Saturday, 27 August 2022

Change is coming: AI renderers

Midjourney rendering of a geisha woman
Midjourney Geisha. All the images in this post I 'created' with text prompts using Midjourney.

And when it arrives, it will be quick.

I've been enjoying the Light & Magic series on Disney+, about ILM and all the creative marvels who have worked there. Their enthusiasm and idealism are infectious; these are not so much documentaries as recruiting commercials. 

One thing that did strike me: when the shift was made between traditional and digital effects, which reached a tipping point with Jurassic Park, it happened fast. Very quickly large numbers of people were packing their bags and heading out the door. When the lead stop motion animator saw the first Tyrannosaur walk sequence created in 3-D software, he said to Speilberg, "I feel extinct." Spielberg loved the line so much he said he'd put it in the movie. 

This particular gentleman managed to find a berth as an animation director once his lifelong vocation became obsolete.

A stylized portrait generated by AI Midjourney
A stylized portrait

I remember taking some courses at Sheridan College back in the day, when I toyed with the idea of getting into 3-D (I'd been accepted to their animation program but elected to pursue illustration instead). The security guards there were all former master typesetters; their careers had evaporated with the introduction of digital layout programs. 

The Luddites are generally smeared as an ignorant bunch of louts who were afraid of technology. They were actually skilled weavers and textile workers who were being put out of their jobs by mechanized looms, at a time when there was no social safety net: losing your job could well mean starvation and death to someone in 1780.

An AI generated picture of a squid man in the style of Heironymous Bosch
A gluttony demon in the style of Heironymous Bosch

Change is constant, and rest assured friends, it's coming for you. We have to be prepared for the rug to be yanked out from under us by some new fangled invention.

Which, incidentally, is one reason education has to be a lifelong endeavour. 

It will also be increasingly necessary, I suspect, for higher education to be free for all adults. The US gave tax cuts worth 1.9 trillion for the rich, which is apparently the same as the size of the US student debt, according to Robert Reich. Empowering young Americans should be a greater priority than providing tax cuts for oligarchs to fuel trickle down economics that don't, and never have, worked. It's also disturbing that corporate America is now pumping more money into ephemeral stock buy backs than they are into R&D. It jazzes the stock price for shareholders, but it does squat otherwise.

An AI generated landscape in the style of Bosch
A deceptively peaceful landscape, in the style of Heironymous Bosch. No hint of the coming turmoil.

As we move more and more into a knowledge based economy, education will becritical to establish competitive advantage over other nations. We should really get manufacturing of essential items back, as well; just in time production is so lean it provides no cushion when shocks hit the system. 

And AI rendering is going to be a shock to the system. A major shock. Currently the renderers are limited in what they can do, particularly when it comes to specifics. But they will get better. We should not feel complacent just because of what they can't do today. They are going to get better and better and better. 

Four AI Generated cats in the style of Bosch
Cats in the style of Gustav Klimt. You don't have to just pick one influence, you can combine several, creating something (potentially) distinctive.

I have been playing with MidJourney, which seems to 'photobash' images. Stable Diffusion, however, seems to have some 3-D renderer sensibility behind it. Things Midjourney can't do, Stable Diffusion can. Check out this article on it here. I haven't gotten my grubby little hands on it yet, but I intend to. 

And there are more renderers coming out.

From the linked article on Stable Diffusion

"Artists and other creative professionals are raising concerns and not without reason. Many will lose their jobs, unable to compete with the new apps. Companies like OpenAI, Midjourney, and Stability.ai, although superpowered by the work of many creative workers, haven’t retributed them in any way. And AI users are standing on their shoulders, but without asking for permission first."

Where will they be in ten years? Hell, where will they be in one year? 

Midjourney has only been out a few short months. I believe Stable Diffusion has just been opened to the public in the last day or two.

This is new. 

This is a huge moment in the development of visual arts technology, and we're all here to witness and play with it. It's an historic moment, only unlike with the fall of the Berlin Wall, you can sit at home in your pajamas and participate with a hot cup of cocoa.

I have pumped out an insane number of images using Midjourney over the course of two weeks, more than I could have done if I dedicated myself to painting full time for a year (or more), and while there are defects in the renders, they are also in many ways highly sophisticated; it would take a great deal of effort to match them, technically, when at their best. 

An abstract Midjourney generated cat
An abstracted cat. I think. 

When the tipping point is finally reached, AI Renderers will revolutionize commercial visual arts. They may upend fine art, too, while they're at it. Image making will become property of the masses; those with great skill may still be needed to refine things, but as AI gets better, their numbers will dwindle.

The real question to my mind is whether or not there are limits on what the AI can do. Can developers solve the issues they face now, or are some of them insurmountable? If so, that's good news for artists.

I can see AI 'script bashing' movie blockbusters, too; if not today, then soon. They're mostly written for an international audience, so they're light on dialogue and heavy on effects and action sequences. It wouldn't surprise me if AI could take 1000 action movie scripts and hodge podge together something indistinguishable from a Michael Bay film. 

A stormy scene of a radar tower under clouds
A Soviet monument, in a storm

Aaron Sorkin may be safe, but for how long?

Skill with actual physical materials will leave some artists safe for the time being, but sooner or later there will be robots with paint brushes.

What room is there for the human creative spark? 

The Empire State building made of pink feathers, generated by Midjourney
A featherly Empire State Building

Years ago I read about a computer program that came up with ideas for advertising posters by juxtaposing elements, such as making the dome of a stadium a basketball. 

No doubt it can do better now. 

I've greatly enjoyed playing with Midjourney. I am somewhat relieved it is limited, but I can see incredible potential, too. 

It will be a few years (maybe a few decades) yet, but it would not surprise me if seismic changes occur soon enough for me to witness live and in person.

I hope they aren't planning to replace the security guards at Sheridan with robots any time soon...

With any luck, I'll have retired before the AI Renderers take over. And I for one welcome our new robot overlords...

I made these images with tools from @midjourney, you can sign up for their private beta here http://bit.ly/3J2NNVs

An Oni demon mask generated by Midjourney
Ghost Oni in the machine

Painting of an emo artist walking contemplatively along the shore as a storm closes in on him; in the distance looms a great AI rendered castle.

AI generated image of two people fighting in an office
An illustration of two people fighting in an office, with an Art Deco feel; this is what Midjourney delivered back. 


No comments:

Post a Comment