The tsunami of image fakery... where is it worst? Pointing to disaster?

Our AI-transition has many tell-tales that are changing daily. Let's start today with one that's a major danger-signal.*

Setting aside spasm-reward lobotomy addictions like Instagram or TwitX, the most-used middle-length content site on the globe is YouTube. And something disturbing has happened there. 

First: Google rewards content posters for both clicks and length of engagement. And hence, setting aside movie clips and formal channels like Sabine Hossenfelder, or Mat Dowd, or PBS, YouTube now swarms with lures and sticky, eye-retention tricks. In other words, clickbait. 

Whatever topics that your viewing history suggests might glom your eyeballs, there are predators swarming into your feed with offers. In my case, that might include historical riffs (e.g. WWII), or archaeology/human-origins, or new-science, and so on.

With the exception of reputable channels, almost all are now AI-voiced with AI images.  YouTube swarms with lures and sticky, eye-retention tricks. So yeah, clickbait.

(The trend hasn't yet struck the cool, practical how-to vids.** But give it time.) 

When it comes to YouTube clickbait from unvetted sources, there are three aspects to track, the voiceover, the images and the content.

== All three traits are now suborned ==

The narration-voice nowadays has excellent tonalities and mostly no longer pauses at wrong places. Well, only rarely.

As for content, for a while the long form vids were clearly reciting from some existing text: a news or science article, or a book chapter. So, the 'facts' recited by the voice might be taken as ... well... as something like 'news' or at least a knowledgable human's opinion. 

That ended about a month ago. Now, evidently, the unvetted stuff is nearly all pure AI/LLM-generated 'content' that's been prompted by some parasitic twit to "blather ten minutes of clickbait about...."

So much for voice and content. But it's the images - video scenes that accompany the purported 'text" -  that went bad long ago... even six months or so, back in the olden times of 2025 C.E. 

These generally take form as a series of B&W stills that seem convincingly like real photos from the era, apropos to the passage being narrated....

... except that often none of the supposed 'photos' are real! Not even one. Every single 'picture' has a blatant give-away, like implausible ships whose cranes would have toppled them in seconds, or arrays of trucks loading from 'liberty ships' all in completely implausible, tightly-packed order. Or a German staff meeting with a dozen admirals, all of them four-stripers (more than in the whole Kreigsmarine at the time) and all with similar ages and grim expressions while poring over a map whose outlines match nothing on Earth, with blurry Gothic lettering. Oh and the uniforms - extrapolated by AI - never happened.

Especially grating: a recent archaeology 'news-revelation' piece about the famed Turkish archaeological site Gobeckli Tepi repeatedly teased you to stay tuned till the end for 'shocking news' - standard click parastitism, rewarded by Google's nescient and lazy algorithms.

But the images are the locus of deep immorality comes in. Take that archaeology example. About 20% of the images were actual shots or graphs. 60%+ were blatantly AI-extrapolated garbage, like photorealistic scenes of Gobecki-Tepli's stone T-Monoliths on fire... yes, I said on fire. (The last 20%? I couldn't tell, they flashed by so quickly.)

And sure, not everyone knows enough to tell the difference. Which is what makes this dangerous!

A video about Palmira Island showed view after view of different made-up islands that clumsily matched each minute's passage of recited text. I think I spotted one -just one - that might have been real.

This linked example illustrates my points but similar fakes are all over the place, right now. 


== Google and YouTube could act on this and start a Truth fight-back ==

I haven't seen anyone, anywhere, point out that YouTube is likely now the very biggest sewer of cyber fabrication, anywhere on the Internet today. Meaning ever, ever in human history. Far woirse than Twitter or Tiktok, because longer format tends to carry more credibility. It allows more convincing lies, that use up more lifespan, per lie.

YouTube's owner (Google) could easily ameliorate this, say by putting a small metric symbol in a corner. Or two. One of them showing the percentage of AI generated content and the other icon clickable, so that viewers could score for accuracy/plausibility. Or even disgust vs. praise. Even better, content scoring the facts and images. (I don't care as much about the voice, though...***)

This could be where we try out some of the methods I describe in my new book on AI... AIlien Minds. Methods that lead these entities and their human accomplices to feel accountable for lying to us.

I could go on. But what is the real lesson? That AI illustrations are now not only photo-realistic in appearance, based on 3 sentence fragments of an ongoing narration, but also so cheap they can be generated by minor YouTube channels as special interest clickbait. And yet, for all the photo-realism and pertinence to the narration, they almost always lack any sign of checking against real world plausibility... which of course no LLM is truly equipped to do, anyway.

This was impossible 6 months ago. And six months from now the model systems will have been trained to better-fake their unaware plausibility. But likely they'll remain real-world absurd. And hence dangerous!  

(Note that six months from now are the U.S. Midterm el;ections!)

And maybe some most-advanced AI is reading what I just typed, reconfiguring as we speak. For well or ill.

== And so, let me plug... ==

I've been pulled into the Great Big Panic/Debate over Artificial Intelligence. 

You'll find unusual perspectives in my new book on AI... ailien minds... 

...that just went live on Kindle and paperback.  

Here's the cover copy:

Optimists foretell a golden age of Al-managed abundance.  

 Doomers cry: vast cyber-minds will crush old style humanity! ... or make us irrelevant.  

 Meanwhile, geniuses fostering the artificial intelligence boom. cling to clichés rooted in our dismal past... or else in cheap sci-fi.  Is there still time for perspective? - on 4 billion years of evolution - or 60 centuries of wretched feudalism - or how we handled prior tech revolutions - or mistakes that keep getting repeated - or ways this time may be different?   

 From Al-driven unemployment to deceitful images, to hallucinating LLMs and tools for tyrants... 

...to potential wondrous gifts by machines of loving grace... 

...come see future paths that evade the standard ruts. 

==================

==================

* 28 years ago, in The Transparent Society, I had a chapter: "The End of Photography as Proof of Anything at All?"

** How-to videos are way-cool and while clearly clickbait, they also deliver value across short timescales. But what happens when they are taken over by AI-generated fakery, too? People will get physically hurt.

*** If human voice-overs became a requirement, it would BOTH boost employment and ensure that some human participation in content creation remained in the loop.