172: Shifting our mental model
A segue from jazz, through creative practice, to benefiting from AI
Let’s begin with the critical balance of building muscle memory (i.e. craft) contrasted with the practice of freeing your mind to create fresh, remarkable, behavior changing ideas. Have you heard of Effortless Mastery (Amazon) by the pianist Kenny Werner? Decades ago I was a jazz major at the University of Cincinnati attending a National Jazz Educators conference and our department head, Rick VanMatre, said, “Look—attend the sessions which appeal to you, or sleep in if you want, but you are required to be in the hotel lounge every night to listen to the Kenny Werner trio.” It was easily some of the best education I ever received, listening to and watching Kenny demonstrate profound story telling through in the context of piano, bass and drums.
Anyway, effortless mastery is pretty simple: Practice and play carelessly.
And of course those four words are meant to provoke and I hope they instigated some reaction in you. The gist of it is, of course, to care less about the outcome, to create without care—because, as Kenny describes—when you do, the result is typically much better. Easier said than done, as we all know, which is why he wrote the book.
You can learn more about Kenny’s concept in this Google Talk, or this six minute clip with Nick Bottini. Or, you could jump into this awesome Third Story podcast interview from 2019 with Leo Sidran (Apple, Spotify). The podcast reminded me to break out my old copy of Effortless Mastery, which is one of the most useful books I’ve encountered.
I bring this up because we remain in the midst of tumultuous times. The world of AI techniques, platforms and tools expands and begs our attention. All while the work remains to be done. And listening to Kenny last week reminded me of the distinction between practice/craft and performance/output, which I think matters when it comes to using AI. So on the one hand, AI can be technique, the tool itself. Which means you have spend the hours, as Ethan Mollick suggests, to get comfortable with the lingo, the UX, the personality of the software. To practice. And then there’s the performance—using AI to actually get tasks done, to help you deliver the deliverables. Those are two very different things, the practice and the performance. It’s important not to confuse them. As Kenny points out in the Nick Botti video above, part of your brain is wired to reap the benefits of practicing, of learning. A very different part of your brain takes over when you’re creating. Does it make sense to approach the tool (Midjourney or ChatGPT, or the piano) the same way for both? Probably not.
And you need to practice to have any hope of creating to the level you aspire.
All this to suggest you might want to consider your AI practice, your AI skill and craft building as distinct from creating and getting the work done.
Because, as the business strategist Rishad Tobaccowala recently suggested:
“Work will change more between 2019 and 2029 than it has in the five decades before.”
(I think we could tighten that timeline to 2025-2029.)
Rishad has a book coming out about this change. What does it mean, to “change more?” I think we’re already experiencing some of it—the uncertainty around knowledge, especially. The fact I can, with the help of an LLM, gain a useful level of fluency in a topic I previously had none in, almost instantaneously. At the root of a lot of this change is an undulating question.
What is AI for?
It’s probably “for” too many things at this point to make answering the question easy.
A few weeks ago the analyst Benedict Evans wrote,
“Part of the pattern of disruption is that the new technology is always bad at the things that mattered to the old technology. You couldn’t use a 1980s PC to replace mainframes running a bank’s back-end, nor build Photoshop inside Netscape in 1995. The new thing is bad at the things that matter to the old, but it unlocks something else that matters just as much. And so, testing an LLM by asking it to do precise information retrieval as though it was SQL might be the same mistake as asking if an Apple II could run for a year without a reboot as though it was an IBM 360: ‘no, but that’s not the point.’”
When I’m consulting on AI I use a slide that asks the audience to try and resist applying their prejudiced existing definitions of software onto AI as available today. We need to shift our mental model when trying to gain advantage from AI. We have to practice differently, to care less about how previous software’s rules were applied.
Over at The Neuron, they described working with ChatGPT this way: “Things humans find hard (like complex math) are actually easy for AI to learn, while things humans find easy (like walking or catching a ball) are incredibly difficult for AI.” Or as NYU prof (and prolific AI critic) Gary Marcus wrote, “Implementing basic (but often hard to articulate) knowledge that is readily present in most ordinary humans has turned out to be devilishly difficult to capture in machines.”
Remember, AI is unlike all other software we’re used to.
What it’s “for” is something we’re going to have to practice before we figure it out in pragmatic application.
The new year is upon us.
The legendary marketer and idea person John Hegarty welcomed it with an affirmative exhortation:
“I implore you to throw significant effort into having ideas.
New ones. Fresh ones. The sort that capture imagination, and ensure that your brand continues to drive conversations this year. Marketers have spent much of this decade obsessing over data and trying to anticipate consumer behaviour.
But the greatest businesses don’t predict the future: they create it.”
Amen.
And let’s add to that context with a brilliant post from Bob Lefsetz.
“Giving people what they want is commerce.
Doing what you want is art. But that does not mean everybody will be interested in what you do. But when you get art right, it’s forever, when you get commerce right, it’s evanescent.”
(If you know Bob, I think you’ll agree this was one of his quintessential posts.)