STOP AI FROM REWRITING HISTORY
And Other Evils
This video is nearly 14 minutes long. Below the transcript in this post is COMMENTS, in case you want to just skim or skip most of the transcript. The comments are great too, including mine at the end.
VIDEO TRANSCRIPT
With all the crazy technology we deal with today on a daily basis, I often find myself wondering what these brave soldiers from the past would think about our modern way of life. Oh, wait a second—those aren't real soldiers. All those pictures are AI-generated.
Now, you may not have heard about this because nobody's talking about AI right now. In fact, I'm probably the first one to make a video here on YouTube about AI. But generative AI has really been taking off lately—it's getting faster, cheaper, and scarily good at creating realistic photos and videos. Calling out fake historical media has always been a staple of my page. In the past, I've covered pictures from historical reenactments, war movies, and even pictures of action figures that have been passed off as real historical photos. But now AI is starting to blur the lines of reality in ways that we've never seen before.
Today, I want to cover some of the ways that I've encountered AI being used to alter the past, why this is dangerous, and—believe it or not—I'll even cover some uses for generative AI relating to history that I think can be good. …
Now, unfortunately, to understand this AI nonsense, I think we need to start by talking about the AI cesspool that is Facebook. Facebook has become ground zero for all types of AI-generated slop, but I especially see a lot of content on there related to history and the military. Let's get into how Facebook created the perfect storm for this AI garbage to exist, so we can understand why these fake historical posts are being shared in the first place. …
The first reason why Facebook is perfectly suited for these AI posts is that, unlike most other platforms, you can monetize photo posts on Facebook. It's much easier to set up AI to generate and post these photos compared to other platforms like TikTok and YouTube, where generally you have to post a video in order to monetize your content.
Second, Facebook's older and international user base is often more likely to fall for these AI fakes. The comment sections of these posts are generally filled with supportive messages from elderly people, people who aren't viewing content in their native language, and bots. Keep in mind, if their page is monetized, they're making money off of all those comments.
Third, Facebook's content enforcement is very lax. They seem to have no interest in labeling or removing AI content, which most other platforms have taken steps to do.
All right, now let's take a look at some examples of fake historical photos that I encountered while roaming the unforgiving trenches of Facebook. I've grouped these into categories so we can take a look at a few different post formulas that these AI history pages typically use.
This first affront to the eyes is an AI-generated copy of a real photo from World War II. The caption on these types of posts is typically the original caption from a real World War II photo. These are super common, and this type of post is just very perplexing to me. In my mind, it seems that it would have been much easier to just post the actual World War II picture. Most of these photos are in the public domain, so it's not like there are any copyright issues—not that these pages would probably care about that anyway.
The original photo probably would have generated more engagement and wouldn't have required all that extra effort. But instead, they take an original photo and caption, use AI to generate a weird version of that photo, and then upload it with the caption from the original photo. I would guess that the reason why this doesn't make a lot of sense is because that decision was probably made by the AI bot that's programmed to upload stuff on these pages. I really don't think there are people out there prompting AI to generate weird copies of these historical photos so they can post them as if they were the real thing—or at least I hope not.
The next category is what I call patriotism farming. These are posts meant to farm engagement by kind of guilting people into sharing, liking, or commenting on the post—because if they don't, they might feel like they don't care about their veterans, their country, or whatever. Some of these posts share real stories of veterans accompanied by weird AI-generated photos, while others just post something vaguely war-related that's clearly fake.
These are very similar to the posts you've probably seen where it's like an AI-generated old person with a caption saying that you need to wish them a happy birthday or thank them for their service or something like that. These are very popular, and if you ever go on Facebook, you've probably seen people unironically sharing these posts.
This last category is probably the one that's at the highest risk of causing real problems. These are just fake historical photos, often with no basis in reality whatsoever. As usual, most look pretty bad, but some look very believable at a glance. To make things worse, these pages often post a mix of real and AI-generated photos—sometimes in the same post—so it just blurs the lines even further.
Like I said, Facebook is the epicenter for this AI epidemic, but these types of fake historical photos pop up in all types of other places too. Every other social platform has their own version of this problem, and I've also seen this a lot lately from stock photo websites, which often have their own embedded AI tools. When you Google image search something history-related, there are often suggested results from these stock photo websites that are fake AI-generated photos, which some people might mistake for the real thing.
So what's the impact of this, and who cares if someone wants to make silly vintage AI photos? Well, over time, these photos really can blur the line between historical fact and fiction. I mean, studying the details of historical photos is one of the best tools we have to learn things about the past. I look at World War II images all the time to figure out what weapons and resources troops had available for specific battles, which troops were present for specific events, and much, much more. Inserting fake images into the mix can seriously alter how we understand these important historical events.
I know we all love to think we're immune to falling for these AI posts, and a lot of them are very obvious. But AI just keeps getting better and better, and it also keeps being used in new ways where we don't even think to look out for it. Sure, there are certain groups of people who are more susceptible to being duped by AI, but it's happened to me, it's happened to you—it happens to everyone at some point, whether or not you've realized it.
There's a very real risk of these AI photos getting mixed in with the real ones. Also, in my experience looking into fake historical photos, the original post usually is not the big problem. Let's look at a non-AI-generated example to understand how this cycle often works. This is a photo that was taken at a World War II reenactment photo shoot years ago, along with a few similar photos. At some point, it was uploaded online—probably by the photographer who took it. But then this photo got re-shared without the context that it was from a reenactment. As soon as that happens once, all bets are off. Tons of other pages then grabbed this photo and started sharing it, often with their own captions added to it.
Some even included the name of a real 101st Airborne soldier from World War II and claimed that the photo was of him. After a while, sellers on eBay who sell prints of photos listed the photos for sale, claiming that they were real photos from World War II. Now, this photo is one of the top search results when you look for pictures of the 101st Airborne during World War II, and the average person would have no idea that this was taken at a reenactment.
Even if you tried to research this photo, you'd find countless posts from people over the years claiming that it's real. I would love to say this is uncommon, but there are numerous examples of this happening with photos online. Over time, these pictures essentially become indistinguishable from the real ones. These AI history pages are relatively new, so I don't think many of their photos have had the time to go through that cycle yet, but believe me, they will.
Even some of those AI photos that aren't super convincing now—when they get shared over and over again, and the quality is reduced, and people add filters and stuff like that—it can make it harder to pick out those identifiable AI features in the photo. In fact, covering up those distinguishable AI features is one of the main reasons why AI historical photos are often so hard to identify in the first place.
I remember at some point hearing someone talk about how some of the most convincing AI photos and videos are those that are made to look like they were taken by security cameras. We're all used to seeing low-quality security camera videos, and that lower resolution can hide a lot of the telltale signs of AI. I think this is even more applicable with historical photos because old film photos are often grainy and sometimes damaged, and that can hide a lot of things. When you consider that some of the most famous real photos of D-Day look like this, it's easy to see how a blurry AI photo could be mistaken for a real one and distort our view of what happened in the past.
Take a look at this slightly grainy photo, for example, and see how many of the typical AI identifiers you can spot before I point them out. All right, did you find anything? Well, I hope not—because that's a real World War II photo. I would guess that a lot of you were starting to pick apart and question some of the details in that photo, which is completely understandable. It just goes to show how AI is distorting historical reality—the real stuff starts to look fake, and the fake stuff is looking more and more real.
Lately, I've been seeing a lot of comments on posts of historical photos accusing them of being fake when they're very real. These commenters are probably skeptical because maybe they've been fooled by fake historical photos before. It's good to be a bit skeptical, but this can also discredit real historical photos, which is just as bad as introducing fake ones. We risk getting to a point in the future where people can just write off any historical photo they disagree with by claiming it might be AI.
We've already seen the trend lately of politicians and public figures casting doubt on anything that makes them look bad by implying that it might be AI-generated. Now, let's move on to talking a bit about what we can do about all this to at least slow down the growth of this fake AI history.
First, do not engage with AI historical content. Doing that makes money for the people who posted it and for the platform hosting it—it just perpetuates this weird AI ecosystem. I think the platforms should create better guidelines on AI usage, but they aren't going to do that if they keep making a bunch of money off people consuming the stuff.
Next, I'd say view any historical media with a hint of skepticism and consume historical content from reputable history pages. There are a ton of great ones out there to choose from. If you find a page that's hosted by a real person, or that's affiliated with a reputable institution like a museum, or that just posts information that you can verify with other sources, those are some good signs that the page is at least more reputable than the faceless profiles you come across.
And finally, talk to the people in your lives who fall into the demographics that are generally more likely to be tricked by AI—children, the elderly, sometimes people consuming content that isn't in their native language. These are often the most likely people to be duped by AI.
Okay, we've talked about how this technology is dangerous and how it's scary and it's going to take over the world and rule over all of us someday, blah blah blah. But I'd like to end on something a bit more optimistic. Are there any good uses for AI when it comes to historical media?
Well, in my opinion, absolutely. My favorite application for AI in this space has been for restoring and upscaling historical photos and videos. If you guys haven't seen this, it's pretty incredible, and I highly recommend checking out some YouTube videos of this when you have the chance. Basically, there are programs out there that use AI to enhance the quality and frame rate of historical footage. Some turn out way better than others, but I've seen some really good ones out there that make 100-year-old videos look like they were recorded with modern technology.
I do still think we should be careful with this stuff because you are generating new pixels and frames when you use these tools, but generally it doesn't affect the content of the footage. I've heard of similar technologies being used to restore historical documents, and again, as long as it isn't changing the content of the document, I think that's great.
Another popular use I've seen lately is using AI to animate historical photos into moving clips, and I'm still not quite sure how I feel about that one. Honestly, I find some of these a bit weird and creepy, but there have been some very popular uploads of these online, so clearly some people really like them. If these animations are helping some people connect with the past and better relate to the people in these photos, I think that's a great thing.
In my opinion, one of the hardest parts about teaching history is getting others to view these historical figures as regular people, like the ones we interact with every day. I've seen a lot of posts where people say they imagine anything before a certain date happening in black and white. Even though I know those people are mostly kidding, it really can be easy to feel disconnected from the past when we're only seeing it in grainy black-and-white footage. It can be hard to imagine what it would have been like to be there for the Hindenburg disaster, for example, when we've only seen it in spotty old footage.
If AI can help people relate to the past by turning this grainy old footage into something like this while preserving the original content, I think that's great. After all, making the past more relatable to people today is one of the main reasons I make these videos.
Now, what do you guys think? I've heard so many mixed opinions on AI as it relates to history. Do you think this won't really change much because the technology to alter historical photos has already been around for decades? Do you think AI will permanently blur the line between what happened and what didn't?
Please let me know your thoughts in the comments—I'm really interested to see some more opinions on this. Also, I want to give a huge thank you to my hot-shot-tier supporters on Patreon: Corey Mhler, Emily Hensley, NCR Trooper Scout, and World War Historian22. Thanks for having my back, guys—I really appreciate it.
With that said, thanks for watching, remember to thank a brave AI veteran on Facebook, and I'll see you in the next one.
COMMENTS
{As usual, the video viewers have good comments & suggestions, some of which I’m copying below.}
@warrenharrison9490 — 3 days ago AI should have mandatory water marking.
@hanspotterman — 3 days ago As an artist I’m already PISSED about these so called "ai artists" but now it’s getting into my other interests and I’m so done with ai at this point
@LewisTaylor-pc1zw — 3 days ago I googled World War II a month ago and the first three recommended images were AI and that was the most sad thing I think I’ve ever seen in my life.
@bubba98 — 3 days ago Y’all remember when things being photoshopped was the biggest worry when it came to images?
@Novice_Rifleman — 3 days ago I greatly despise AI. I tend to get a lot of fake images and videos regarding several Napoleonic topics recommended to me. AI voiced videos, fake images, HORRIBLY inaccurate info, etc.
@littlehousebrower7752 — 2 days ago Honestly, well said. Coming from an artist who wants to get into animation, this subject hurts. face-purple-crying I said it once and I will say it again, AI should not be there to replace humanity, it should be there to assist humanity.
@rizluz396 — 3 days ago In Norway any advertisement that has "retouched" the models have to write that in a disclaimer. A new (universal) law should come in place making it mandatory to write in the caption that it is ai generated etc.
@Azeria — 3 days ago We’re going to need public institutions to archive and then make accessible archives of photographs more than ever before. It was just a brief window in time where we could more often than not rely on a google image search and it’s over now. This is the 5th video like this I’ve seen a creator make in the last week, too. It’s just sad how pervasive it’s got so fast, that people either can’t see the harm in it or don’t want to. I’m tired man.
@WarbirdComics — 1 day ago As a historical artist, I've never been happier to be well versed in verifying information, and not rely solely on digital sources... My collection of books for reference is already somewhat formidable, but count on it to grow in the light of this. As will my use of digital archives and museums, and thorough cross-checking of any information I come across.
@peyton-rproductions380 — 3 days ago 4:27 "Today is my birthday because I'm poor" Such inspiring quote from a veteran {apparently that was shown in this video as an example of fakery, as on Facebook}.
@jon-paulfilkins7820 — 3 days ago One of the biggest fibs in historical studies is that the Camera never lies... from the moment the photographer decides to photograph this direction rather than that, someone has been filtering the image, it gets treated during development, cropped, presented with captions that may change its context. Photographers even used to compose 'shots' in the field, moving items into or out of shot, some in the development room created composite photos effectively stitching different photos together, then you have the infamous use of proto photoshop, airbrushing! Photos can be useful, but never trust them at face value.
@mitwhitgaming7722 — 3 days ago Something similar happened with photos from space. People are so used to sci-fi and space shows where you can see the stars that they accuse real space photos where you can't see stars as being fake. Not realizing that's how cameras and exposure lengths work.
@TheLeviathan000 — 2 days ago have you seen those weird ai history channels too? They all usually have “Painter” and “Explained” in the channel name and are all extremely inaccurate.
@DraganKKWCZ — 2 days ago IMHO ai copies or imitations of historical pictures should be banned by the law
@d3nza482 — 3 days ago ... It's literally the opening of Chernobyl:
"What is the cost of lies? It's not that we'll mistake them for the truth. The real danger is that if we hear enough lies, then we no longer recognize the truth at all. What can we do then? What else is left but to abandon even the hope of truth and content ourselves instead with stories? In these stories, it doesn't matter who the heroes are. All we want to know is: "Who is to blame?" ..."
@Jorgerome9557
3 days ago
What also worrying is that people didn't see this coming. Foresight is what history serves as the main course. Yet, I couldn't find anyone agreeing with me even before this mess.
@legogonegaming
3 days ago (edited)
10:09 “I’ll stop kicking this dead horse when it stops shitting out money”-Eoin Reardon
@jehru5
3 days ago
Forgotten Weapons recently released a video addressing AI generated images as well. He keeps getting sent "historical" images and being asked to identify the guns in them.
AI is such an amazing tool. My work has been implementing AI to help us search complex technical documentation and compile procedures across multiple documents. It's incredibly helpful. It sucks that people are using AI to deceive others.