Sora: End of creativity part 2

It costs about $1.9M to make a single minute of a Pixar animation.
That’s $31,000 per second. Many traditional movies exceed even that.

Let’s talk about OpenAI’s Sora

Just a short while ago Sam Altman said he needs about 7 TRILLION dollars to push AI truly forward. The main needs are in building AI-specific microchips and thermonuclear power plants to run the models of the future.

Raising that much money is definitely going to be difficult, but to the surprise of almost everyone one week later they showcased SORA — an AI prompt-to-video generator.

Text to video isn’t all that new. Models like that have existed in the past, as well as models that were able to animate static images and photos.

Good movies can cost almost $2M per minute to make

What’s remarkable here is the quality, visual fidelity, and the fact that the videos can be up to a minute in length. In the example above, if you could generate one minute like that and still have a good story and voice acting you’d be saving almost $2M.

The split

The obvious use cases are of course non-professional videos. YouTuber B-roll materials and educational content that may just need a voiceover — likely done by some other AI from a script generated by chat GPT.

All parts of the production can be automated this way

Of course, you can guide the AIs or even manually modify and adjust things but let’s be honest here. Most people won’t do that and simply spit out massive amounts of completely automatic content.

This will result in a couple of things.

1. There will be a lot more content to choose from

2. It will be much harder to find quality content

3. We will be overloaded with choice
   Take a look at the screen below.

Modern streaming services are famous for their impossibly large collections of content. It’s actually so much stuff, that it often takes an hour to just decide what to watch.
This is cognitively tiring and it’s the same with YouTube and other video platforms.

When we drop a lot of lower quality, SEO-article-like AI-generated content into the mix this will only increase the fatigue and decision paralysis.
It can have the opposite effect with people simply deciding to watch less YouTube instead of more.

The SEO trap

I often bring back the infamous “SEO article” as an example. It’s a special kind of content written for search engine bots and not humans. Their sole purpose is to be packed with keywords so tightly that they’ll appear in the search results above other, more real content.

And while both Google and Microsoft claim they’re fighting these you often end up on those sites when looking for something anyway.

At first glance, you’ll able to spot that this is not what you wanted and click away.

I wrote an article about the internet split recently that fits well with what we’re discussing here today.

The tale of two internets

With robust, almost magical tools like SORA, the split of the internet will be even larger. Now it will not only include written articles, but also video content.

AI-generated content can of course be good, but the main reason people use those AI tools is because they want to save time, not because of higher quality.

So while a small minority can still deliver an amazing experience with these tools, the overwhelming majority (I’d say 99.9%) will just deliver the visual equivalent of spam.

Human-generated videos will gain more traction often because of their imperfections and will be prioritised in our choices due to them being made by someone we can empathise with.

The honesty of low production value will often trump the blandness of high production. Take a look at Zuck’s latest Vision Pro Review. It was recorded on a couch, with bad lighting and some potato camera. He just talks about stuff, as naturally as he can and that’s it.

High production by humans

Then the highest regard will go to highly produced videos by humans. I feel MKBHD will be safe even if others try to emulate his crazy robot crane shots with AI. That’s because he’s human.

Welcome to the Matrix

Remember this scene from The Matrix? The bullet-time rotating camera thing was so jaw-dropping when I watched it in 1999 I rewatched this scene dozens of times.

I was wondering how the Wachowskis did it. It was intense and insane both at the same time.

This is when it hit me. We wouldn’t go to the Louvre to see the Mona Lisa if it was just an AI printout.

No, we want to see it because it took SIXTEEN years to paint it.

Historians discovered da Vinci applied very thin, nearly transparent layers of oil paint with his fingers over many months to slowly build up the glowing, softly focused image of Mona Lisa. In fact, he would apply 20 to as many as 40 layers of paint.

We go to see it because we are in awe of true mastery. That’s also why we watched The Matrix. It was new frontiers in visual storytelling — made by humans, for humans. Even if in a VR simulation.

The future of video?

My prediction from The End of the Internet still stands here. It will be only slightly different with video.

The majority of the internet will be mass-generated, ultra-low-value AI content fueled by ultra-low-effort AI advertising.

The quality content will be either login-walled or pay-walled and will have a human touch all over it. In the case of using ANY AI in that content, it will have to be clearly stated and annotated so the users won’t feel cheated.

Entertainment will definitely go that way and initially AI content will take over the attention of customers. But it will quickly tire them out by sheer quantity and lower average quality.

That thought of no human process behind it will also linger somewhere at the back of our heads.

Educational content

The biggest problem I believe is not in entertainment though. It’s in education. Both video courses and free YouTube edu materials are currently the only way for many people from developing countries to learn new skills for free or very cheap.

That AI / Human divide will leave them with garbage content that overloads their brain and further limits their ability to grow. It will be like a digital wall dividing two completely different worlds.

That will lead to more world inequality and tensions.

It pains me to see this happening, but it feels like this shift has already started. And once that first AI hype clears up a little bit, those who can afford it will simply move to the better content.

The rest will be stuck with mostly garbage.

Liked the article? Share it!