132: Skill still matters
More AI class projects: 2 GPTs and a video game
It’s easy to see why generative AI can be frustrating.
On the one hand, decades of science fiction have led us to believe AI tools will magically infer our intent, hide complexity, and “just” deliver a desired result. Our favorite sci fi movies and TV shows didn’t dwell on this current period of AI development where the code is relatively infantile. But here we are; living in a space where patience and curiosity are necessary.
This is also a time and place to reflect on the value of art history. Before the camera existed, you and I would have spoken about a proposed image using terminology which proves useful today when coaxing generative AI. It boils down to thinking of your image request on at least two levels: 1) the shape, architecture, structure, or “bones” of the request, and 2) the style or façade, i.e. color, lighting, and surface textures of the request.
Last week I shared a student’s final project which leveraged Structure (inside Adobe Firefly), as well as SREF, Image Prompts and Toner (in Midjourney) to help produce very specific visual outcomes. Then I ran across this post from SystematiqAI (registration likely required but worth it) which diagnosed the topic further, to wit: What to do when you’re “asking Midjourney for a lot.” In this case below, a specific facial expression, a specific environment, with specific visual attributes.
Note how the team at Systematiq organized their prompt below—first with an Image Reference, then the text direction, followed by two Style References (one of which is also the Image Reference).
Turns out getting an explicit idea out of your head onto a screen, or canvas, takes some skill. Even with AI helping you a long; at least for now. Those with experience in art history, terminology and application have an advantage in this age of generative AI.
🤖👨🏾💻 AI for Artists and Entrepreneurs: Max Giammona and Krish Agrawal
In addition to video and furniture generation, our MCAD students also focused on generative AI coding for their final projects.
Krish Agrawal (LinkedIn) decided to pair up Google Gemini and ChatGPT to develop a web-based “asteroids” game (play it here—make sure your cursor is over the playing area; use arrows to move and space bar to fire). He used GPT 4 to author the initial code—his first prompt was, quite literally, “let’s build a video game like Asteroids.” With the initial result in hand, Krish asked Gemini to evaluate, clarify bugs and suggest enhancements, which he pasted back into GPT 4. You don’t necessarily need to know how to code, to code.
Max Giammona admitted healthy skepticism towards AI at the onset of class. But then he discovered a utility for his work in storytelling and animation. First, can an AI help organize storytelling to fit specific constraints, like a chapter limit? And second, can an AI help a storyteller with continuity, and evaluate new script ideas against an existing story bible? Turns out the answer to both questions is “yes,” if you build a custom GPT (which anyone can do now, and for free). Here’s Max pitching his work:
You can use Max’s two GPTs in OpenAI’s GPT Store. His ChronoPlotter GPT is here and “T” the Lore Keeper is here. Both are free and available to anyone.
From my standpoint in marketing and advertising, I can see how GPTs like a lore keeper can and will support diverse teams in keeping a brand on track. Imagine an AI trained on your brand’s voice—every headline or piece of copy which exemplifies how the brand speaks at its best. At the very least, this tool gives brand managers and account executives a means of considering how and why new writing can work. While I am fully aware too many will simply default to using a GPT as a grading tool, the smart ones will use this type of AI to protect great but risky ideas.
AI+Creativity Update
👨🏻🏫 Wharton prof Ethan Mollick published another mid-year summary of how to think about and engage with AI across a range of creative and strategic tasks. As always with Mollick, this one is very much worth your time.
🎉 If you’re in Minneapolis, Adobe is hosting a Create Now event on Thursday, June 20 starting at 5:30pm at the American Swedish Institute. I’m hoping to attend. Maybe we can ask them about the whole ToS kerfuffle!
🤖📊 “In the latest McKinsey Global Survey on AI, 65 percent of respondents report that their organizations are regularly using gen AI, nearly double the percentage from our previous survey just ten months ago. Our 2023 survey found that AI adoption did not reach 66 percent in any region; however, this year more than two-thirds of respondents in nearly every region say their organizations are using AI.”
👨🏼🏫 Shelly Palmer offers some useful ways of thinking about classifying, weighing and integrating external data into one of the major frontier models (ChatGPT, Gemini, et al). The more you engage with AI, especially for strategy, the more this type of information will prove essential.