because we've seen it happen; but this makes it hard to not be perceived like a complete loon. What I tell myself to retain sanity is that there is no truth and never has been: everything is subjective. That way, I can take a step back and not be upset all the time about things I cannot change, and focus on those that maybe I can. Even if you're working on a high risk, high reward moonshot, you can only do one of those at a time.Coming back to your image. I had 3 MSM live reporting streams on a separate laptop yesterday and the one posted on Truth Social was the only image I saw. I first thought it may be fake and I still think it may be edited. Looking at the picture quality, it also looks like a photo taken of a photo (or a deepfake). I found it interesting that someone else thought that too - an anon asked my question before I could, when bief57 said there was a fake circulating.
The important thing that changed is we now have enormous choice of sources. If dishonest actors lose out on our attention to honest actors, there's at least a mechanism in the system that rewards honesty.
I don't think the conclusion I draw from the exercise of tracing this image is that "there's at least a mechanism in the system that rewards honesty."
The fellow who first traced the fake image posted his very thorough thread within an hour or two of the fake being posted, yet I was far more likely to encounter the fake image than I was to encounter his debunking of it.
To optimism's point of not seeing the fake image on MSM: score one point for the establishment.
There are also mechanisms to proliferate dishonesty.
I didn't say the mechanism incentivizing honesty was dominant, just that it exists. I don't think it existed before.
I think there have always been long term benefits to integrity. The reason why a significant portion of the MSM stopped caring is because these may not align with short and mid term profits.
I think the long-term benefit was staying on the right side of the feds, whose permission was required to continue broadcasting.
The outrage they'd be on the receiving end of over not covering their asses on stuff like this is enough reason for them to be more careful than, say, the usual FUDsters on the bird app.
There is still some journalistic integrity in the world, but we also know that there's a narrative going on and they have no choice but to report those narratives. I found it interesting that for example Al Jazeera fell into the "legality" discussion trap yesterday, much like the BBC, until an analyst remarked that this is a dumb distraction move. It's far more interesting to find out who the insiders were, and what's in it for them.
While this is true, it's also the case that the usual FUDsters on the bird app are not as likely to be the target of a serious pressure campaign.
Back when media was a few thousand journalists, people who wanted to control the story had a much easier time identifying which throats to choke. These days, control is some sort of aggregate business, as we were discussing the other day, it's become more of a sybil attack.
I'm also not convinced that outrage affects anyone in the media business (MSM or small time FUDster) very much. They make money from the outrage. Worst comes to worst, they give it a day or two and move on...because everyone else certainly has.
And of the many meanings of journalistic integrity that we might come up with, I hope that it is closer to "finding out who the insiders were" than "getting the details all exactly right."
Because you can just give em some cash and they'll sing your song; broken 60s jukeboxes have more integrity because at least they won't play what they don't have. So you're right, because you don't have to pressure if you can buy. At a bigger scale, this was also how I interpreted that whole "buying the NYT" recently.
It's the one where you make people believe that they want to be controlled by you. Because your censorship, lies and self enrichment are the best thing that ever happened to them, of course. Oh and let's not forget the new wing of that college that was built with your name on it because you earmarked 50M of creditor rektness for it in a 7000 page spending bill - that is the most awesome thing anyone has ever done.
Many do, but I think for some, rectification hurts. Yes, there are still some newspapers left in the world that actually do this. Total self-own of course, bad for profits. They could pump out 300 stories per day but these losers only do a 100. Sad! Maybe we should invade them, kidnap their CEOs and control them too. So much money lost! Think about the money!
Why?
Agree, but I'm not sure that this is the case.
This is definitely true. Emphasis on never has been. Yet, it is a difficult way to live, because it is so easy to fall into the pattern of trusting a source or trusting what you see. It's also difficult to build with such a world view: we can't spend all our time second-guessing our priors. And of course the flat-earthers are just stupid.
We are all probably used to the idea that words can lie, but even before the days of a camera in every pocket, I think we had a sense that it was harder to lie with a photograph (and even harder with video). In the 80s and 90s (when cameras really became widespread), an image could trump the source...I think -- not that it always did, but it could.
Then there was a brief shining moment, maybe 2010 - 2020? where digital cameras were suddenly in everyone's hands, while the means of believably faking images was not -- it felt like all of a sudden you could get as it was happening photos and videos of events around the world. There were still many ways to distort the truth or lie, but it felt like you could learn something from an image itself, regardless who posted it. This is what I think has broken.
In that decade, I think we all got somewhat used to seeing an image and believing it or being able to tell somewhat quickly that it was photoshopped. If we couldn't tell that it was faked, this was due to the changes being relatively subtle.
Trump's photo of Maduro easily could have been any slightly graying latin man -- the face is mostly covered. Using an actor was always a viable way to fake the image, yet the sheer prevalence of digital cameras made me at least think that fakes were harding to achieve...someone would leak a photo of the actor getting ready, chatting with a soldier or taking a smoke break (remember the fad of covid actor images where the morgues were full of bodies but then other images showed us the same bodies getting in a last puff before getting into character?).
These slight hopes at accessing truth are gone now. Perhaps you're right and we are only back in the same square as we were in the 90s, but it feels like we lost a little edge we had gained in the battle over control of information.
The unsettling bit is we’re training ourselves to ask “who posted this?” instead of “what can be verified?” That tradeoff might bite hardest during real crises.
That's a good question. How would you answer that question for this situation?
Good question In this case the answer is to look for real world follow through, not just the image or the source If something that big were true, you’d see independent confirmations and second order effects pretty fast when the world doesn’t move that’s usually the strongest signal.
Are you saying that the narrative of deep fakes was all an elaborate ruse to rouse us into trusting sources, as opposed to the evaluation and forensic study of primary evidence? Maybe that wasn't what you said, it could have just been the voice of the general from Ghost in the Shell, hacking my ghost as I read ;-)
I was thinking about the authenticity of media accounts while reading some the other day. I suppose another valid concern in addition to hacks would just be the general nature of assistants human or otherwise. I mean, am I to believe that the social account of politicians and leaders are the actual person posting these notes themselves in real time?
No elaborate ruse, just the forces of human civilization sloshing us about.
The rational parts of my mind say obviously not...and yet I so badly want to believe that is Elon or Trump who posts from their respective accounts. I'd guess more than half the world holds similar feelings passionately enough that they're willing to stay in never never land and just believe.
Fair and logical assessment.
I didn't mean to suggest that all important public facing social accounts are not their own, just that there must be a large swathe that are too ingenuine to actually put themselves on the line. Some actually do come across as unfiltered and genuine.
I'd imagine national government and opposition routinely run there posts through a whole team of PR and legal scrutiny before they post more than 'GM' or a picture of their latest campaign photo-op.
Honestly, I think it is fair to call every big account into question. There is so much money to be made in the world of social media, and when we are talking about people like Elon or Trump or Taylor Swift it seems laughable that they might be doing their own posts (I mean, how do such people even have the time?), yet they are trying I think to make us believe they are. How else to explain Trump's unique style of tweeting? or the wild things that Elon says?
I'm going to try to answer your question!
For Elon, public owner of x, richest man, and generally smart dude who exhumes righteousness and serves as a reflection of the indignation of the righteous suffering of the downtrodden collective censored west, it serves the platform well to maintain a human presence that embodies its values. Particularly as Jack Dorsey gave birth to the platform and steered it well, before it became what it became, before now.
As for The Don, my memory stretches back to the dreamstate typo kofefe mystery which captured the imagination, and if I was to build a persona, it would be that of a non-digital-native age guy, a quirky but astute tychoon.
They definetly both appear to post the least orchestrated bulletins. There have been acccounts taken of control of, maybe. So, hard to say how else to explain! I'd say there's a need to foster belief in the dictat coming from the horses mouths. In the end, I'm in the same ballpark, where I choose to believe and I give it a pass. We both can assume policy originates from where it originates, and that's not the mouth of one or two horses.
If such high-level leaders, political or otherwise, can enter a live debate, talk coherently, humanly, and convince us they have more than an uncle Jim level of undersanding of the world, I think it lends more credence than a polit-bureau speaking for individuals (or a nation) at minimum, as a reflection of a free society with checks and balances on the direction of a nation.
And when you actually think this is bad it might just get way worse ai is really getting out of hand
this is the point of web of trust. let accounts and sites post fake shit and watch their WoT score go down. and it's all from your perspective, so if you don't trust cnn then you wouldn't include them in your WoT and their content would be ranked appropriately (for you). i recommend listening to interviews with pip like this one https://trustrevolution.co/episodes/s02e16-pippellia/
WoT is awesome, but I'm not sure that it will help us with discovering the truth. Seems more likely that it will help us find the truth our tribe agrees on. Maybe that's just fine, as you say "it's all from your perspective" -- but at the end of the day, I wonder if it allows us to have a better grasp of what is actually happening.
Could be wrong but I thought there might be some software that can tell if an image was AI generated. Could also be easy to route around.
Normally when a news story breaks like this. You read it and are like... What!?
I like to take a step back, let the dust settle, then dig in as much as I'm interested.
There definitely is software than can tell. Just upload any image to Gemini or ChatGPT or Claude or whatever is your favorite.
I have not yet seen any studies about the prevalence of false positives. Also, I think it is very likely that image generation tooling will quickly beat detection. It's not even going to be a contest.
It seems that there are no trustworthy sources these days, only those that are less untrustworthy
The steps you took to determine the origin of the images are what we need to teach our children, though I doubt most netizens (me included) have the inclination to sustain their detective work for two hours
If I'm honest, I spent much more than two hours (writing the post probably took an hour itself). Unclear if it was worth the effort, but sometimes I get a thing in my mind and I have a very hard time moving on.
it's a wild time to be alive, hard ot beleive anything, but some of these pics are hillarious. i still remember the ones of trump being arrested that pepole thought were real
Can you imagine doing historical research on this stuff? It's already hard enough to figure things out 10 or 12 hours later. Trying to look back 10 or 20 years is going to be a mess.
So which is real?
I dunno, but I guess I believe the Trump one is.
Yes but isn't this basically because we just trust he wouldn't make one up?
In this case, that's probably true, but we're clearly fucked if we don't find a better way to verify long-term.
Well we know he reposts AI stuff all the time but this isn’t a repost
How awesome would it be if Maduro had actually evaded capture and Trump just got Gemini to make a believable photo of his capture?
Lololol
But no that’d be not awesome
Towards the end of the last election, I told people that it would be the last election that you could belive anything you saw or heard the candidates do. There will be deep fakes, and real things that are said to be fake. It's going to be chaos.
Yeah. We are already way past the point where you could trust a picture.
The first one seems the most realistic
The first one looks like a Darthcoin meme lol
I like this one (which actually was posted before the popular one got posted):
Oh yeah. I can see why
Also [this one]https://x.com/winus_ai/status/2007386728718586367) was pretty good -- it's so interesting to me why it didn't catch on like the other:
https://twiiit.com/winus_ai/status/2007386728718586367
https://twiiit.com/pt0j/status/2007386798427901964
This one had me fooled
https://twiiit.com/talhagin/status/2007417679712559604
When an image appears across different mediums and formats our brains lean toward thinking it must have multiple independent sources. That repetition is a psychological trick whether intentional or not.
The reliance on source over evidence is a natural shortcut in a world overflowing with information. It feels efficient to simply decide that if the individual or platform is familiar and historically credible the content must be accurate. But this shortcut is becoming increasingly fragile because both the source identity and the content can now be fabricated with minimal skill. A deepfake video or AI image gains more legitimacy if placed in the hands of a trusted figure and that kind of placement is easier than ever in the age of account hacks coordinated misinformation campaigns and cleverly staged recontextualization.
In the end the greatest risk is that we are being conditioned to prioritize provenance over proof. If this conditioning becomes widespread the barriers to high impact deception will collapse in moments of political or economic crisis. The discipline to demand factual verification even from the most trusted figures may be the single most important information habit to protect in the years ahead.
Look what they did while I was writing this!
view on x.comhttps://twiiit.com/PaulDMauro/status/2007624542156660903