Serial Novel: ISLA
Missed something? Sketchbook | Stories | Dispatches
Shaving trees into toothpicks
I read a post about a new wave of agentic capabilities in AI1 — that’s where a system can operate autonomously — and I’ve been thinking about it for the last two days. The article has been viewed about 75MM times so it’s safe to say it struck a nerve.
In it, the author, Matt Shumer (CEO of HyperwriteAI) paints a pretty bleak picture for, well, everyone, but especially up-and-coming folks. Simply put: if you’re just starting out in any number of trades: AI will take your job. Too bad, so sad.
But what if you’re already established in your field? [Office Space Lumbergh.AI] Uh, yeah... we’re gonna need you to move to the basement, too.
It’s no lie that companies far and wide are jumping on board with this idea.
Why? Because AI is smarter, cheaper and faster at doing your job. This is all justified by the steamroller of progress, you see. It can’t be helped.
Shumer says this evolution can be slowed if you’d spend time — just an hour a day — using the tools. Think of this as training for how a new cog (you) might adapt to the future. Thank god there’s a way through this, right? Not quite. Your job is still in jeopardy but at least they’ll have some of your money in the meantime.
Now, I’m not a luddite and I think there’s value in researching new tools. We should seek out tools that enhance us. I don’t do art well, so I use Photoshop, Pixelmator or Express. I like spreadsheets but loathe Microsoft Excel so I use Google Sheets, lately with some scripting help from Gemini. And I frequently use ElevenLabs reader app for listening to my stories (Burt Reynolds is my favorite) and have even used their licensed voices for audio versions of my stories.
Let me say this clearly for those who mistake my intent: at no time is my hand off the wheel relinquishing creative control. I’m still the one making all the poor gramerical2 errors. Because, for my money, the subtle pen might be mightier than the sword but a keyboard in flow state is a howitzer by comparison.
But there was something in Shumer’s salesmanship of investing in this technology gave me pause.
Just look at the marketing of any LLM and you’ll see a system to replace writers, software engineers, even teachers... for only $20/mo. The frugal side of me says “that’s a paltry sum to pay for such expertise!” But forget the dollar figure and look at the *real* cost.
To find that, we have to look under the “Nice Price” sticker to see the obfuscated fine print. An LLM isn’t a calculator for doing math faster, these are systems that mean you never have to do math again. Or composition, history, and any number of learned studies that feel too hard. I mean, with 100 trillion pages of data at your fingertips, why bother with school at all?
This is where I see the rub. Leaning heavily on these tools feels like a no-brainer... but that might be exactly what it sounds like. When you take away the friction of learning something, failing, trying again and, hopefully, succeeding that hollow sound between your ears becomes deafening.
When you abstract the “work” of learning 2+2 by jumping to 4 it’s the same to me as shaving a tree down to simply make a toothpick. Not to mention that all the while the AI is kissing your ass and calling you brilliant. Craig Mazin on the Scriptnotes podcast had a similar take.
In the end, it’s a loss of something far more valuable: the frustration and exhilaration of learning. Of learning to think critically. Maybe, of having your own opinion.
After all, if the solution is always at you fingertip, what’s the point of life?
At the behest of velocity we lose the value of looking out the window to wonder and to dream.
And dreaming — not hallucinating — is what gives us life.
If you’d like to read something written by a human, we have a ton of books at Tiny Worlds written by fabulous authors like Lyndsey Resnick, Sean Thomas McDonnell and Hanna Delaney
I have a story you might like there, too. It’s about AI on the surface, but it’s not. Grab a copy of The Zeno Paradox for free with code: ZenoZero
I use “AI” because this is the common marketing term but what I really mean is LLMs (large language models). These are the predictive engine underneath the veneer.
😉
I should've gone into the trades. It'll be a while before an LLM can install a toilet. (Although, I'm sure that, given an appropriate robot body, it would probably try. Or maybe it would just say, "Ew. Plumbing? That's human work.")
I think it's really the same old problem when considering AI, garbage in garbage out. The assumption when AI is used is that you are getting valuable information. We've probably all heard that a very large percentage of research has been faked. So just like asking any "expert", take it with a grain of of salt and remember that common knowledge is often wrong. Always consider that for anything requiring a conclusion there are always assumptions buried in the calculation.
I agree that it's a powerful tool but it IS just a tool.