I have seen the future, and AI Art is embedded in it.
Case in point, the above image was generated by typing nothing more than the words "a dachshund doberman mix with black fur and light brown and medium brown coloring riding a skateboard" on AI art service, DALL-E.
Four different variations were rendered in under one minute, I picked one version where I chose a few more variations, clicked upscale, and this is the output.
In under two minutes start to finish, I had created something magical (to me, at least).
The experience brought me back to Neal Stephenson's 'The Diamond Age,' which introduced the concept of a Matter Compiler.
Think 3D Printing, Generation 100.
The Matter Compiler of The Diamond Age was an atomic and sub-atomic assembly and render engine. Pretty much anything, organic or inorganic, live or not alive, could be output from a Matter Compiler.
If that is what the fully realized end-state looks like, then the potential of AI Art is some subset of that.
A Sample AI Art Gallery
Here is a sample gallery I created in Midjourney, another emerging AI Art service.
Using the specific phrases you see on the each page, and then creating variations, and then iterating those variations until I got the looks that I wanted, is how I came to a finished instance. To finish, I upscaled the final candidates to the max.
Sidebar: The "finished product" is akin to an MVP but for visual prototypes.
Think of the word strings on each page of the gallery as akin to how we all got good-ish at fine tuning google queries to get the pinpoint return we were looking for.
This is best thought of as a render query, and it demonstrates AI Art's potential as:
- A new "compose-able" medium
- An artistic medium for creative dabbling
- An engine for product design
- A platform for extrapolating design patterns
I have two thoughts here.
One, most disruptive innovations start as "toys" and then as they ramp up the power and utility scale, find their niche and grow to dominate.
Keep that in mind in assessing the unfinished quality of what AI Art is at, and how magical it already is, anyway.
Two, by reducing the effort required for deep exploration and prototyping from a scarce, complex activity to a simple and unlimited one, AI Art creates a fertile environment for a wave of meta artists (and technicians) to emerge.
This is similar to the way twitter turned blogging from a niche universe to a 140 character tweet that anyone could instantly create and/or consume.
Segments Ripe for Disruption
Stock photos and stock images are one such example where this type of service could be a disruptor, but how about renderings of buildings, or of master planned communities?
But while AI Art for Images is pretty damn compelling in its own right as a native experience, AI Art is not just for Images, but for Music, Games, Video and Writing, too.
One use case I can see here is AI rendered post production services to overlay digitally rendered video, imagery and sound into movie scenes.
As social media showed, there will be an ever-growing content base, and all of the creative activities pursued by users through their usage will "train" the systems to yield better output.
This will by its very nature, MEANS accelerated learning patterns, and a virtuous innovation cycle as it kicks into high gear.
AI Art has the potential to be transformational for multiple industries:
- Point of Interest (POI) libraries will have real value in how
they enable better modeling and extrapolation - As a driver of growth, the scaling of Capturing,
Rendering, and Output will be interesting to watch - Gilder’s Scarcity and Surplus Observation suggests that not only will this wave yield a lot of new value creation, but great monetary wealth
Deep Fakes or Parody: The Intellectual Property Question
A final thought. A question that both Midjourney and DALL-E are already grappling with in different ways is the question of intellectual property (IP) and use of likenesses and recognized brands, and how heavy-handed they should be with each.
At one extreme, you have deep fakes, counterfeiting and co-option of someone else's identity, brand and/or likeness in ways that invade privacy or damage one's rights to own and define that which is theirs.
At the other extreme, you have parody and satire, which is largely protected as freedom of speech and artistic expression.
Midjourney and DALL-E are in beta if interested in checking out.
Update: I LOVE this analysis by Sequoia Capital on #GenerativeAI, which they define as "A powerful new class of large language models making it possible for machines to write, code, draw and create with credible and sometimes superhuman results." This essay does a great job of codifying the different layers of the stack, and the applications they engender.
Update:Very strong analysis by Kevin Kelly (Picture Limitless Creativity at Your Fingertips) on how Artificial intelligence can now make better art than most humans, and will transform how we design just about everything.