16:28
Hi, welcome to another episode of Cold Fusion. What's up, guys? Uh, so Justine uh tweeted a really good article how to do these character swaps that really are going viral uh everywhere. Um, I posted two yesterday and I think the day before and uh we almost got like 10 million views. So, the technique he just saw is called a style transfer. It basically allows anyone to record themselves and digitally transform themselves into anyone they like. This recently went viral on X, but the technology has existed since the late 2010s, and we've covered it here on Cold Fusion back in the good old days. But the difference now is that it's photorealistic and anyone can do it. This is a spray of cold water to the face. But it's just
one example of how generative AI is wreaking havoc on the very medium of video, from online news to social media. But YouTube, the largest video platform of them all, is currently undergoing a deluge of AI slop content. slot being low-effort AI generated content that doesn't even have to make sense. Its sole purpose is to make a quick buck. Essentially, it's a scam in video form. Here's probably a common experience for many of you. You've got older parents that are watching AI videos without them even knowing it until you point it out. And dangerously, much of those videos are filled with facts that aren't even
true. YouTube is slowly being overrun by AI slob. Today I'm upset because all I want to do in this life is listen to YouTube videos in the background while I do other things. And that experience has been ruined and compromised for me because of AI channels on YouTube. But right now, it's quietly becoming a landfill for AI generated junk. The biggest YouTubers on the platform are quitting because they feel like they don't want to compete with AI slop anymore. AI has destroyed the internet in less than 6 months. In the past year, you may have noticed that more of your YouTube suggestions are AI videos. Have you seen those 2-hour boring history videos with an AI voice and made up facts? Because of the long running time,
those videos make a lot of money. Lazy people have noticed, so much so that it became an automation scam on X. One user claims to be making $21,000 per month. Unfortunately for us, since YouTube destroyed the search function to increase algorithm engagement, the algorithm is feeding us these AI videos and it's turned into a bit of a hostage situation. The spread of AI content is only going to increase with each coming year. With the right optimization, AI systems can pump out 30 videos in a day, whereas a humanmade video can take days, weeks, or even months. There's just no way for humans to compete with that volume.
Here's how you make $500 a day posting kids songs on YouTube. Step one, go to YouTube and look at nursery rhymes and kids videos and you'll see all of the cartoons and the things that they always watch. Then you're going to copy that, put into something like notebook LM, and that's going to transcribe the entire video and give you a summary. Then you take that, put it into Claude and say, "Turn this into a nursery rhyme." Then you take that and put it into Cling 3.0 or Higsphere, then say, "Generate this into a video." Then you take that, then you post it on YouTube, and voila, now you're monetizing. So the question is, what happens when the slop takes over YouTube? According to some reports, YouTube is already 20% AI slop on average. It seemed like in a
couple of years, most videos on the platform would be AI generated. All looked hopeless until mid2026. Finally, YouTube CEO Neil Mohan woke up. YouTube will now feature an automatic AI detector on videos that informs users whether they're watching AI content or not. That's good on the one hand, but on the other hand, Neil is pushing more AI tools on the platform and encouraging creators to use them. We miss you, Susan. In this episode, we'll see what's going on with YouTube and their battle with AI generated content, the absolute mess it's caused, what's being done to stop it, and what the future holds. We'll also answer a
few other questions. Are there even any good ways to use generative AI in content? You are watching Toll Fusion TV. Generally, a lot of people are sick of AI and assume that everything is AI until proven otherwise. Even for me, after running this channel from 2008, I still catch stray bullets from newcomers who assume I'm AI, and that's just because of the way I talk and my voice. The funny thing is, I'm the one who's had their voice stolen to use on other YouTube channels. Here's an example of my voice being stolen for a true crime channel. A federal law enforcement officer investigating mail theft gets arrested for mail theft. while
investigating mail theft in the neighborhood that called him there to investigate mail theft. But it gets worse. Other channels have had AI theft in various ways. The channel in question, Loose, took a screenshot or downloaded a portion of the video they wanted to steal from, and then they used AI video models to replace elements to match their voice over. We can tell it's the exact same scene because the dark splotches on the table are identical, as is the texture on the floor, the position of the light, and the general composition. Same thing here. We see here that the style is identical, yes, but so is the pattern of bricks in the background. No longer do you have to tediously copy someone's work.
Now the process is automated. before it was taken down because of my plagiarism complaint. The channel actively deleted comments that negatively mentioned AI or plagiarism. I'm sure AI has some good uses for YouTube, but I don't think ripping off other content or automating the human thought that goes into a YouTube video is one of them. Even for me personally, I'm not comfortable with using AI to write scripts or doing the voice over because I just feel it's disingenuous. Sure, it would be more efficient, but at the end of the day, it's not even the thoughts of a human. So, what value would I be giving to you if that was the case? But let's slow down and zoom out a little bit. Let's
see the current battle between YouTube and AI slop. I think AI slop on YouTube has a place for mindless consumption. But the speed and the volume of the increase is a concern. A Guardian analysis revealed that almost 10% of YouTube's fastest growing channels were AI slop. Billions upon billions of views worth. On YouTube, the makers of low-quality AI generated content are banking about $117 million per year. The AI slop channel from India, Bandai Apna Dost, features an anthropomorphic monkey and a dollar store Incredible Hulk. The channel now has 2.4 billion views, and it's estimated that rogue monkey rases in over $4 million yearly. So, let's watch one and see what all the fuss is about.
Someone typing a prompt into a text box is a pretty simple way to make money. And you know the saying, monkey see, monkey do. So now a bunch of people see those dollar signs and they're all getting in on the slob action. Max Reed, a journalist who has written extensively on AI slop, tells the Guardian, quote, "There are these big swaths of people on Telegram, WhatsApp, Discord, and message boards exchanging tips and ideas and selling courses about how to make the sort of slob that will be engaging enough to make money. Even if they're a documentary orformational channel, they don't care about the information. They just want to make a quick buck. This kind of thing is all over X." YouTube niches that will pay at
least 5K a month a part three. The hottest channel in the YouTube automation space is Zen. having 7.3 million views in just one month on their biggest video with his amazing idea, what did ancient humans do at night. This one video started an entire new niche, what did ancient humans do? blank, which we see another channel posted, how did ancient humans kill boredom? Another channel just 2 weeks ago. First video at 240,000 views. This one at 1.3 million views. And one of the craziest examples are those users who make tutorials and courses on how to generate fake female influencers to create income. Really, what they're optimizing for is the most monetary
return for the least amount of effort. It's popular to have AI slop that's all about following trends, appealing to children, or appealing to adults who want to watch mindless but visually interesting, unusual images on a screen. No shame if that's what you're into, but a friend of mine put it this way. It's like the adult version of being an infant entranced by a baby mobile. So, anyway, in pure numbers, just how bad is the AI slot problem? A study conducted by Capwing estimated that for a fresh unsigned in account, YouTube is already 20% AI slop. And to me, that's fascinating. Imagine being a young adult or a kid today opening up an account only to be fed sludge. The country with the largest
growth in lowquality slop content was Spain. With Egypt at number two, then the United States, Brazil, and Pakistan rounding out the top five. In terms of view count, South Korea loves it with Pakistan in the US completing the top three. The study also took a look at the generic YouTube experience without signing into an account. If you consider shorts, it's a complete disaster. An alarming amount of content fed to the account was meaningless. The study explains, quote, in total 104, 21% of the first 500 videos were AI generated and 165 33% of those 500 videos were brain rod. Of course, YouTube only serves up what it sees as popular or engaging. So, people must be watching this style of AI video. When you consider that for creators, AI generators mean very
lowcost or even no cost content, it's not hard to see why there's been a sudden rise. And if YouTube reinforces that by pushing the videos and rewarding the creators, the rise will only continue. End quote. Another sad thing is AI music playlists. YouTube used to have one of the best algorithms when it came to underground music, but now a lot of those mixes are just AI slop. The internet itself has a widescale AI slop problem. Some numbers have estimated up to 50% of published articles on Google are AI generated and some social media could see up to 79% of content having some form of AI involvement. And this includes online
news and media. I recently did an interview talking about AI slop and soccer media. With the World Cup coming up, it's an interesting take on the epidemic. You can check it out here. But coming back to YouTube, for the longest time, the problem wasn't being helped by YouTube management. It seemed like they just didn't get it. Up until May 2026, YouTube was barely doing anything about the problem. Bizarrely, they were leaning into AI video tools for creators. It's one thing to help creators with analytics using AI, but actually doing the video generation for creators is a completely different thing.
The YouTube channel Jack Films tried out one of the new AI features with horrific results. So, what you're seeing here is the original short paint on a pendulum dripping out paint and makes these cool uh designs, right? Jake pores paint. That's the YouTube channel if you want to watch more from this creator. Now, on the bottom of the shorts, there's a few different dots and these are AI remixes of the original short. So, here's YouTube's AI version. No way. What is happening? THIS IS INCREDIBLE. I CAN'T believe it. No way. Pretty cool, right? YouTube is pushing this. YouTube is like, "Hey, if you liked that cool organic humanmade video, how about a clanker version?" Your ears are
assaulted with the most AI sounding robot voice ever. No way. Incredible. And you, and if you read here, it says remixed from Jake Pors paint. For this clip, it's a real video of Red Bull launching a real human through a ring of fire. And here's the AI version. YouTube, can you please give people the ability to make this worse with AI? Oh my god. What the wa? That was close. Wow, look at the lights. What does this mean? What is that? For some reason, YouTube thought it would be a great idea to create senseless AI remixes of shorts. At this point, these changes seem like ideas from people that don't even understand the users of their platform. So, what does YouTube have to say about AI? The
CEO of YouTube, Neil Mohan, was asked about his views on AI in 2025, and he stated to Wired, quote, "The genius is going to lie. Whether you did it in a way that was profoundly original or creative. Just because the content is 75% AI generated doesn't make it any better or worse than a video that's 5% AI generated. What's important is what's being done by a human being." End quote. In an ideal world, sure, but that's not how the real world functions. Probably some of the most viewed videos if you go by views are AI generated shorts. That is a bleak future, one that a lot of people have expressed their concern over, hoping that YouTube would step in and maybe do more to stop the platform
just becoming a cesspool of just AI trash. Neil, on the other hand, the CEO, believes that's actually a really good thing because it's giving people that lacked the skills and equipment the ability to put that in front of you. In 2026, as every lazy person on Earth understood that they could make money from bad AI content, the scams continued. In this video, you're going to learn exactly how you can make an animated YouTube channel completely with AI with zero workers. This channel, Lost Legacy, is getting a ton of views. Just take a look at the view count on their videos. So, we're going to copy their niche. We can press this copy style button, and this is key for what we are
doing here. Now, this is actually going to analyze all of their videos. It's going to scrape all of their scripts and it's going to create a script writer that writes in the same way that they do. As the sloppification of YouTube gradually got worse, the CEO pivoted and changed his tone slightly, explicitly saying that the platform will combat slob. Quote, "The rise of AI has raised concerns about low-quality content, aka slop. to reduce the spread of low-quality AI content, we're actively building our established systems that have been very successful in combating spam and clickbait and reducing the spread of low-quality repetitive content." End quote. But not long after saying that, he announced that users would soon be able to use AI generated shorts using their
own likeness. Then on the 28th of May 2026, Neil finally woke up and he announced that YouTube would be rolling out a new feature that autodetects AI content. If it looks real but was made with AI, viewers will know immediately. We're also making disclosure a bit more seamless. While you're still required to manually disclose realistic AI use, starting this May, YouTube is rolling out internal signals to help. If YouTube systems detect significant photorealistic AI and it hasn't been disclosed, will now apply that label automatically. If you think that the system has gotten it wrong, don't worry. Creators are still in control. As this technology continues to improve, if a label is applied incorrectly,
you can jump into Studio and update the status yourself. The only exceptions where the label does stay permanent are if you used YouTube's own AI tools like VO or Dreamcreen, or if the file contains metadata, proving it's fully generative. Most importantly, for creators, these labels alone do not affect how our videos are recommended or whether they can earn money. This is purely about giving viewers the right information at the right time. Let us know what you think in the comments. Leaving aside the conversation about potential accuracy, this is a good move. It lets the users decide and choose for themselves if they want to watch AI content. The only reason an uploader would hide the fact that they're using AI is to mislead
viewers. So, let's be fair and more rounded here and add a bit of nuance. Generative AI can technically be used to make highquality content, but personally, I haven't seen much of that, and I think it's because the incentives are pointing the wrong way. With AI, it's much easier to quickly smash out something that gets engagement and earns money rather than building a grand story that took months of writing with AI just being the final step that brings your vision to life. To see what using AI properly could look like, here's an example of a short piece made by Simon Mayer. It's AI generated but not complete slop. It runs for about a minute and a half. Honestly, it's pretty simple to make films with AI. You just press a button. That's it.
I'm Thomas. I'm an AI artist, whatever that means. That's it. That's how we make films. I usually make about five films a day, sometimes 10. You know, ads, Super Bowl commercials, feature films. I do it all. So easy. Want to give you guys a little tour. So, this is our social media department. Lisa here handles all our LinkedIn stuff. What's going up next? So proud to announce that we pressed a button today. The future of storytelling is here.
Clients are always like, "Is this legally okay?" And I'm like, "Don't worry about it. We have an excellent lawyer in house. It's very safe. I checked everything. 100% safe." That clip felt different because it was created by a person who actually works in film. He came up with the concept, storyboarded it, and used AI to bring it to life. Now, that's more in line with using AI correctly as a tool to enhance your skills, but not just do the work for you. But in saying that, there's even the wider question of how these models were trained in the first place. While schools like London Film Works
have courses on how to integrate generative AI, the truly unfortunate thing is the vast majority of AI content uploaders right now are sloppers. Defined by Wictionary as quote, "Someone who was overrely reliant on generative AI tools, a producer of AI slop, careless, low-quality content generated using automatic computer applications and distributed to farm views and subscriptions, or sway political opinion." End quote. Throughout May of 2026, YouTube threw down the gauntlet when it came to the fight against AI slop. 4.7 billion views worth of AI content were deleted. The sloppers on X were panicking that the YouTube AI automation
grift was coming to an end. Some went to extreme lengths to appeal to YouTube. This AI slop channel hired an old man from Fiverr to appeal their case. I'm making this video because our channel Fluffy Animal Tales recently received a message from YouTube saying our channel was demonetized for inauthentic content. Every video on this channel is built carefully from the ground up by us. Our channel is about meaningful animal rescue stories. While researching this video, a few of those massive AI generated channels that made headlines were deleted. For example, the channel The AI World based in Pakistan made AI generated shorts of catastrophic flooding. The titles were a blast. Classics like Poor People, Poor Family, and Flood Kitchen. It had 1.3
billion views, but YouTube nuked it. Not long after, YouTube added an option for users to vote if they felt that a video was AI slop and to what degree. I suspect this data played a part in training the AI flagging algorithm. But in some aspects, it seems like YouTube has overshot. YouTube's AI for determining whose channels should and shouldn't be banned, frankly, has turned out to be a disaster. I don't know the figures, but innocent people have been catching stray bullets. They've been banned by YouTube's automated flagging. Another example is this Korean channel. She created handmade stopmotion cooking videos inspired by her mother.
The YouTube bot saw this and thought it was AI, so it banned the channel. It was reinstated after a public outcry. Now, this one's egregious. The channel's splash plate was banned because someone lazy stole and reuploaded his content. Although the lazy re-upload channel got terminated, the original splash plate channel also got terminated. YouTube's AI wrongfully assumed that this was another re-upload. YouTube basically ignored him when he protested. Quote, "This is the email I've just received in response to YouTube support. They believe I'm still circumventing a ban and my termination is correct. This is despite all the
proof that this is a blatant error in their AI." End quote. It took months, but he did finally get his channel back. Nanny Josh was another example. He had a channel terminated for spam or scamming even though he painstakingly made animations and 3D models himself. His appeal was rejected five times. And if you go on to X, there are so many YouTubers who have been banned because the AI hallucinated. Let's just say this clearly. AI in its current state just isn't ready to make such decisions. And because AI should never be able to be the judge, jury, and executioner, it should never have the ability to terminate a channel. So, where does this leave us? To fix the main problem, YouTube needs
to restore trust, and the AI flagger on content is a great start. But the thing is, it has to work with a minimal error rate, and that could be a hard ask, but we'll have to wait and see if it does work correctly. That kind of method should be the standard across all platforms that contain video content. I think the speed and the breadth of AI generated content did take YouTube by surprise. Hopefully, they can navigate it without being overcome, but it's going to be a messy road. So, look, I do think the technology of generative AI itself was fascinating from a technical standpoint. That's what I documented from its rise in 2016. But today, it saddens me to see how it's used for exploitative, loweffort
content. The content that's disingenuous, misleading, or for pure profit. But the thing is, that's what scammy people have always done. So, why should I be surprised? On this channel, I've talked a lot about how relying too heavily on technology can sometimes weaken the very skills we're trying to improve. And that's why learning how to actually think through problems is becoming more important than ever. And that brings me to something interesting Brilliant has been working on. This is the new Brilliant, a personal tutor for math and coding that sits right there with you while you learn. It can follow your reasoning, guide you through problems stepby step, and adapt in real time to how you think. What I
like is it doesn't feel like traditional studying. Brilliant turns learning into something much more interactive, almost like solving puzzles instead of memorizing information. I've been trying some of the lessons myself, and one thing that stood out is how the teacher nudges you in the right direction without just handing you the answer. So, instead of passively consuming information, you're actually building intuition as you go. If you get stuck, Cooji can help you out. For instance, I was working through their thinking and code course, and I got help directly from Cooji to figure out the next steps. Their curriculum covers everything from core maths concepts like functions and algebra all the way through to calculus, coding,
logical reasoning, and debugging. It's all designed by educators and experts from places like MIT, Harvard, and Stanford. What's especially interesting is how well this fits into everyday life. Unlike traditional tutoring, there's no scheduling or waiting around. It's just there whenever you want to jump in and learn something. So whether you're learning from scratch, brushing up on old concepts, or just trying to keep your mind sharp in an AIdriven world, it's one of the most engaging approaches I've seen. You can try Brilliant's new tutor for free by scanning the QR code on screen or heading to brilliant.org/ org/coldfusion. And if you want full access to all courses, Brilliant is also offering Cold Fusion viewers 20% off an
annual subscription. Thanks to Brilliant for sponsoring this part of the video. Anyway, what do you guys think? Do you think that in 5 years, if all videos were AI generated, nobody would care because that's just the standard? Are you of the view that it doesn't matter if an AI created it because all that matters is if the content is good enough? Or do you believe in human created content? And lastly, do you think YouTube's new measures go far enough? Let's have a discussion in the comments section. Hey guys, thanks so much for watching the whole way through this episode. It really means a lot. I also want to thank you for all of your support. It's been a crazy month, but yeah, just wanted to say thanks. Anyway, if you want to watch
anything else on Cold Fusion, there's plenty of stuff on science, business, and technology. And yeah, that's about it from me. So, you guys have a good one. Cheers.
