AI Remixes. Humans Push Things Forward

Recently I was sitting in Cabo airport on the way back from a company offsite. My flight to Dallas wasn't until 2, so I sat with Caleb Peffer, the CEO of Firecrawl where I work, and a few others while he waited for his earlier flight to San Francisco. We got talking about how I got into making videos, and Caleb asked me how I saw the future of content creation in the age of AI.

I wasn't expecting the question, and it wasn't something I'd ever really thought about, so my answer wasn't very good. This post is the answer I came to after having more time to think about it.

The short answer? I don't really know.

AI moves so quickly that predicting anything feels pointless. There's no way I could have guessed that Opus 5.5 would be such a massive leap over Opus 5 in such a short amount of time. It's funny to think that Opus 4.8 only came out this year. LLMs have progressed so fast at coding, at building 3D worlds, even at making full 3D games, that I genuinely can't tell you where AI video will be a year from now.

I know, I've just said I can't predict anything, but there is one thing I feel sure of. There will always be a need for a human. Humans are the creative ones. Humans are the ones who push creative mediums forward.

AI games aren't new games#

Take a look at the computer games people have made with AI. Someone writes a simple prompt, maybe lets it loop for days or weeks, and out comes a game. The results are impressive, but they're games that already exist. They were made very quickly, sure, but there's nothing new in them.

When I think about the games I've played and the things I've watched that genuinely felt new, they were all handcrafted by a human. Someone who looked at everything that came before and deliberately tried to do something different. I haven't seen AI do that yet.

To be fair, it's not exactly a fair fight. A one-line prompt on a loop versus a studio spending years on a game. But that's kind of my point. Put a human with real intent behind the AI and things get interesting. Take the human away and you get a very fast copy of something you've already played.

Predictable stories are easy to generate#

I've watched a lot of films, and with many of them you can guess what you're getting from the title, the opening scene, or who made it.

I watched Fantastic Four recently and could have told you the plot before it started. They're superheroes, there's a baddie, they defeat the baddie, they win. That's fine, but if a story follows a pattern that common, AI can generate it too.

Do something different, though, and it becomes much harder for AI to predict.

Poor Things, and not knowing where it's going#

Poor Things, starring Emma Stone and Mark Ruffalo (yes, the Hulk), is one of the best films I've seen this year. Maybe one of the best I've seen in my life.

What made it so good was that I had no idea where it was going, the entire time. I won't spoil it, but here's the premise. A scientist finds a pregnant woman who has just taken her own life. He brings her back, but swaps her brain with the brain of the baby she was carrying. So you have a grown woman with the mind of an infant.

From there the story took so many twists and turns I didn't see coming that I was desperate to find out what happened next. I'm not saying AI could never make a film like that, but it would take a lot of human creativity to get there.

Death Stranding is gloriously weird#

The same goes for games. I haven't played many in the last few years, but one of the best I've played in the past decade is Death Stranding, for exactly the same reason. It's almost impossible to predict.

It's essentially a walking simulator. You play as a postman in a post-apocalyptic world, delivering packages from one side of the US to the other. Simple enough, until you hear the rest:

• There are enemies that humans can't see with the naked eye. They're black, humanoid figures that have umbilical cords of their own.
• To sense them, you carry a newborn baby in a pod of fluid, umbilical cord and all, strapped to your chest. It's hooked up to a device that flashes when they're near, and even then you never see them fully.
• There are human enemies too, but you're not supposed to kill them. You can shoot them with real bullets, but if they die, their bodies cause a massive explosion.

And all of this is in a game where your job is delivering parcels.

There's no amount of time I could spend prompting an AI to make a sci-fi game that would land on something with that level of creativity and whimsy. Maybe I'm just personally drawn to weird things like this, but I think that's exactly the point. Nobody would ask for it, and nobody would predict it, until a human made it.

The same is true for software#

I think this applies to software as well. Most of the software being built with AI is stuff that's already been done. People are being creative, making tweaks and remixing existing ideas, but I haven't seen anything built by AI that pushes software development forward. A lot of it is being made to make money quickly, which is fine, but it isn't new.

I'm no Hideo Kojima, but if I did want to push software forward, here's what popped into my head while thinking about it:

• Keep the camera always on on my Mac, tracking my eye movements to work out what I'm actually trying to do.
• Keep the microphone always on, so I can talk to the computer about what I'm editing.
• Track my fingers so I can move things between screens and navigate without touching anything.

Basically, ways of using a computer beyond the mouse and keyboard. Now, I'll be honest, none of these are new. The Apple Vision Pro already uses your eyes and a pinch of your fingers, and Macs have had Voice Control for years. They're just the first examples that came to mind, and if you asked an AI to brainstorm, it would probably suggest the same ones.

What I don't think it would do is pick one strange idea and commit to it far past the point anyone thought sensible, the way Kojima did with a game about delivering parcels. You can't prompt an AI with 'you're a video editor, push software forward' and expect that kind of stubbornness. That part is still ours.

Humans push things forward#

To be fair to AI, neither of my examples came out of nowhere. Poor Things is based on Alasdair Gray's 1992 novel, which itself riffs on Frankenstein. Kojima wears his influences on his sleeve. Humans remix too.

The difference is what we do after the remix, and that comes down to taste and intent. A human chooses which weird idea is worth years of their life, brings their own experiences to it, and makes something nobody asked for. AI is trained on things humans have already made. It's incredible at reproducing them, quickly and cheaply. But moving a genre forward means making that bet on something nobody has done before, and so far that's been a human job.

So if Caleb asked me where content creation is going today, I still couldn't give a confident prediction. What I would tell him is that, for as far ahead as I can see, there'll be a case for humans when it comes to creating things, because humans are the ones who push things forward, not machines trained on what we've already done.

For most of us making everyday content, I think that means AI will eat the predictable middle: the tutorials, videos, and posts that look like every other one. What's left, and what will be worth more, is the taste to make something people don't see coming.

That's my take, but I'd really like to know what you think. Do you agree, or do you think I've got this completely wrong? Especially if you've seen something made by AI that genuinely feels new. Let me know on Twitter/X.