086: Developing a generative AI curriculum
I'm curious what you make of this one. Class begins mid January.
I’m imagining the very first college level “Intro to Photography” courses circa 1830-40. Clearly they focused on technique, but they must have wrestled with appropriation, realism, and all of the “end of painting” bromides. Better still, imagine writing a curriculum for, say, writing, where you don’t acknowledge or use writing instruments of any kind or word processing or dictionaries. I suppose you could pull it off. But that would be a weird course.
Which is another way of saying I can’t imagine building a curriculum for “AI for Artists and Entrepreneurs” without using generative AI tools.
First, a huge thanks to Olaf, Robert and Sanjit at the Minneapolis College of Art and Design for asking me to write the curriculum and teach the course which begins January 15. But also many thanks to Jason, Jeff, Julian and the robust crew of fellow travelers lead by Ross our fearless leader who’ve gathered via Zoom so many Friday mornings to share our generative journeys.
Two caveats upfront:
I didn’t expect—and my process has proven it true—for ChatGPT et al to save me lots of time, or “do the work for me.” This is my fourth original college level curriculum, and the process and time required to create it have proven largely unchanged from circumstances without generative AI assistance. But as I said above, it seems inconceivable and even rash to develop a GenAI curriculum without collaborating with GenAI.
It’s a brave new world. I’m grateful for all the other academics authoring perspectives and resources, like those at: Barnard, Columbia, Penn State, Stanford, Northeastern, Tufts, Boston University, Harvard, UW Madison, NYU, MIT Sloan, and Carnegie Mellon. The tech changes day to day (see below). Teachers and students are building airplanes in the air. As Wharton Prof Ethan Mollick often reminds us, generative AI technology is as astounding as the lack of documentation on how to use it. I guess that keeps some of us employed.
How I built this
The good news is I began with an existing course description, which I had written initially without any generative assistance—and which was edited by my department chair and his boss. But then I wrote the prompt above to seek perspective from various GenAI platforms. Those results helped me unlock and refine the 15 sessions. And I’ve been bouncing various parts of what follows off the same tools just to see how they react. Sometimes things get better!
COURSE DESCRIPTION: No other technology has effected creativity, the arts and business as rapidly as artificial intelligence (AI). Comprehension and fluency in Generative AI (GenAI) tools is fast becoming necessary for a wide diversity of illustrators, designers, image makers, film makers, writers, strategists, and entrepreneurs. This course seeks to bring artists of all types up to speed in the verbal and visual GenAI landscape, including working within ChatGPT/Dall-e3, Midjourney, Stable Diffusion, Runway and other platforms in order to strategize, and create. We'll focus on understanding and leveraging these tools within the context of an artistic and entrepreneurial journey, and discuss how new technologies impact important topics including ethics, copyright and collaboration.
And here’s where we’re at with Course Outcomes.
Class participants will:
Gain productive fluency in verbal, visual, motion and multi-modal GenAI technologies
Be able to characterize the purpose and role of various GenAI technologies within their artistic practice or entrepreneurial venture
Produce a variety of verbal, visual and motion outputs that can support and enhance their portfolio of work
And now, here’s an outline for how the 15 sessions might flow...
Session 1: Introduction and Expectations
Introduction to the course, terminology, platforms/tools, instructor, fellow students
Text tools: ChatGPT, Gemini, Bing, Claude, Seenapse, Text.ai
Visual tools: Midjourney, DALL-E3, Firefly, Stable Diffusion, Emu
Sound/Motion tools: Runway, HeyGen, DiD, Pika Labs, ElevenLabs, Boomy
Overview for exploring GenAI in the arts and entrepreneurship; How disruptive technologies (i.e. historically computers and the Internet, now AI) affect strategy, creativity, the role of the creator, artistic practice, process/operations and outcomes
Discuss AI’s ethical, security, policy, copyright and societal impacts; Review MCAD’s rules on plagiarism; Tool vs Author—which carries responsibility?
Homework: Sign up for required platforms + Write your GenAI usage statement
Session 2: Conceptual Models and the Underlying Mechanics of GenAI
How does an LLM function? What is a model? A neural network? A GAN? How does diffusion work? Discuss the basic technical characteristics enabling GenAI tools and platforms
The major players: OpenAI, Google, Microsoft, Anthropic, Adobe, et al — Discuss the evolution of GenAI and its various business models
Review implications of GenAI in strategic and artistic ecosystems over the past 12 months, specifically noting disruptive and enhancing capabilities, e.g. what changes now?
Discuss how GenAI acts as a design material and collaborator
Homework: Use tools to respond to Assignment
Session 3: Intro to Generative Text Tools
Compare/contrast the UX, taxonomy and general outputs from ChatGPT, Bard, Bing, Claude, Seenapse et al
Prompt writing basics
Practice using GenAI text for business analysis and strategic problem solving
Practice using GenAI text for initial creative development
Practice writing versus editing text with GenAI
In-class workshop and potential guest speaker
Homework: Use GenAI text tools to respond to an Assignment
Session 4: Intro to Generative Visual Tools - Part 1
Compare/contrast the UX, taxonomy and general outputs from DALL-E3, Adobe Firefly and Meta Emu for generative visual content
Prompt writing basics for visual outcomes
Practice using GenAI visuals for audience profiles and behavioral moments
Practice using GenAI visuals for initial creative development
In-class workshop and potential guest speaker
Homework: Use DALL-E3, Firefly and/or Emu to respond to an Assignment
Session 5: Intro to Generative Visual Tools - Part 2
Compare/contrast the UX, taxonomy and general outputs from Midjourney/Discord and Stable Diffusion for generative visual content
Evolved prompt writing (i.e. seeds, styles, etc.)
Practice using GenAI visuals for product designs, logos, graphic design
In-class workshop and potential guest speaker
Homework: Use Midjourney and/or SD to respond to an Assignment
Session 6: Intro to Generative Motion Tools
Compare/contrast the UX, taxonomy and general outputs from Runway, HeyGen, Pika Labs, DiD for generative visual content
Prompt writing basics for GenAI motion
Practice using GenAI motion for talking head, product demo, background animation, etc.
In-class workshop and potential guest speaker
Homework: Use one of the motion tools to respond to an Assignment
Session 7: Intro to Generative Audio Tools
Compare/contrast the UX, taxonomy and general outputs from ElevenLabs, Adobe, Boomy, etc. for generative audio content
Prompt writing basics for GenAI audio
Practice using GenAI audio for speeches, teaching, commercial support, etc.
Incorporating GenAI tools into music creation
In-class workshop and potential guest speaker
Homework: Use one of the audio tools to respond to an Assignment
Session 8: Blending it All Together - Part 1
Evolved prompt writing—chaining, feedback loops, leveraging emotion, etc.
Exploring multi-modal systems; Using GenAI to help prompt GenAI
Introduction to Adobe Photoshop Generative Fill
Creating your own workflow across various tools
Exploring OpenAI’s GPTs
Midterm: Use a combination of GenAI tools to respond to an Assignment in class
Session 9: Application—GenAI for Artistic Practice
Use GenAI to
Inspire, write and edit storylines and concepts
Nurture visual exploration
Help develop characters, worlds, and aesthetics
Evolve episodic arcs
Workshop and/or Guest Speaker TBD
Homework: Research and report on artists using GenAI
Session 10: Application—GenAI for Entrepreneurship and Startups
Use GenAI to
Clarify milestones and outcomes for an entrepreneurial venture
Discern and evaluate the validity of startup opportunities and unmet needs
Research and write SWOT analysis
Evaluate contracts and investment information
Write and refine startup business plans
Workshop building an initial startup business
Session 11: Application—GenAI for Brand and Product Design
Use GenAI to
Understand Brand and Product Design fundamentals
Research and refine audience and merchandising insights
Develop initial brand and/or product design directions
Workshop a brand and/or product design presentation
Homework: Research examples of brands or products developed using GenAI
Session 12: Application—GenAI for Advertising Campaigns
Use GenAI to
Comprehend how ad campaigns are made
Research and refine elements of a creative brief
Develop initial ad campaign concepts
Create visual and motion assets
Workshop ad campaign creation from a brief
Homework: Research examples of ad campaigns developed using GenAI
Session 13: Project Work
Use GenAI tools in class to initiate, evolve, and produce an artistic or entrepreneurial outcome
Potential guest speaker
Session 14: Project Work
Continue refining your artistic or entrepreneurial project in class
Potential guest speaker
Session 15: Project Presentation and Course Wrap Up
Present your project and GenAI methodology in class
Discuss course learnings and perspectives on futures of AI within creative and entrepreneurial settings
So that’s the basic syllabus.
What would you edit or add? What am I missing?
I gave the document to Claude and ChatGPT and asked them to play the role of Academics Dean, and both suggested increasing the volume of ethics conversation across the span of the course. Claude suggested articulating “2-3 core questions or themes the course will coalesce around.”
What do you think?
AI+Creativity Update
🤔 Analyst Benedict Evans has released his annual presentation. This time (as you’d expect) it’s “AI, and everything else.” As he puts it, so much generative AI tooling suggests, “If you can ask anything… what do you ask?”
👍🏽 The team at Addition have built a prototype autonomous (generative) video editor. Imagine automatically scanning TikTok for keywords, downloading relevant clips which are then repurposed with original audio commentary.
🚀 Both Meta and Google released new generative tooling.
1️⃣ Meta’s released a web-accessible version of its generative image tool. Access it here. (You’re required to login. Oh, and the Emu image synthesis model was trained on 1.1 billion publicly visible images on Facebook and Insta. Does that render look like you? Well, now you know why, notes Arts Technica.) I was surprised by Emu’s speed; it’s very fast. Image quality seems close to if not on par with Midjourney v5. Meta’s tool doesn’t recognize aspect ratio in prompts, every output is 1:1. You get four images per render.
Prompt: A harried professor works late into the night on a syllabus with the aid of brilliant but mercurial Otter, in the background are lots of fax machines
2️⃣ Meanwhile, Google has released Gemini, its answer to ChatGPT. Here’s the blog post. Here’s the tool. I need to spend more time testing, but the handful of initial prompts I tried came back quickly, were well written and didn’t seem to be hallucinating. For example, I submitted the prompt above written for the AI syllabus to Gemini. You can view the result here. Not awful! But also pretty much what you’d expect, which is good?
Alberto Romero wrote a deep summary with an emphasis on Gemini’s native multimodality, i.e. a “multimodality approach resembles much more how the human brain learns from multisensory contact with our multimodal world.”
I’m a fan of Gemini’s UX. Always offering three drafts suggests the role generative tools should play, in my opinion, which is as a collaborator with options versus a dictator with only one solution.
And fine tuning options at the end are quite useful.
What have you discovered using Gemini?