I played with Midjourney last weekend. Not because I agree with their ethics, or have changed my mind, but because it's inevitable and I need to understand it. The experience, at least during trial, is pretty terrible: you're stuck in a Discord chat room full of other newbies, everyone furiously typing prompt after fevered prompt, the screen filling up with their generated content faster than you can actually see anything. It's like hanging out at a party where everyone's on speed and talking at the same time. But it IS uncanny.

I decided to test the platform after reading this article about Midjourney v5, staring slack jawed at its inroads into full-on photographic territory. Some of it is sterile and stunted, but that's changing and it's already perfectly suitable for certain markets. I didn't have access to v5 on a trial account, but it's unsettling to witness first hand what can be achieved with a few choice keywords, even on v4. The limitations are obvious, but so's the potential, despite a certain lingering emptiness—because you're participating, but barely, leaving very little sense of fulfillment... for now.

I can imagine a visual generation engine that would learn from our images, that we'd train and control completely. There are already apps out there doing this to speed-up processing. If I can create a subject, place it in any location, real or imagined, choose the image format, the camera model, the lens, f-stop and shutter speed; if I can place one or more light sources and select a POV; if I can then virtually move within this scene and create a set of images that I then send to Lightroom to edit...

It won't be photography as we know it; it won't be a fleeting moment captured midstream; but then, neither is a painting.

Talk about full-circle irony.

The truth is, we're f#&ed. Photography won’t disappear, but many of its current applications will. Another article showed off a $17 headshot service powered by AI: you send a couple of pics and get back 100 "professional-looking" headshots, all in different settings and styles. And you know what? We can mock all we want, but this is horses vs cars. Because those types of services are in their infancy, they're not even toddlers yet, and the growth will be exponential. The scale here isn't years, it's weeks and months.

We're not ready for any of this, not ready for the speed at which AI is being developed with very few guardrails, and especially not ready for the push behind it, as everyone and their neighbours race to add an AI bullet point to their marketing material. I just hope it eventually grows into more than a way to cut corners, or steal other people's vision.

And that's the 100K question: can a proper generative AI exist without the entire Internet feeding it the work of others? Or is infringement the price of power?

One last point, which is sort of a follow up. I often see the argument made about all art being derivative and every artist stealing in some shape or form, to defend AI training. It’s true: none of us is an original, born with a magical and unique gift no one else can match. We all copy at some point.

But when we do it’s usually out of awe, because we’re inspired by an artist’s work and wish to learn by internalizing it. It comes from a place of recognition and alignment with someone else’s vision. And while we may succeed at replicating a style, we’re still interpreters: who we are will contaminate the output over time, and something new will emerge. Artists steal, but only through sweat and a genuine quest for knowledge.

AIs just eat the world to spit it out.