This past week, I listened to two Twitter Spaces that discussed AI-generated art. One was focused on gaming with Jonah Blake on Onchain TV and the other focused on art with Claire Silver on Rug Radio.
So what the hell is AI-generated art? What else is being AI-generated these days? Letās talk about DALL-E.
What is DALL-E?
What does it stand for? I actually have no idea haha.
Fun fact - Years ago, I was on a flight and decided to watch WALL-E since I heard great things. It was in French which I thought was odd, but it was a Disney Pixar movie and the animation was great. As confusing as it was, I was able to piece the French movie together. After about an hour, I finally turned to my friend and asked him why WALL-E was a French movie. The look he gave me told me all I needed to know (for some reason the movie had the French setting turned on).
Moral of the story - TPan is stupid. Consider unsubscribing.
Alright, time to redeem myself. Quick facts on DALL-E:
DALL-E is a version of GPT-3 trained to generate images from text descriptions using a dataset of text-image pairs. They have 12 billion parameters (Iām assuming this means data points) that have trained this model.
A text-image pair
DALL-E and GPT-3 are products by OpenAI, an AI research company focusing on AI applications that benefit humanity.
Employees include
Greg Brockman (ex-CTO of Stripe)
Ilya Sutskever (ex-Research Scientist at Google)
Sam Altman (ex-President of Y Combinator)
Non-employees include
Adam DāAngelo (CEO of Quora)
Reid Hoffman (Partner at Greylock, Co-founder LinkedIn)
Will Hurd (ex-US House of Reps)
Tasha McCauley (CEO of GeoSim Systems)
Helen Toner (Director of Strategy/Research Grants at Georgetownās Center for Security and Emerging Tech)
Aka: Team is stacked.
Some examples of DALL-E prompts and outputs:
š¤Æš¤Æš¤Æ
Have you seen those memes that look like this?
Yup, thatās the work of DALL-E. This Twitter account has a lot of weird and funny prompts and the resulting outputs. Iām sure there are more on each social platform.
BTW, the waitlist to access DALL-E is 1 million people long. You can join it here.
If you canāt wait, a couple of DALL-E alternatives:
Midjourney (25 queries for free, need to join Discord server first, 600k+ users, wow!)
I gave Prose Painter a shot out of curiosity and this is what I got, pretty creepy and pretty cool.
Lolā¦For some more context, the default query and image looks like this.
That combination of words and image made me think of the KPop supergroup BTS (they have a popular song called āDynamiteā), and I was hungry, hence āKorean bulgogiā. Add in the autocorrect, and the rest is history š.
Have you listened to a BTS song before? If so, subscribe!
Where are we seeing AI applications?
Iāll break this down by current (or in the near future) and future applications from a content/art perspective.
Gaming (inspiration from Jonah Blake)
Current: Character (NPCs, playable characters, bosses, etc.) ideation and storyboarding.
Future: Character ideation and rendering, ideation and rendering of how characters can move, sound. Item and weapon creation, environment/world/level creation.
Media
Current: Character and content (whatās the plot? where is the story setting?) ideation and storyboarding
Future: Writing whole scripts and themes
Art (inspiration from Claire Silver)
Current: Generating art and images for inspiration and ideation
Future: Full-on art creation including the individual artistsā historical style and finishing touches based on past work
Content
Current: Short (ad copy) and long form (blogs) content ideation and generation.
Future: New content based on a companyās historical content and social media commentary to create FAQs and blog posts automated with a posting schedule.
Iāll focus on the content category since Iām more familiar and Iām doing a lot of writing these days. Whatās going on there today?
Writing content on Twitter? Twitter writing prompts are here! This isnāt quite AI, but this is one step closer to half-baked Twitter threads that are AI-generated. Youāll need to put in your personal touch or make sure the thread flows well, but future Twitter threads could be 80% complete before the author even starts typing.
Need help with your ad copy, social media, blog, homepage copy?
Is this turning into an episode of Black Mirror? A little, but itās important to note that itāll be a while before AI can fully take over the human touch and discerning eye. However, AI will speed up the ideation/brainstorming steps significantly.
Will we lose our jobs? At some point in the future, yes. Especially if weāre average at whatever weāre doing. But until then, Iād love to see Copy.ai write in my style š. Eventually, Iāll need a vacation and want AI keep my 5 posts/wk streak going, right?
Dall-E 2
As of yesterday evening, I wasnāt sure what Iād write about today. However, as the evening went on, I figured itād be nice to fun to write about DALL-E especially since it was brought up twice in Twitter Spaces.
And write right (hehe) before I started writing this post, I came across this tweet. Todayās topic was meant to be!
Why is this a big deal?
Initially, DALL-E generated images were not allowed to be used for commercial purposes. Meaning you couldnāt type in a prompt and sell the output as art or NFTs (BTW, AI generated NFT collections are already a thing with limited success, for now). Could you use DALL-E as starting point for ideation or inspiration, then create your own work or materially modify the output and then sell it? Sure.
Thatās changed with DALL-E 2.
Remember those applications I mentioned above? Things are going to get interesting.
Fortunately, the OpenAI team understands that its products are powerful and may have major implications down the line. They expand on risk and limitations here. After all, with great power comes great responsibility.
Imagine: OpenAI unintentionally becomes Skynet in 10 years šā¦šØ
What is art?
The Twitter Spaces with Claire Silver (popular digital and NFT artist) on Rug Radio was interesting not only for discussing DALL-E, but also the more philosophical conversation around what is art?
Claire had some great thoughts about this topic, employing the spectrum framework while also keeping an open mind to various camps and schools of thought. I often employ this mindset with topics that might be controversial or hypotheticals.
I generally agree with this framework, and the general framework applies to other fields, eg: Copy.ai and marketing content creation.
Letās look at what this might be withā¦newsletter content of course :)
This gets pretty crazy, huh?
What does TPan think?
Iām torn here. Iām a fan of technology as an enabler and tool. However, I believe itās important to have heart and soul in the work that we do, especially when it comes to art.
As of today, I can get behind most of the AI spectrum above except for the far-right. If fully disclosed, I think it could be great for specific situations:
If the artist is a master storyteller that utilizes AI-generated capabilities to amplify the story
If theyāre also an accomplished āmanualā artist
Iām sure there'll be more as more live use cases emerge
Also, it is currently easy to determine if someone is using AI-generated art end to end and claiming it to be manual: Just ask them about their process and watch them live. Same thing goes with writing (ghost writing might be different for scale/volume purposes).
That said I can see a future where we respect and appreciate full-on AI Art and itās proudly displayed at museums around the world.
Weāll see if/when I start incorporate AI into my work. For now (and probably for a while), itās all me š
See you tomorrow.
