AI: Hype, Hallucinations, Smoke and Mirrors

The debate over AI is devolving into a circus

EDIT: I’ve rewritten this article. I don’t think I make my point clear the first time around.
EDIT: After finishing this rewrite, I still think there’s so much to say I should just keep writing until I can deal with the other topics.

The chatter about Google’s Gemini woes reached a fever pitch before settling down a bit. No lasting solutions, though. We didn’t get to any conclusions as to how diversity vs History should be resolved in the long run — from “diverse” Popes to non-Nazi Nazis, we’re bumping into problems that occur when images are created by statistics-driven generative systems. By their nature, they adhere to universal principles.

I arrive on the scene a bit late, but hopefully still in time. The overall narrative and the different voices are a cacophony now and I am still contemplating how my voice can rise above that ambient noise, not to become just another ghostly echo.

Fanfare of pundits. As usual, pundits look like they're about to pick up a fight. (Image by the author on custom SDXL system.)

The AI dialogue is a jam-packed arena, with everyone from Elon “Muskarade” to Sam “Virtual Visionary” and a host of CEOs turned full-time commentators — a veritable legion of pundits — all seeking attention with their relentless blogging, newsletters, and media stints.

Do they actually work, or has the C-suite morphed into a glorified PR circus?

AI has become an exasperating sideshow swarming with jesters, each claiming they know best, they serve the best flavor of the AI utopia and they deserve the spotlight. There is religious fervor in their concerns “for the development of mankind”, usually related to the fact that their company’s valuation should be the highest one.

Town becomes a circus and all the clowns make it hard to reason. (Image by the author who is adamant about NOT ever using Unsplash)

Generative AI Can’t Get It Right, But What’s So Wrong With Martian Popes?

AI as a technology sill interests me, as there is a lot of potential and a lot of problems to tackle, like Diversity Popes. But it seems to me not enough has been said about the quirks and kinks of AI horses and AI Cadillacs.

If you’ve ever worked with these systems, if you’re counting your images in the tens of thousands as I am, you know they’re unpredictable. Infuriating. Exasperating. Aggravating. Anything but “tedious”, as they always surprise you when you are in a rush.

Conceptually, is this level of absurdity worse than The Woken Popes? (Supposed Cadillac supposedly driven by a blonde was another attempt by the author to get an image that bears no relation whatsoever with *this*.)

Getting what you need can be a maze of complexity

Trying to get what you need from AI can feel like solving a puzzle without it’s picture. Take ComfyUI, for instance— it’s basically Stable Diffusion with a huge twist, packed with hundreds of modules and a sea of parameters. It’s a beast to master, and yet it’s still a roll of the dice whether you’ll get the outcome you’re aiming for.

Gemini, Midjourney, DallE, all the different Stable Diffusion solutions on the web (like Leonardo AI, for example) are a mishmash of randomness and probability — what marketing people like to call “AI.” Whether it’s a diversified Pope, a Nazi mermaid or a Cadillac with a ghostly chauffeur, they’re all just as likely (or unlikely) to pop up.

When I end up with a jazz quartet where three musicians are playing tenor sax, or that ghost-driven Cadillac, isn’t that about the same as quirky Popes and outlandish Nazis?

I get it, with all the Diversity Wars going on and all the confused actors rewriting History in so many ways at the same time, we’re all sensitive to some Q-Plan regarding AI Popes.

But other than that, it’s all part of the same unpredictable AI game: more often than not, it gets things wrong. When it does something cool, THAT is showcased, and suddenly everyone’s led to believe — courtesy of the Marketing & Media machine — that AI can follow instructions to a T.

Here. I like this image. The legend explains how simple it was to create it.

That image might be part of a Noir game that the author might or might not be involved with. The author does know it was generated in Midjourney, detailed and enhanced in Topaz AI, then sent to Photoshop, extended to fit in the 16:9 format, then I remember a bit of a fight with Adobe’s random “I’ll extend this image for you” generative thing, then there were a few regenerations on the sides, then some smaller parts were re-re-re-regenerated, then the whole thing was sent to DXO Silver FX to get the right contrast and mood and BW vibe and some grain, than back into Photoshop to be properly saved and then into RIOT to be compressed. But, YES, it’s REALLY simple & easy to get great results with “AI”. My EYE!

As some of you know, I am a photographer (and a philosopher and soon, hopefully, a game creator). Pulling off a workflow like this in 15 minutes is part of my routine. But this image, like many others I’m working with, has almost “zero AI” to it — all I got there was in my brain way before I even decided to go through the motions.

The unseen images that quietly fade away in the digital graveyard of my machine aren’t mourned. Yet when someone appreciates an image, suddenly it’s “AI is flawless, it’s stealing from photographers!” Maybe. Maybe it’s showing us that Unsplash was a real bad move. That’s another debate entirely. Right now, for me, AI is a tool that gets one of my jobs done and frees me up to explore visual storytelling in gaming — something that would be cost-prohibitive otherwise.

Generative AI is a tool

Generative AI is reliable as a tool, sure. It’s more dynamic than a spellchecker and makes tools like Photoshop seem more approachable, but it’s not the ultimate solution.

Referring to it as ‘artificial intelligence’ might be overstating things — it’s machine learning at best. There’s no real “intelligence” there, which would suggest understanding, purpose, and context.

We’ve got intricate pattern detectors and number crunchers, not conscious computers. The industry runs on excitement, investors are blinded by their greed, and the media is churning out more fiction than fact, losing sight of the real picture.

A big issue is that many lack the technical insight to grasp what’s happening, which is, in my opinion, the core problem. We’ve all become part of the AI Media & Marketing Bros. Circus, where illusion is the name of the game — not the firmest foundation for meaningful discussion.

As the talk around AI has turned into a media frenzy, where the wilder the claim, the more attention it gets, we’re missing thoughtful conversations about AI’s mechanics, its limits, and the ethical considerations of its use.

My worry grows. This isn’t just about AI’s inner workings; it’s about how we as a society use this extension of our intellect and creativity. It’s time to reduce the fanfare, send Musk off to Mars with Altman and Branson (no communication devices!!), and consider AI with a healthy dose of skepticism and open-mindedness.

Let’s think about what’s really at risk. Do we need to keep damaging our planet to create more electronics by extracting precious elements in order to generate more heat? Why aren’t we discussing the impact on jobs and governance? Why do we pay attention to all this at a time when the world should be focusing on survival instead of AI?

You want it darker? (song by Leonard Cohen, image by my own fears and imagination)

I’ll be diving deeper into this as time allows—a challenging week lies ahead. But consider this a prompt for reflection, an invitation to look beyond the spectacle and ponder the real implications of AI, stripped of the hype and hyperbole.

Thanks for sticking around!