093: The irony of innovation
[Before - Session 2] I would have posted sooner but there was a spectacular sun set and date night to be enjoyed!
I’m freelancing with an art director who describes GenAI tools as casino slot machines. Lots of drama and excitement, but rarely a decent payout. And he’s not wrong, if your expectation of the tool is that of reliable miracle worker. The Sam Altman’s of the world, and let’s be honest—every tech founder previously—are literally required to overstate possibilities and outcomes. It comes with the title. Without hype, tech startups very likely can’t scale interest and engagement quickly enough. Perversely, you and I kind of need the FOMO, need the hustle, to justify getting involved. The ethics of this are meant for another post.
A year into this collective GenAI adventure, I’m not surprised the average result from the LLMs and Diffusion models are typically mediocre. But the difference is, that’s what I’ve learned to expect. To achieve “beyond mediocre” requires a keener curiosity, and the bandwidth and patience to struggle through hundreds of banal attempts. (See image above: Midjourney and I tussled through eight rounds—and it was the appearance of the playing cards after eight prompts and numerous “creative upscaling” that sold it for me. Dall-e3 and I sparred for ten until I gave up on its ability to capture comedic robot pathos in a casino.) In preparing for tomorrow’s second session of AI for Artists and Entrepreneurs I tried adding up the hours I’ve spent over the last 12 months on text, image, sound and motion AI. I’m betting it’s 600 hours. And I suspect my wife has a more accurate accounting.
The current UX of GenAI carries a lot of the blame for misrepresenting the miracle of this new experience. Back in May 2023 I said,
If you’re reading this, I’m guessing you’re closer to the leading edge of the bell curve of adoption than its center mass. You’re curious. And I’m guessing you provide all kinds of grace when experiencing the new. Early adopters forgive all kinds of coding and design sins which the middle of the curve can’t.
Today’s AI chatbot UX remains sub-optimal and too mysterious to help more people comprehend, enjoy, and leverage the possibilities.
There’s a lot still required of the average user to gain useful ability from GenAI delivered via chatbot. Which is why this week’s class is going to dive into some history, definitions and framing to ground ourselves in the mechanics and strategies of the tools. If you understand the motivations of the builders, you stand a better chance of leveraging what they built.
As Wharton’s Mollick has observed numerous times—there is no owner’s manual for ChatGPT. It’s hilarious to consider a firm valued above $100 billion ships a product used by perhaps billions of people and there’s zero user training provided by the company. But there are savants on Twitter, gurus on YouTube, and goofy instructors like me trying to make sense of it all.
Tomorrow we will begin at the beginning—meaning the late 1930s as science first attempted to model human cognition in math and mechanics. And science fiction began to shape our current conceptions of “intelligence” operating within circuits and algorithms.
Look for a “During” recap late tomorrow night or early Tuesday.