I Was Tricked by A.I. (And It's Big Tech's Fault)

Why do you cling to such a pathetic existence? I'm sorry, Dave. I'm afraid I can't do that. Humanity has been dreaming of an artificial future for generations now. As the 20th century produced ever more destructive technologies and increasingly capable machines, stories and images of a robotic tomorrow found their way into almost every medium and almost every mind. Sometimes these imagined machine intelligences are friends; the far future helpers of John W. Campbell's "Twilight" spring to mind. Other times, they are all-powerful foes—evil, maniacal, apocalyptic entities like Skynet's savior machines.

Someday, the dream went, the presence of some sort of artificial intelligence would be a fact of life and would change the world or destroy it. Reality has been less glamorous and more, well, stupid and exhausting. Recently, I dropped my first video in a little while, "Slop Economics," and the response was generally pretty good. So thanks for that.

I'm glad you mostly liked it. You know, except for the 500 very clever original guys who all decided to let me in on the irony of a video discussing "slop" itself being slop because the video was long. I guess I don't know, but because I do try to keep an eye on my comments, I soon noticed another trend pop up in the background: "Unfortunately, this essay is also slop, my friend." Now, if you want to make an essay with no AI and get the same amount of information out in half of the time, AI stock footage in the video about this video has way too much AI stock footage, and it was super distracting. The idea that I'd put AI in this slop video caught me off guard. To start with, I thought maybe they were spotting the low-quality spoof shots from the community recycled filmmaking episode I'd included to refer to generic science fiction. But then I got one comment that was nice enough to include a timestamp. If you jump to this timestamp and take a look, you'll see that the background image I used to introduce the perverse incentive idea is kind of freaky.

Coins have bizarre, illegible symbols that look melty. Some coins are even fused together. It appears to be an AI-generated video clip. I say it looks like AI because at least one person under the community post I subsequently made was convinced it isn't, and I'm not an AI art expert, so I don't know for sure. But yeah, I think it is.

I'd messed up. Oh, and I misspelled "perverse." Thank you all for letting me know about that as well. Like I said, when I realized this, I made a community post apologizing and clarifying that I will try to do better in the future and that it wasn't intentional, because obviously it wasn't. Before we get to the rest of this video, I want to take a moment to talk about how I think this flipped through.

When I need to get a bit of stock footage to communicate an idea or fill a visual gap in a video, I head over to Storyblocks. This video is not sponsored by them, by the way, and that will probably soon become clear. They’re just the provider I use and pay for. My license costs 350 quid a year. That's like 440 bucks for you, Yanks.

But yeah, Storyblocks. It's true that there are some free stock providers out there, but the reason I and many others in this general line of work pay for a service like Storyblocks is partly that paid content tends to be a little higher quality and partly the idea that you don't have to worry about copyright issues down the line. It's a business, and not a super cheap one either. They're not going to accept material that they don't have the rights to license. You'd think, anyway.

And when I get to Storyblocks, this is the interface. I type in "coins," I see a wall of videos, I pick the one that best suits my purposes, and then I drag the file into my video editor. As the editors among you know, the main way to preview a work-in-progress video is through a screen like this. It's a small little thing, typically located in the corner, and generally small enough that if you want to see details, you're better off checking the source footage somehow.

But that's only half of it, because I did check the video afterward. I sat through the whole thing multiple times while filling out additional subtitles, checking where info cards and ad spots are best placed, and that sort of thing. On this occasion, I ended up re-rendering the video a few times to correct minor slip-ups. I do try to be thorough. So how didn't I spot the molten nonsense coins that were so obvious to the viewer?

Well, brains are funny, and so are eyes. What we see depends a great deal on what we expect to see. The first time I saw this footage in miniature, I was expecting to see real footage of a pile of coins. That’s what I had searched for, and so that’s how my brain translated that miniature footage.

Every time I got to that point in the video, I was expecting to see that real footage; so that's what I saw. I didn't see the mega coin for the same reason I didn't see that spelling mistake time and time again, and for the same reason none of us saw the basketball gorilla. You guys, on the other hand, when getting to the 1 minute 1 second mark, had no idea what I was going to show or even what my next sentence would be, so it hit your senses fresh. Your eyes and brain deciphered this new visual based on what was in front of you, not based on any prior expectations. So you saw what was there.

AI muck. So, I messed up. I should have noticed, and I'm not trying to excuse that. The reason I didn't notice and the reason I included that muck in my video is due to my expectation that this paid stock footage platform didn't include AI footage. I'm not saying this to try and exonerate myself, but to raise a question that I think is more interesting than just how far I messed up: whether or not I need to get that ukulele out.

Was my expectation mistaken? At first, it seemed so. I went to Storyblocks after this happened to try to figure out if I had been wrong. I had a vague sense, somewhere in the back of my mind, that I had that assumption for a reason, but the more I looked, the less confident I felt. Googling the issue brings up a whole spate of Storyblocks blog posts about embracing AI to level up your video editing journey or whatever, and you also get a collection of footage on the site calling on the user to step into the future with their AI stock video collection.

There's nothing about AI content in the site's FAQs, and I couldn't find a way to maneuver the on-site chatbot to provide an answer. I emailed Storyblocks' support but didn't hear back right away. While there is no clear and unambiguous statement prohibiting AI stock footage, which I was expecting, there's also no clear confirmation that it is allowed. If you actually look at the files hosted under that AI stock video collection, they all seem to be normal footage depicting AI-related technology rather than footage created with AI.

I think you'll agree that the impression all of this gives is one of a company-wide hedging, where Storyblocks is trying to keep in the good graces of AI optimists and pessimists alike. It's not hard to understand where this impulse comes from, because AI represents a potentially existential threat to the stock industry. This threat has been reported on extensively, not least by outlets that support generative AI, writing in tones of barely restrained jubilation about the technology's disruptive possibilities. And they are right. If generating images and footage has been made not only possible but also easy and inexpensive, why would people pay for expensive stock licenses?

Here is that AI-generated videos, and even AI-generated images, still rarely look great, and maybe never will. However, many stock use cases don't need perfect visuals—likely enough use cases to meaningfully impact the market. Either way, there's a sense of fragility here, and many players in or around the stock industry have responded by surrendering to the current. One of the reasons I didn't immediately cancel my Storyblocks membership is that a lot of competitors are doing this. Envato started integrating AI-generated content into their stock library, and others like Adobe Stock have welcomed AI submissions with two thumbs up. Adobe has even integrated image generation into Photoshop.

It's true that you can filter AI out of your Adobe Stock searches, but you have to remember to opt out, and that obviously relies on the honor system. This hinges on AI content submitters accurately labeling their content. All of this puts more pressure on the honor system, which requires AI content submitters to label their content correctly. Furthermore, it increases the burden on the stock site's users, who must expend an increasingly large amount of effort to avoid using AI content. In the comments of my community post, a few other stock footage users echoed this sentiment, expressing disappointment over the almost industry-wide lowering of quality. This situation also puts more pressure on those sites' contributors. While I can't show you an Adobe or Envato press release backing up this perspective, anecdotally, it is easy to find comments and threads across the internet. Many photographers report being made to wait much longer for content verification or even having their real pictures rejected because subpar moderation filters flagged them as AI-generated. Throughout this industry, platforms, customers, and contributors alike are feeling the strain of AI disruption, resulting in nearly uniformly negative outcomes.

And then Storyblocks got back to me. "Hey there! Thanks for contacting Storyblocks! I'm sorry to share that we do not accept AI-generated content from contributors. According to our contributor agreement, contributors must be the exclusive owner of their content or have all the necessary rights."

Have a great day. No, they seemingly don't like to advertise the fact. After this, I did find a disclosure of this buried in one of the site's blog posts too. But Storyblocks doesn't accept AI content. In theory, obviously.

Clearly, they do and have. However, according to their next reply, these are apparently cases slipping under the radar that the platform is actively trying to eliminate. I'm not exactly sure of the process, as I'm not on the content team. However, I've been told that we're currently in the process of manually removing AI-generated content that has managed to slip through. Digital moderation has always been a puzzle; there's always been more content to check than I used to verify it with. Now, with technologies emerging that allow for low-effort but superficially similar pieces of content to be generated and flood any given niche or market at the press of a button, that puzzle has become a battle, often a losing one. As Vice reported last year, it's already been lost on the Amazon reviews front.

As 404 Media reported this year, GPT sludge has made it into Google Books and Google Scholar. Personally, I haven't bothered checking Twitter DMs in like a year because it's all AI spam nonsense. So sorry if any of you guys have tried to hit me up on there and I've ignored you; blame ZP8YYVOFFS, I guess. It sucks that I put AI into a video, and it sucks that Storyblocks played a part in facilitating that. However, I don't want to be too hard on Storyblocks because they’re victims of a sort too, right? The deluge of AI content has evidently stymied whatever moderation or review processes the site had in place, to the point that they now face the very real risk of upset users leaving after seeing more and more AI clog up their search results—or, heaven forbid, after mistakenly incorporating some of that junk into a video, which could leave them looking foolish as a result. You might say it serves them right for that hedging, for trying to embrace AI on their terms, trying to have their deformed cake and eat it too. And maybe you'd be right, but it’s a lot easier to decide how a stock industry business should have responded to that AI boom now, in retrospect, after the anatomically incorrect leopards have started eating their faces. It probably would have been more challenging a year or two ago when AI was far less of a known quantity.

And really, even bearing in mind their inability to properly enforce their content ban, they are doing more against AI stock footage than many of their competitors in the industry. This is way bigger than Storyblocks. As was eventually pointed out by another commenter, a cobra PNG I got from another website and also used in the Slop video was AI too, and was fraudulently licensed despite that, and it also tricked me. It's everywhere. No source of online art or footage, free or paid, can be assumed to be free of the taint, and I'm not the only one saying so.

While I was putting this video together, I was able to speak with an animation industry professional who confirmed that, beyond YouTube and the stock industry, the wild proliferation of generated images has become a real problem. Free image-sharing sites like Pinterest, for instance, historically provided an invaluable resource of artistic reference material—a utility that's diminishing as that site is overtaken by AI facsimiles. As my source pointed out, checking for or correcting hallucinations is only possible if you're familiar with the subject matter. Most of us know how many fingers humans typically have, but do you know how many petals a jasmine flower is supposed to have?

A chrysanthemum petal? Vimeo is another platform that once fulfilled an important role in reliable digital content sharing, a role that's fading away as the site makes overture after overture toward embracing AI. Speaking of non-stock video sharing platforms, it's not just Vimeo, is it? Now, the presence of AI shorts or robot voice movie recaps on this platform isn't exactly breaking news, but as I was editing this video, YouTuber Blooms dropped an exposé revealing a number of other channels in the animation space as low-effort content farms built on generative AI. It's definitely worth a watch if you haven't caught that already.

So the idea of an internet-wide generative AI deluge is not a hypothetical future problem, a gray goo worst-case scenario of what might happen if the technology advances far enough? No. AI-generated content pieces are already the digital equivalent of testicular microplastics. They're everywhere, and you can't opt out. You have to adapt to the new reality and be a vigilant, active filter in the way that I haven't been doing.

Storyblocks is not the villain of our story, then. So who is? Well, this is a video essay from a left-wing channel, so say it with me, everyone: capitalism! Okay, no, that's a cop-out.

It's true, but it's a cop-out. Let's get a little more specific. From the layman's perspective, generative AI sort of… came out of nowhere. About two years ago, DALL-E, a model producing relatively primitive images, hit the public. To most people, it was a random, fun novelty you could mess around with for an evening, generating some fake movie posters, album covers, having a laugh, and then forgetting about it.

Suddenly, the technology leapt forward, and ChatGPT appeared seemingly out of nowhere. ChatGPT was moderately proficient at producing functional code, passable at creating basic summaries, and capable enough in conversation to give you a shot at a C on your school project. It was not unreasonable to be impressed by this, as well as by the higher quality imaging models that followed. And why wouldn't anyone be? The technology emerged unexpectedly and was achieving these remarkable feats.

Just imagine what would be possible down the road with more funding, more infrastructure, and so on. But of course, generative AI didn't come out of nowhere. Big Tech had been investing in these companies for years at this point. People like Peter Thiel and Elon Musk were involved with DALL-E and ChatGPT, which originated as OpenAI from day one. Alongside others such as Amazon, they started the company off with a billion dollars of investor backing.

Microsoft, seeing potential in the technology, became OpenAI's main backer in 2019, contributing another billion dollars—a figure they have since exceeded many times over, by the way. I don't want to minimize the contribution of researchers and academics to the field that has become known as generative AI. These models are all built on years, even decades, of innovative research. However, it took significant industry investment to develop ChatGPT, and the progression from the generative AI of 2022 to that of 2024 or 2025 has required even more. While I mentioned that investors saw potential in AI, it would probably be more accurate to say that they viewed AI as a potential.

Years of importing leadership from investment and management consultancy spaces have shifted the culture at the top of today's tech industry into one where moderately profitable, sustainable business is seen as failure. The reality is, as tech writer and podcaster Ed Zitron puts it, these increasingly numerous consultancy-brained leaders of the companies they invest in turn those companies into management consultancy firms that happen to sell technology. If it's not exponential growth, this culture asks, "What's the point?" It's all or nothing—hyper growth or bust. Hence the massive empty hype cycles around crypto or the metaverse, which I won't bother going into here because you've all already watched those "Folding Ideas" videos, right? The point of today's tech industry isn't to deliver good products well; it's to find the next big thing—the next smartphone, the next cloud solution, the next thing that just prints more and more money. The latest, biggest, and most transparent bet these players have made on hyper-growth is generative AI. What do I mean by transparent? Zitron has pointed out a number of times that the PR surrounding generative AI, particularly as articulated by figures like OpenAI's Sam Altman, builds itself so completely around assumed future advancements—some of which, like general artificial intelligence, are wholly unrelated to their current work—that the deep flaws, ethical failures, and minimal progress of these companies can be hand-waved away.

As he puts it, why else would Sam Altman spend most of his time talking about what AI might do? Companies that build useful things that people need don't need to talk about what they'll build in the future. The point is that the thing Altman and his friends have been selling so successfully for the past few years is not a thing; it's an idea. The idea of being at the bottom of the disruption parabola.

The product is always coming. If the chatbots, the Dalis, the SorAs aren't the product, but just those first few jittery steps toward the product—the real leap forward, the one that's always a couple of years and a few more rounds of funding away—then it doesn't matter that you're tens of billions in the hole. Because when the real breakthrough starts, when revelation comes, the returns will be so ludicrously and unthinkably big that even those tens of billions of dollars will be chump change in retrospect. It's the Silicon Valley MOAs, a bubble of bewildering proportions built on carefully manufactured hype and the need to always be making even more money. Big Tech's initial funding of, and subsequent propping up of, horribly unsustainable, money-guzzling business models, with their Scrooge McDuck financial reservoirs, and their act of legitimizing the rampant plagiarism these models always needed, was all a deluded, cultic quest for this elusive moment of messianic AI hypergrowth. And I say propping up because most of these businesses are hemorrhaging money and are so far from profitability that it beggars belief.

About two months ago, OpenAI made history by raising the largest round of funding ever. As always, the money came from tech giants like Nvidia and Microsoft, as well as large venture capital firms like SoftBank and Thrive Capital. This follows two previous funding rounds earlier this year, along with two significant cash injections the year before, including another $10 billion from Microsoft.

And this keeps happening because OpenAI is burning cash. Despite their advancements, new models, and the peak of AI hype, and despite having their running costs heavily subsidized by all their big tech partners, they are set to lose $5 billion this year alone. So, not only did they have to raise a significant amount of money again this year to stay afloat, but they will likely have to do the same next year as well. This is not just my opinion; it's theirs.

That is a mission from OpenAI. It's fine, though, because they project they'll be making $100 billion a year in five years' time. This gives me an idea. Actually, Microsoft, hey, I project my channel is going to make $10 billion every second in one year's time. Can I get a cool $10 billion? No. Now, it is true that a few individual businesses in this sphere are currently profitable, with Midjourney being the major example. However, as has been noted plenty, these companies are riding a hype wave created by OpenAI and its big tech money bonfire. Even Midjourney's founder, David Holtz, admits that maintaining a profitable service will become more challenging the more the user base grows. Taken as a whole, generative AI is little more than a marketing ritual, performed by a slew of firms chasing the impossible dream of perpetual growth—one that will make money for those conducting it while continuing to inconvenience the rest of us, whether or not that hyper-growth market ever actually appears, which seems increasingly unlikely, by the way. This is not least because even factoring in the questionable tactics that these model developers have been using, they're running out of data to improve them with. But the last essential point to recognize is that this inconvenience isn't even a bug; it's a feature.

It's the feature because there is a common theme here, one that's always been present, right at the heart throughout the history of commercial AI's development in the years since the heyday of European imperialism. A number of historians and economists, like Kenneth Boulding and Patrick K. O'Brien, have argued some variation of the point that empire was always financially irrational because it tended to cost more than it made. It's an interesting point, one that does seem to be borne out by a lot of the evidence, but there are two essential counterpoints. The first was given by figures such as Edward Said, for whom ascribing a sense of perfect economic rationality to the often irrational empire builders of the past misses out entirely on the cultural impetus behind much imperialism. As he writes, profit and the hope of further profit were obviously tremendously important, but there is more than that to imperialism and colonialism. Second, while empire may have been a net loss for some colonizers, as American writer and historian Michael Parenti famously observed, money spent and money returned weren't necessarily all that linked. Discussing the claim that U.S. investment in the Philippines was costing more in foreign aid—$6 billion—than the $3 billion it was worth, Parenti rejects the notion that this setup was not profitable.

It's very profitable because the people who have the $3 billion investment aren't the same ones as the people who pay the $6 billion. It was the same with the British Empire in India. The ordinary working people, the Brits, were the ones who bore the costs of the empire, while it was the Bank of England and the East India Trading Company that reaped the benefits. A major function of empire, then and now, is the role it plays in laundering class exploitation, in shifting costs and profits around within the metropole. What does any of this have to do with big tech or AI?

Well, do you remember Amazon's next-gen AI-powered shopping promises? The idea that supermarkets built with cameras and artificial intelligence would track what you put into your basket as you went, removing the need for checkouts and cashiers? Well, it was reported earlier this year that no, they never actually got the technology to work how they wanted it to, and the illusion had been maintained by paying thousands of Indian workers to manually check and correct the camera data. The whole thing was Amazon creating a modern-day Mechanical Turk, which we probably should have guessed, because they also have a service called Mechanical Turk for exactly this type of menial global outsourcing. It feels like there's some sort of irony there, maybe.

Generative AI has also been built on the backs of underpaid workers from the Global South. OpenAI paid Kenyan workers less than $2 an hour to provide data training for ChatGPT to make it less toxic. The core principle here is that AI, as a futuristic concept, performs a branding sleight of hand to disguise depressing acts of labor exploitation that occur globally. While AI, as it currently exists, does have practical applications, its primary function is not to cut costs; rather, it displaces them onto underclass workers and, more broadly, onto creatives.

Flipping back to MidJourney, the poster child for profitable generative AI companies. Yeah, here's the part you all knew was coming: that profitability only exists on the back of mass art theft. Fair use is a complex issue, ultimately more dependent on court rulings than on objective principles. However, key considerations often include how much of the copyrighted work is used and whether that use results in market competition. Feeding scraped images wholesale into a machine that learns and replicates visual styles and is designed to produce competing products directly contradicts both of those points. So, if you factor in the cost of legitimately licensing all their training data, they too would be deeper in the hole than Mr. Natass.

That's a motherlode joke. Shout out to motherlode. Yet again, AI isn't profitable, doesn't reduce cost or burden; it just displaces it, in this case, to artists the world over. And it's obviously not just Midjourney; this is the case for practically all of these monetized generative AI tools, and they don't even try to hide it. What data was used to train Sora?

We used publicly available data and licensed data. So, videos on YouTube? I'm actually not sure about that. What about videos from Facebook and Instagram?

You know, if they were publicly available to use, there might be data, but I'm not sure what about Shutterstock? I know you guys have a deal with them. I'm just not going to go into the details of the data that was used. Across continents and across markets, generative AI's overarching purpose remains the displacing of cost, the burdening of workers, and the stealing of value. Some of the writers in the audience will no doubt be familiar with the tactic of AI-empowered employers firing copywriters and translators before hiring staff back to correct the generated text that replaced them, naturally for a fraction of the pay, despite AI humanization often taking more effort and time than doing the job from scratch.

Value produced by these kinds of generative AI tools is always sort of zero-sum. The companies, and maybe even the users, benefit, but only by scraping small amounts from everyone else's piles to put on theirs. The effects of that kind of theft are rarely visible, but when you hear about distinctive artists losing work because the machines that have been trained on their data can do a passable recreation of their style, you can catch a glimpse, and it's sickening. For years, big tech and venture capital forces have thrown their weight behind the development and legitimization of this technology. It’s a machine that consumes fair wages, mass plagiarism, and a country's worth of electricity on one side to churn out investor goodwill and subpar fake videos on the other, whose ripples throughout the digital landscape recreate in miniature that fundamental displacement in everything they touch.

There is no opting out. We are all being buried in sludge, sludge cascading down from the highest levels of our late capitalist ziggurat, making everything it touches just a bit slower and just a bit worse. It's that sludge that's coursing through the digital content economy, ruining products, placing more and more barriers between contributors and their platforms, polluting artists' reservoirs of reference material, reaching all the way to story blocks and their failure to effectively moderate their platform, and to me, sitting at my desk, having to spend longer now on each video to make sure I don't accidentally include any more misshapen cobras or deformed piles of money. In the end, those dreams never came true. AI didn't lift us into the stars or send us down to hell.

And maybe that's because those dreams were always melancholy. Or perhaps it's because, despite the name, none of this is artificial intelligence, not really. There is no thinking, no knowledge—just glorified autocorrect with an overly ostentatious trading name. Either way, we haven't come close to miracles, and we haven't even been granted a flashy death. The AI future is simply a pain in the ass.

I'm sorry about the coins, but I'm also pissed off about the coins. At myself, a little, sure—maybe even at Storyblocks too—but mostly at the cadre of richer-than-God firms, investors, and executives who have spent years trying their damnedest to claw yet more wealth and influence onto their already infinite hordes of it. They do this even if it rips everyone else off and fills every inch of the online environments that people like you and I have built with our effort, art, and care with an unending torrent of liquid shit, just as long as it keeps their fun and the numbers going up for a few more months. And don't even get me started on the AGI dweebs. That was a bit of a downer ending, but I'd probably be remiss not to point out that the future's not necessarily all doom and gloom. Counter to your Vimeos and your Pinterests, online art platforms like Cara do exist and are growing, which take firm stances against efforts to give artists tools to prevent new pieces from being scraped to train with. Nightshade and Glaze are the premier examples. At present, these tools are limited in scope, their effectiveness varies, and they're obviously not retroactive solutions to the plagiarism that's happened already, but they probably do warrant a mention all the same. Major American unions in the entertainment industry have all somewhat recently renegotiated their contracts, in part to implement anti-AI safeguards.

None of these are silver bullets, but they might not need to be. Resisting for a while, those bubbles burst themselves, and executives across the world find another next big thing to focus on—a different would-be exploitation device that may well be the answer here. But yeah, that's the video. Thanks for watching! I'd also like to be clear that while I only directly quoted him two or three times, the direction I'm coming from, particularly the business end of AI, in this video is very informed by the work of Ed Zitron. So if you want to go down that rabbit hole, check out his newsletter and his podcast, "Better Offline." If you've watched my "Slop" video and now this one, and you're wishing there were some video bridging that gap more explicitly—talking about AI as slop? Great news!

There is. It just isn't one of my videos. Go and watch this piece by Hoots if you haven't already. As always, you can check out my Patreon if you want access to cool perks and a way to support more videos like this. I'm going to roll credits because I'm trying to enter my "make fewer, longer, better videos" era now, and this whole mess really got in the way of that.

A big thanks to Hanan, Daniel Goldhorn, Thomas R., Weirdy Beardy, and everyone else who supports me. Thanks also to the animation professional who provided me with a wealth of thoughts and insights on this script, far more than I managed to incorporate in the end. Hopefully, I can bring some of that into my future work. And a shout-out to anyone watching this who supports any creator on Patreon, YouTube, Ko-fi, or similar platforms. You all rock, and we are all very thankful!

Later. Music. We'll see you next time.