Caution, doors are closing. Next station - Matrix

Yes, neural networks are already quite overwhelming; they are being talked about from every corner. Now AI must be part of any startup; otherwise, what kind of startup is it? And I wouldn’t be wrong to point out that not a day goes by without an article on tekkix about neural networks – how to use them, which one writes code better, who uses it, and so on. But the discussion will not even be about them or programming, but rather about neural networks for creating visual content and the speed at which they are evolving.

In just three to four years, we have gone from the first attempts to generate something realistic with Midjourney to generating videos indistinguishable from real ones. Yes, just three years, and all social media is already flooded with videos completely generated by neural networks. And the saddest thing is that now it's unclear whether this talking head is real or just a digital avatar. Is the voice real or not? Does the location in the frame actually exist in reality? It’s already scary to admire kittens on social media because it’s hard to tell how real they are. With each new neural network, Will Smith eating pasta looks increasingly reasonable and resembles more of a pasta advertisement than a rehab ad.

And let's not even talk about music; a lot of tracks are completely generated through neural networks and make it into charts, recommendations, and go viral on social media. Two years ago, two guys released a neuro-rock cover of Asti performed by Hetfield from Metallica, which was just an informational bomb, and it took them quite a while to create such a full-fledged track.

Millions of views/listens, journalists were conducting interviews, many people came to their community in anticipation of the next remixes. And now, in just a couple of minutes and prompts (conditionally), you can have either a complete ditty about the adventures of a mermaid or Billie Eilish singing "Beautiful Far Away," and it’s no longer surprising. Incredible rapid progress, and at the same incredible speed, all of this is seeping into our everyday lives. Well, what can I say, the first neuro-info gypsies have emerged, selling courses on how to create an avatar and how to earn millions. Yes, yes, there are people with 5-20k followers, and from all their videos, there won’t even be a million views in total, but they will happily sell you a course/guide/intensive based on information that is publicly available on the internet. And they will spin tales about how successful they are, how their digital avatars bring in several hundred thousand rubles a month, and they only spend 15 minutes a day – such geniuses with knowledge inaccessible to mere mortals. Well, what can you do, a sucker is not a mammoth, as they say.

Before neural networks, there was so little that was real; Instagram and TikTok were flooded with pseudo-millionaires, pseudo-successful people, and pseudo-happy individuals. Yes, one needed to keep in mind that whatever you saw on the internet, you could safely divide by 10 and not be mistaken, but there was still room for sincerity – in problems, worries, complexes, achievements, etc. And now, there will be even less room for that, if there will be any at all – five minutes in a neural network, and there you have a realistic former firefighter without an arm, or a girl with burns on half her face, or a dwarf speaking in a bass voice – anything, as long as imagination holds out to attract attention, capture it, and monetize it. And this is already being mass-produced, so it’s just a matter of time before there is so much of it that people with their real, albeit not so "exciting" fates, but real ones, will simply fade into the background. And in fact, every release of such tools as NanoBanana, Kling, etc., is like another station that we rush past on a train at great speed, and there’s no turning back. We’re all moving together to the final station, to the matrix. To some environment where there will either be no boundary between reality and fiction, or both will be very thin, barely noticeable. Therefore, I want to slow down this locomotive a little before all this madness goes too far.

And I wouldn't be surprised if there are plenty of people who consciously choose to ignore this boundary, just to get their dose of dopamine – to be president for five minutes, to score the decisive goal in the Champions League final, to spend an evening with Miss World: anything can be generated in a couple of minutes, you just need to subscribe. Events that are impossible for 99.9% of people will become available – just ask the neural network. Yes, they are not real, but who hasn't imagined themselves in someone else's role or image in front of a mirror? And here, you won't even need to imagine; you'll see everything you want, and everything will be exactly as you want it – how can one resist that?

But for me, I'm spoiling the events a bit. Right now, the average citizen is delighted by generated kittens, empathizes with nonexistent people and their fictional destinies, and wonders if people are really jumping into snowdrifts from the roof in Kamchatka. Yes, for now, there are still artifacts in neuro-video; it's still hard to just write a prompt and get a good result, even making something for 10 seconds is difficult – more details are needed, several attempts are required, special effects are necessary to hide the unnaturalness of the generated character. But this is now and for the time being; two years ago, it was impossible to generate a running person no matter how you asked, and now you can specify how they run, what they are wearing, how the camera pans, how the hair flows, how the light is set, etc. – and here you have a shot that looks like a full-fledged commercial with a budget of a couple of million. We are still far from this final station, but the train is just picking up speed and has no brakes. And in all this neural chaos, I am overwhelmed by two opposing feelings.

On one hand, I feel a sense of Spanish shame about everything happening. From this ton of fake videos, from these sometimes unnatural avatars, from people who do not understand such obvious generations. It's particularly amusing to see bloggers selling masterclasses on using a few prompts to work in two neural networks, most likely through Higgsfield. And of course, there are promises of golden mountains and successful success, because they succeeded – take their word for it. On the other hand, I have a sense of déjà vu, as I reacted similarly to Bitcoin when it first appeared, thinking who needs it, well, someone is mining it, so what? I also reacted to videos where people on TikTok danced viral dances or repeated viral trends, streamed games, or filmed pranks and challenges. I had the same feeling about recommendations to buy land or invest in real estate, even though mortgage rates were at 12% – and showed price growth and current events, those who bought and invested probably won more than they lost. And now I wonder, could this be a new blue ocean? While some consume such content, others create, mastering new models, prompts, techniques, etc.

And already now, businesses prefer to play neuro-tracks in their premises, since no one has to pay royalties and no one will punish you for copyright. In 5-10 years, will there be parties with completely generated audio tracks?

Right now, many cards on marketplaces are generated by neural networks, both the descriptions and the visual part – photos, videos. And obviously, someone was paid for this. Illustrations in neural networks for articles and books seem to have always been with us, but what lies ahead in a couple of years? Despite all this rejection of neuro content, what’s happening sparks immense interest in me as an IT person. Can this be automated? I have a home server, n8n, Openrouter – can I whip something up from that? What if I prepare prompts in one model regularly through an API, pass them to another, and constantly publish new content on schedule? And what if I create not one avatar, but, say, seven?

Or come up with a concept for a neuro-series, the continuation of which will be determined by the most popular comments under the last part? For example, perhaps it makes sense to move closer to the driver on the train, rather than trailing at the back?

And all of this is incredibly frightening. Everything and all at once. How easily one can create an avatar, come up with a story, generate content out of nothing. And, most likely, it will look just like the real thing. The authors don’t need to know anything for this, not even a decent camera or at least an iPhone, nor how to set up lighting in the location. Just $20 a month for a subscription and voilà, you are a blogger in 24/7 mode. And there are those who will pay for such content and those who will consume it. And all of this happened in just a few years, literally right before our eyes. Judging by the increased demand for memory and the skyrocketing prices for it, we are just starting to pick up speed.