I have fallen into a rabbit hole: the world of AI generated art.
There are two major platforms for AI art generation:
- DALL-E2
- Midjourney
Both are free, and both create output that boggles the mind.
DALL-E2
![dall-e](https://6104926.fs1.hubspotusercontent-na1.net/hub/6104926/hubfs/dall-e.webp?width=1120&height=781&name=dall-e.webp)
Let's start with DALL-E2, which is built by OpenAI - the same organization that builds ChatGPT, and the login/registration process is identical. Put in your email and phone (you can use your Google/GMail account), and you are off and running.
Once you are logged in, the screen looks very different:
![dall-e-screen1](https://6104926.fs1.hubspotusercontent-na1.net/hub/6104926/hubfs/dall-e-screen1.webp?width=1120&height=890&name=dall-e-screen1.webp)
To trigger the AI to create art, you simply give it a prompt. Like this:
"A photo of Michelangelo's sculpture of David wearing headphones djing"
Output:
![michaelangelo-wearing-headphones](https://6104926.fs1.hubspotusercontent-na1.net/hub/6104926/hubfs/michaelangelo-wearing-headphones.webp?width=946&height=940&name=michaelangelo-wearing-headphones.webp)
...or...
"A photo of a teddy bear on a skateboard in Times Square"
Output:
![teddybear-skateboarding](https://6104926.fs1.hubspotusercontent-na1.net/hub/6104926/hubfs/teddybear-skateboarding.webp?width=940&height=936&name=teddybear-skateboarding.webp)
The output of the platform can be refined, and comes in sets of four, like this.
Prompt:
"A large medieval arabic city with docks and boats on a lake, digital art."
Output:
![dale-prompt-output](https://6104926.fs1.hubspotusercontent-na1.net/hub/6104926/hubfs/dale-prompt-output.webp?width=1120&height=892&name=dale-prompt-output.webp)
In my experience, I found that DALL-E created better output based on the number of examples that exist in the real world. For example, is does a great job creating renaissance oil portraits.
DALL-E2 is free for up to 15 prompts per month, then it is ~$0.13/credit, sold in blocks of 115 for $15.
Midjourney
![midjourney-landing](https://6104926.fs1.hubspotusercontent-na1.net/hub/6104926/hubfs/midjourney-landing.webp?width=1120&height=780&name=midjourney-landing.webp)
Things are going to get a little weird.
Midjourney is an AI art generation platform that describes itself this way:
"Midjourney an independent research lab exploring new mediums of thought and expanding the imaginative powers of the human species.
We are a small self-funded team focused on design, human infrastructure, and AI. We have 11 full-time staff and an incredible set of advisors."
The project is lead by David Holz, who was previously the Founder of Leap Motion, and a researcher at NASA. To say that David is an interesting dude, is a gross understatement.
The way Midjourney works, is very different from DALL-E2. The art generated comes fast and furious and is always conducted on Discord, a private-server social media platform. To register, you must have a Discord account (probably to keep the uncool people out) and once connected there's not much to explain to you what the heck is going on. It looks a little like this:
![midjourney-on-discord](https://6104926.fs1.hubspotusercontent-na1.net/hub/6104926/hubfs/midjourney-on-discord.webp?width=1120&height=924&name=midjourney-on-discord.webp)
From here, I had to do a little Googling: "how do I use midjourney? If you are lucky, you find your way to the Midjourney Quickstart guide which reads a little like this:
- Join the Discord
- Find a Newbies channel
- Use the /imagine command (this is where to tell it what you want)
- Processes the Job (aka, wait around about a minute)
- Upscale or Create Variations (push one of 8 buttons that show up)
- Rate Images
- Save your image
- Subscribe to a Plan (you get 25 free jobs, then you have to pay)
Remember: all of this is done with at least a few dozen people in the same Discord channel. So what do it look like? A little like this:
![discord-midjourney](https://6104926.fs1.hubspotusercontent-na1.net/hub/6104926/hubfs/discord-midjourney.webp?width=600&height=495&name=discord-midjourney.webp)
Let's look at some art (pulling a few random samples of what people were asking for).
/imagine thermonuclear godzilla as a fairy
![thermonuclear godzilla as a fairy](https://6104926.fs1.hubspotusercontent-na1.net/hub/6104926/hubfs/thermonuclear%20godzilla%20as%20a%20fairy.webp?width=1120&height=1118&name=thermonuclear%20godzilla%20as%20a%20fairy.webp)
/imagine dog with pirate hat
![dog with pirate hat](https://6104926.fs1.hubspotusercontent-na1.net/hub/6104926/hubfs/dog%20with%20pirate%20hat.webp?width=1120&height=1121&name=dog%20with%20pirate%20hat.webp)
/image Create a Athletic Clothing brand using earth tone colors and is minimalistic
![an Athletic Clothing brand using earth tone colors and is minimalistic](https://6104926.fs1.hubspotusercontent-na1.net/hub/6104926/hubfs/an%20Athletic%20Clothing%20brand%20using%20earth%20tone%20colors%20and%20is%20minimalistic.webp?width=1120&height=1120&name=an%20Athletic%20Clothing%20brand%20using%20earth%20tone%20colors%20and%20is%20minimalistic.webp)
All of these images are completely AI generated, with no human direction, except the text being input. I'd like to let that sink in.
It never ends
I spent hours on these platforms, and I just only scratched the surface. To quote Alice (from Alice in Wonderland):
“Down, down, down. Would the fall never come to an end! `I wonder how many miles I've fallen by this time?'”
Thinking on it further, to say I've "scratched" the surface misses the point of these platforms: they are generating content as fast as millions of humans can type. As of September, 2022 there were 1 million users of DALL-E2; in the same month, Midjourney reported 2 million users.
While I am a person who likes to contemplate the future, the implications of this technology elude me. I know this will be big, and I know that the world changed in 2023, but I have no idea where this will take us. If you do or if you have any thoughts as usual, I'd love to hear from you.