100: Being patient, interesting and useful
[During] Part 1 of generative image creation; a history of Mass Media; and a different take on SB58
Happy 100th episode/post/letter! Thanks to everyone who subscribes—and thanks especially to those who send notes. I deeply appreciate your comments and suggestions.
This “During” edition covers both Monday and Tuesday classes I’m teaching at MCAD. The first exists on the cutting edge of now, and the second takes a look at how we arrived in this moment.
MONDAY
Maybe the biggest lesson to learn with generative AI is patience.
Especially when you’re generating images.
I just looked at my Midjourney account (/info) and realized I’ve generated 1,915 images in the past 12 months. I have a feeling my usage is very light compared to many.
The point being, most generative circumstances elicit crap for a long time until they don’t. So if you walk into the tools not expecting immediate miracles, you’ll be happier. It also helps if you have a plan.
Aside from patience, planning is the difference between “these tools suck” and reliable, useful outcomes. Monday’s AI for Artists and Entrepreneurs class jumped into Microsoft’s Bing Image Creator (DALLE3 under the hood), Google Gemini, and Adobe Firefly (Part 1 of our exploration of image generation—we’ll dive into Midjourney, Stable Diffusion and Pika Labs next week).
Do you know what you want?
An image of success? A photo representing conflict? A painting that illuminates collaboration? A hyper realistic illustration of product knolling? Even before you consider specifics, can you articulate the intention for your image? What purpose will it serve, what concept will it convey?
Generative tools are lousy mind readers.
So articulating your point of view is a useful place to start, even ahead of prompting. Even if it’s just in your head.
What’s marvelous about this space is you can use generative text tools to help think through and ideate potential image ideas. You can dialog with ChatGPT to work out what you mean to convey, then—pro tip: Ask ChatGPT to write your image generation prompts for you. And within multi-modal environments like Gemini or GPT v4, the platform can also generate the image.
There’s a stereotype we need to leap to prompt writing, and arcane keywords, but really—our primary challenge is in figuring out ourselves first, our intentions and expectations. Then the prompting stands a better chance.
Plan on lots of attempts.
We loosely organized Monday’s class around three assignments:
Thematic (i.e. “Creating a series of at least three images for a presentation on swimming technique”)
Specific (i.e. “Design packaging labels for a new canned soda called Poppi”)
Sequential (i.e. “Build a storyboard for a Super Bowl commercial with at least six frames”)
The first delivers a greater diversity of outcomes, a wider range of concept and style. The challenge with thematic generation is, again, articulating a point of view you want the images to follow.
The second assignment gets to “yes!” much faster; in part because the request is clearer, and the outcome much easier to predict. But even within such a simple prompt can be a world of variety.
The students worked on a lot of “compare and contrasting” taking the same prompt across Gemini, Firefly and Copilot/DALLE3. And quite a bit of refreshing and trying over. But I’ve found topics like packaging design seem to be well represented in any LLM training data, and thus—deliver useful results faster and with greater consistency.
Our third assignment elicited two general categories of result: Either a single image of a multi-frame storyboard, or multiple individual images that may or may not appear to be “of the same family.” Which is another way of say, this kind of assignment shows us the current challenges with these tools. As amazing as they are, things like “keep X consistent across multiple iterations” proves illusive.
TUESDAY
Yesterday’s Persuasion & Marketing class was focused on Mass Media, beginning in the 1700s as newspaper technology emerges, ushering in a need for new new roles and outcomes like Copywriters and Ad Copy. Funny to consider how job titles and deliverables so common now were once very new, and not that long ago.
“The business of attention is about being interesting.”
The author Tim Wu’s book, The Attention Merchants, is hugely useful in helping to discern the transitions from the singular dominance of print advertising into radio, then TV, and finally the Internet. His Google Talk is worth a look, too.
So much has changed since those first newspapers, and yet so much has not. It is both hilarious and depressing to see “fake news” existed then as now.
And I’m drawn to those early days as our industry, the business of attention and creativity, as they begin to establish foundations. Essentially, without the printing press and its newspaper descendants, there is no marketing or advertising industry. There’s a symbiotic relationship as publishers and editors court an audience, in part to trade on its value. And through those early decades, a siren call to be interesting. Even if it meant telling the world the moon was populated with monkey bat people, with a straight face.
Imagine words like “brand,” “positioning” or even “propaganda” coming into existence. They were once new, predicated on evolving media technologies. I recall two newspapers arriving at my family’s doorstep when I was a child—one in the AM, one the PM. Today, no one in my class has even touched a newspaper in weeks—if ever.
Now we might even question the notion of “mass” media at least in contrast to the apex of the 1970s when an advertiser could buy attention from a clear majority of the country via one of three TV networks. Interesting to note this week’s Super Bowl was, according to CNN, the “most-watched American television broadcast in a generation” at an averaged 123.4 million viewers. Maybe we still yearn for those giant, collective moments?
A different and useful Super Bowl perspective
Speaking of huge, singular moments: How did advertising’s big day rate when considering disability and especially representation of disability? Misfit Media, an org focused on disability awareness and training for marketers and agencies, wrote a lengthy and illuminating recap.
“When we consider that disabled people have that buying power, and are shown in ads less than 1% of the time, and usually through problematic tropes, it shows a glaring need for creatives to sink their teeth into disability inclusion with the same enthusiasm we’re all dipping our chips into guac.”
As Misfit notes, Google’s Pixel 8 spot was the clear representation winner. Versus the Kia spot—which suggests all kinds of problematic “savior” porn. And then consider all of the advertisers who could have represented some form of disability, but didn’t (pretty much everyone else advertising).
The situation could be different. And it starts with thinking about it.