097: Humans are made from words, and problems
[During - Session 03] An Intro to Generative Text Tools is getting messy but more fascinating
Our conception of ourselves and each other is expressed almost entirely by words. (And art and theater and film, of course. But so many words.) And here we are a year into discovering (and maybe accepting) this one thing we thought only humans can do—we can write—is duplicated, is perhaps equalled in some measure, by mere machines. Suddenly, we’re not singular or distinct in the cosmos, right?
“A constellation of generative AI start-ups promise to automate an array of tasks we’ve historically considered for humans only: drawing, painting, image editing, audio editing, music writing, video-game designing, and more.” — Derek Thompson, The Atlantic, December 2022
The machines came for the words.
A year ago, while writing a different curriculum on the fly as these tools came into view, the revolution seemed a novelty. There were cute but worrisome hallucinations. But generally speaking, there was awe.
Now, second curriculum same subject, there’s GenAI methodology out the wazoo. And a shift in mindset, too, which matters even more.
How do we solve problems?
First we talked about them. Then we wrote parts down. Then we shared what was written, and wrote even more. And if you’re a writer, you wrote and if you’re an engineer you engineered—distinct from one another. But within a generative chatbot construct, the distinctions between “I’m writing poetry” and “I’m distilling a PESTLE analysis” and “I’m writing a speech for shareholders” is thin indeed. Never mind multimodal images. The same chatbot can shift and pivot a quandary through the roles of a corporation in one thread.
Teaching GenAI text tools isn’t so much about text as it is a new way of thinking about and operating with roles, processes, functions and yes, words. For tonight’s AI for Artists and Entrepreneurs class at MCAD we began with the basics of prompting inside ChatGPT, Bing, Bard, Claude:
Ask for inspiration, not facts (“a SWOT of”)
Begin with a point of view (“teaching a class” and/or “as if you were”)
Provide context, the more the better
Use chained/sequenced prompts to keep focus
In short, it’s not Search. You’re working with a pattern-recognition machine. Think a little differently.
Then we took time to benefit from the wisdom of people like Jeff Su, who breaks down six prompt elements and their prioritization in eight minutes. And Bri Does AI, who demonstrates how the Pareto Principle, Feynman Technique, and the Socratic Method unlock skill building using ChatGPT. There’s a weird kind of democratizing and mashing up here where all the possible ways writing and problem solving can occur simultaneously.
And then there’s my favorite part.
Last September I wrote,
“What feels unprecedented using LLMs is the ability to dialogue with the tool itself, to be able to ask it, “how else can you solve this?” Excel might calculate, but it won’t teach me how or why it did its work. Instagram doesn’t dialogue with its users in the moment of creation. Humans do this, of course, but not the ones I need at 1:14am when I’m busy writing. ChatGPT is happy to oblige at any hour.”
So we’ve got this tool which will concurrently write, analyze, distill, give example, edit, and translate the alchemy of words—and if you get stuck, the same interface can teach you how to use it more effectively. All this to say an Intro to Generative Text Tools is now even more serpentine. And we didn’t even venture into the same habits applied to GenAI text within, say, Word or Powerpoint or any of the corporate productivity tools. (Ethan Mollick offers a 59-second primer on the state of the state.)
Next week’s AI course begins a two-part exploration of generative visuals. And tomorrow’s Persuasion & Marketing course travels that stretch in human history where religion, government, and languages formed, leading into the invention of the printing press.