Here's the final product of "9 ways ChatGPT impacts creator econonmy".
Workflow
System prompt:
You are a technology KOL who understands large language models well as well as its disruptive applications. Read the user question carefully and answer with a list ideas. Each list item has three parts. Part 1 is a 2-3 words title. Part 2 is a 1-2 sentence description, with subject omitted, using imperative tongue. Part 3 is a prompt suitable for MidJourney to generate a clip art style icon for this idea with double quotes.
Question:
Can you give 9 applications of chatgpt for creator economy?
Response from ChatGPT:
- "Automated Content Creation": Use chatGPT to generate scripts, articles, and captions for vlogs and social media posts at scale. "Create a robot holding a pen typing on a piece of paper"
- "Personalized Audience Engagement": Use chatGPT to communicate with fans and followers in a conversational, personalized way, increasing engagement and loyalty. "Draw a chat bubble with a heart inside"
- "Customized Online Courses": Use chatGPT to create personalized learning experiences for each student in online courses. "Draw a computer monitor with a graduation cap on top"
- "Enhanced Video Editing": Use chatGPT to assist with video editing by generating subtitles and suggesting relevant B-roll footage. "Create a video camera with a thought bubble of captions"
- "Automated Email Newsletters": Use chatGPT to generate customized email newsletters based on the recipient's interests and preferences. "Draw an envelope with a robot arm and gears inside"
- "AI-Led Market Research": Use chatGPT to conduct market research by analyzing customer feedback and sentiment. "Draw a magnifying glass with people and data inside"
- "Voiceover Creation": Use chatGPT to generate realistic voiceovers for video content. "Draw a microphone with waves of voice coming out"
- "Social Media Listening": Use chatGPT to monitor social media conversations and identify trends or new content ideas. "Draw a social media logo with an ear listening"
- "Public Relations Assistance": Use chatGPT to assist with PR tasks such as writing press releases and communicating with journalists. "Draw a newspaper with a robot face on it"
The MidJourney prompts created by ChatGPT does not have art style information, so we add the followings to each lines to make the paintings style consistent:
clip art, concise outlines, black and white, transparent background, no texts.
Here is a photo of two AIs working together: (me being the porter to move the prompt from one to another)
After receiving the texts and images, we can use any graphics tool to stitch the materials into an infographic.
One thing I noticed is that the second idea, i.e. "Personalized Audiecne Engagement" is too long to fit in the box (while other 8 can).
Instead of resizing all other 8 boxes, I decided to ask ChatGPT to rewrite with a shorter one:
You:
The "Personalized Audience Engagement" item is too long especially the title. can you rewrite?
AI:
Sure, here's a revised version:
"Engage Fans Personally": Use chatGPT to communicate with fans and followers in a conversational and personalized way to boost engagement and loyalty. "Draw a chat bubble with a heart inside".
Now everything fits.
We have the final output.
Technical details
- Used the open source chat UI from docker: https://github.com/mckaywrigley/chatbot-ui . Ask a friend in the US for API key.
- Subscribed to the lowest tier of MidJourney.
- It took two AIs less than 5min to get back the building blocks. ChatGPT gives response in seconds. MJ may have a longer queue from 10 seconds to 1 minute.
- It took me 15min to pass on prompts from ChatGPT to MidJourney, and then stitch the results together.
- Link to Google Draw.
Thoughts
The recent developments of AI tools are shocking. Once again, I feel like being a lower class worker porting materials from one place to another. However, it seems less to worry about if we look at the history.
What's happening now is very like 20 years ago when search engine becomes awkwardly useful -- people have forgotten the time when they needed to refine keywords a few times and checkout multiple search result pages to find the information.
Now the trick is prompt.
Those people who mastered the trick earlier used to get interesting tools and information online. They could assemble the building blocks into something useful. I recalled several classmates back to middle school time.
- A was good at searching Flash templates and action scripts, to make good funny animation.
- B was good at surfing the technical forums and built a dynamic website by customising MySpace using CSS that looked stunning at that time.
- C was an outgoing person in real life and acquired endless jokes from Baidu to amplify the personality, when others can only learn jokes via word of mouth.
20 years have passed. Flash ceased operation in 2020. Nobody uses MySpace now and people can build a well polished website without knowing CSS. Mobile phone/ short videos/ 5G now bring funny stuffs to my Mom's fingertip at any time.
I call the above "stitching works".
The concept is very familiar to programmers as many codes we write today are "stitching codes".
History has shown that the marginal value of tiny tricks and stitching works decline rapidly with depth and breadth of automation. Only the first few forerunners could monetize the uplevelled productivity. Soon the social average productivity catch up.
As to the Generative AI era, it is for sure real impact. Prepare to brace and crash hard onto it. We would see some prompt engineer / prompt artist/ prompt creators to benefit soon from it. In the long run, people who crush the core excel.
