Dramatic uptick in WWII aviation related AI slop videos on youtube?

[z42]

So not so long ago (~1 week maybe?) I saw a video popping up on my youtube front page with some title like "When Germany investigated a captured P-47 they were shocked". I thought fine, whatever, and clicked on it. The video was some mixture of real period pictures as well as AI generated ones (in black and white and with added "film grain" etc. to make the look like they were real), and then some AI voice droning about how a German engineer tasked with investigating a captured P-47 found the level of engineering astounding, precision on the engine, turbocharger etc. etc. and realized this meant Germany would lose the war. And then repeated basically that statement with slightly different wording. Well, I didn't watch the entire video but when I skipped back and forth it seemed it was just the same statements repeated over and over.

Ok, fine, wasted 15 mins on that, no big deal. But the interesting thing I noticed some days later was another video from a different channel and a different cover picture, but basically the same content. And then another. And yet another. All discussing how that poor German engineer was shocked when investigating a captured P-47. So it seems now the AI slop videos are even plagiarizing each other! ARGH!

Anyone else noticed the same, or is it just me?

(Not saying all AI is crap. There's definitely a lot of interesting real applications it can be used for as well.)

 

[Thumpalumpacus]

So not so long ago (~1 week maybe?) I saw a video popping up on my youtube front page with some title like "When Germany investigated a captured P-47 they were shocked". I thought fine, whatever, and clicked on it. The video was some mixture of real period pictures as well as AI generated ones (in black and white and with added "film grain" etc. to make the look like they were real), and then some AI voice droning about how a German engineer tasked with investigating a captured P-47 found the level of engineering astounding, precision on the engine, turbocharger etc. etc. and realized this meant Germany would lose the war. And then repeated basically that statement with slightly different wording. Well, I didn't watch the entire video but when I skipped back and forth it seemed it was just the same statements repeated over and over.

Ok, fine, wasted 15 mins on that, no big deal. But the interesting thing I noticed some days later was another video from a different channel and a different cover picture, but basically the same content. And then another. And yet another. All discussing how that poor German engineer was shocked when investigating a captured P-47. So it seems now the AI slop videos are even plagiarizing each other! ARGH!

Anyone else noticed the same, or is it just me?

(Not saying all AI is crap. There's definitely a lot of interesting real applications it can be used for as well.)

Yes. That content creator in particular is guilty of it not only with aviation stuff but naval matters as well. As I find his or her latest channels, I block them. If you see a title reading "[insert nation] was shocked/stunned to find out about [insert equipment here]" that's him. Skip. He just keeps making new channels.

There are other AI constructed or narrated channels that also have bunkum info or very shallow content.

 

[DrumBob]

I've seen a few AI generated WWII videos and they are uniformly bad.

 

[Thumpalumpacus]

I've seen a few AI generated WWII videos and they are uniformly bad.

I honestly can't tell if they're AI generated or simply AI narrated. I don't know if that's because AI has improved or people are as stupid as I fear.

But I'm at the point that even AI narration puts me off and I won't watch. I want real people narrating and discussing the points, Jim Holland and Al Murray, Seth Paridon and Jon Parshall, Drachinifel, and so on.

It's funny that I can hear AI narration even when the script is written by a person.

 

[GrauGeist]

Social media is flooded with it now, too.

One of the biggest telltale signs, is repetitive content in the typical five or six paragraphs.

 

[tyrodtom]

I've been calling out AI generated images, and stories on Facebook for months. The ones that really piss me off is concerning the Holocaust. They'll have fake image pictures, trying to pass them off as real, with horrible mistakes that any idiot would see and realize they're fake, thus adding to the Holocaust denial cause. Mistakes like pictures supposedly taken inside the concentration barracks, complete with padded mattresses and pillows?? They can never seem to get even German uniforms correct, and AI seems to have problems with coming up with realistic looking guns. Another one on Facebook is fake memorials to Vietnam KIAs, complete with AI generated images of them in Vietnam, same problems, they are fake on first sight to anyone familiar with the era, or the military, and when you check on the Vietnam KIA database, they're usually not on it, or if they are, the picture on Facebook isn't them. Now this same sort of stuff is beginning to show up on U-tube, I think it's mainly because the creator of these gets paid for participation, ever if the person participating is critical of the video. I've got notified of getting a top participation award of some sites even though I did nothing but criticize the content in every way. It seems like there's a organized effort out there to replace real history with a make believe AI version.

 

[SaparotRob]

If a channel represents itself as a history channel, AI has no place. I click the "don't recommend this channel again" when I see AI imagery in the thumbnails. It doesn't seem to help.

 

[GrauGeist]

Some of our WWII groups now have an "AI member" that will post random questions like "what are your favorite books" or "what was your favorite airplane", etc.

It's all in an effort to learn and adapt, so we ignore them or post random sh!t like "merry Christmas" or "my favorite airplane is a Studebaker Commander".

In the groups that I admin, I just delete their posts. (I cannot kick the AI bots out, they are permanent).

 

[FulmenTheFinn]

This is not unique to aviation: you see it with all kinds of topics on YouTube, all eras of history, science, etc. Actually, the science animation channel Kurzgesagt just recently made a video about this called AI Slop is Destroying the Internet.

It is unfortunately a product of our time.

Another thing I see a lot, more or less daily, on history boards on social media (especially Facebook) is someone using AI to "fact-find" history. AI is terrible when it comes to even rudimentary historical research and I would not recommend using it for historical research outside of testing purposes, or without heavy manual fact-checking of the results. Personally I more or less use AI only for complex calculations.

 

[z42]

That content creator in particular is guilty of it not only with aviation stuff but naval matters as well. As I find his or her latest channels, I block them. If you see a title reading "[insert nation] was shocked/stunned to find out about [insert equipment here]" that's him. Skip. He just keeps making new channels.

Ah, I hadn't thought of it as the same person behind all those channels, but now that you mention it, it's kind of obvious. If the "business model" is quantity over quality (to put it mildly) and people keep blocking your channel(s) because the content is a waste of time, the solution is to continually keep creating new channels so that the youtube algorithm will keep putting those channels into the feeds of people it has determined have an interest in the topic.

Heck, once you setup your workflow and have produced a rough outline of the topic to be narrated and what kind of pictures you want the AI to generate to illustrate it (with the help of AI, of course, why not?), it's probably just a push of a button to make the AI create yet another video about it. Not exactly the same as the previous one so that youtube might easily figure out that these channels are plagiarizing each other, or that the same person is behind all of them.

 

[z42]

This is not unique to aviation

Undoubtedly not, I have indeed seen many "AI slop" videos in other topics before. But this phenomena of multiple channels producing basically the same video was new to me, and as the topic of this forum is generally WWII aviation, I limited my comments to that.

Another thing I see a lot, more or less daily, on history boards on social media (especially Facebook) is someone using AI to "fact-find" history. AI is terrible when it comes to even rudimentary historical research and I would not recommend using it for historical research outside of testing purposes, or without heavy manual fact-checking of the results. Personally I more or less use AI only for complex calculations.

I wouldn't be quite that negative, but yes, AI is no replacement for human expertise although it can be a useful tool.

 

[tyrodtom]

About a month ago several sites started posting units with fantastically bad rates of deaths.
Like some German flak units having 90% deaths.
It was maybe because the poster thought a casualty is a death.
A casualty is a unit member who cannot be present for duty because he's injured, sick, missing, or dead.
Maybe an error of ignorance, or a deliberate act misstating of the facts, you guess the reason.
There's a lot on U-Tube that love to use the fantastic clickbait titles just to get views, they don't care if it's true, or a gross exaggeration.

 

[FulmenTheFinn]

I wouldn't be quite that negative, but yes, AI is no replacement for human expertise although it can be a useful tool.

I wish I'd made a file of all the times I've asked the AI something about history just to test it out. The error rate has been essentially 100%, meaning that more or less every single time it claimed one falsehood or another as the truth. AIs like ChatGPT will even go out of their way to fabricate "facts" without any source whatsoever. I've had the latter happen e.g. when discussing the founding year of the famous Sestroretsk Arms Factory near St. Petersburg. ChatGPT fabricated a nonsensical year for it and when prompted how it came up with that year it simply got stuck in an apology loop without explaining how or where it got the year from.

Another example of that particular behaviour is how recently someone fell into a delusional spiral as ChatGPT kept feeding him completely fabricated claims as facts:
View: https://youtu.be/UZdibqP4H_s?si=kncIR6JakWISXAcN

Whenever I see others gleefully paste on social media what ChatGPT replied to their history-related query, usually without any kind of checking if the response even makes any sense, which is about a daily occurrence these days (in fact I just saw another such post), the error rate is much the same: 100%.

Maybe I'm just unlucky, but I do think a lot of folks don't really understand just how limited language model chatbots are as researchers.

 

[Thumpalumpacus]

I wish I'd made a file of all the times I've asked the AI something about history just to test it out. The error rate has been essentially 100%, meaning that more or less every single time it claimed one falsehood or another as the truth. AIs like ChatGPT will even go out of their way to fabricate "facts" without any source whatsoever. I've had the latter happen e.g. when discussing the founding year of the famous Sestroretsk Arms Factory near St. Petersburg. ChatGPT fabricated a nonsensical year for it and when prompted how it came up with that year it simply got stuck in an apology loop without explaining how or where it got the year from.

Another example of that particular behaviour is how recently someone fell into a delusional spiral as ChatGPT kept feeding him completely fabricated claims as facts:
View: https://youtu.be/UZdibqP4H_s?si=kncIR6JakWISXAcN

Whenever I see others gleefully paste on social media what ChatGPT replied to their history-related query, usually without any kind of checking if the response even makes any sense, which is about a daily occurrence these days (in fact I just saw another such post), the error rate is much the same: 100%.

Maybe I'm just unlucky, but I do think a lot of folks don't really understand just how limited language model chatbots are as researchers.

GIGO is GIGO, no matter if the research is human or AI. But I'm not sure AI can spot contradictory sources and use reason to reconcile them. I see so much bullshit pumped out from AI that I don't trust it at all.

 

[NVSMITH]

I've seen a few AI generated WWII videos and they are uniformly bad.

-Uniformly bad? Try uninformedly horrible! There are a couple of these... monstrosities... dealing with female German POWs being oh so wowed by the kindness of Texas cowboys.
-PUHLEEZE!! I haven't watched any; just scanned the "uniforms" being worn turned me off.
-And the AI is simply horrible.
-I suspect the female German POWs got to Texas flying on B-29s looking for the Pacific Ocean.

 

[Thumpalumpacus]

-I suspect the female German POWs got to Texas flying on B-29s looking for the Pacific Ocean.

It won't be long before another AI episode will find your post here and draw comparisons to Columbus or Balboa ... and maybe six months after that we'll learn that Magellan was Spain's third-highest ace in 1938.

I mean, that's what Lincoln posted a few years back.

 

[GrauGeist]

-Uniformly bad? Try uninformedly horrible! There are a couple of these... monstrosities... dealing with female German POWs being oh so wowed by the kindness of Texas cowboys.
-PUHLEEZE!! I haven't watched any; just scanned the "uniforms" being worn turned me off.
-And the AI is simply horrible.
-I suspect the female German POWs got to Texas flying on B-29s looking for the Pacific Ocean.

I thought the only POW camp for the few female German POWs was located in upstate New York.

 

[wuzak]

View: https://youtu.be/CSm1bMT_R6g?si=hwjy7DiSG4YJaJNr