The AI Paradox: Job Threat or Career Catalyst? What Every Knowledge Worker Should Know

In this post, I answer the question, "What every knowledge worker should know about how to prepare for the coming AI Jobs Apocalypse." The answer: our ability as a species to thrive at work depends on just one character trait: resilience.

TL;DR - The article reinforces the importance of resiliency at work. The adoption of these tools is going to raise averages, perhaps like never before. As an instructional aid, I illustrate how the cumulative effects of AI are accrue in ways that are hard for us to imagine, using different business situations. In one scenario, I use three fictitious daily rates of AI productivity improvements to estimate when nuclear fusion will be ready (assuming a baseline of the mythical "it will be ready in twenty years.").

All images in this post were created using AI, and I leave them as they first render, with no subsequent tweaking. All the writing (with the exception of one paragraph) is my own buttressed by my trusted editing coach (aka, Grammarly ) for the line-level editing.

What does it mean to be average?

I want to tell a story about a line and how fast it changes, but there are many ways to do it.

First, there's the average over time.

I can draw a line in the middle and two dashed lines above and below. The middle is average. The two dashed lines show the width of the data over some discrete time series.

Wider spacing between the dashed lines means greater dispersion. Narrow spreads bring greater consistency and can magnify the impacts (positive or negative) of external effects.

What happens if I change this? What happens if I change that?

What if I told you that there are 8 billion people in the world, and we will pretend that they are all writers, although their skills vary a lot. Through the theater of the mind, we transport ourselves to the point where a magical technology ranks each writer with an Amazon star ranking, 1 through 5.

The Average Rises

But what if we start moving that average UP by N% every day. Assuming no change to the top of the market, the bottom rises. As the bottom rises, so does the average.

The DISPERSION of the data between high and low is also likely to take a massive hit because now, as ChatGPT (and his Neural Net Boy Band of Claude, Bard, and Grok)*, you have what is arguably the same author for a rapacious, gluttonous amount of so-so content. The quality will be as mixed as confident, prominent European royal bloodlines.

Ok, that's great. But now what? How long will it take for there to be a material improvement?

Nuclear Fusion On In How Many Days?

Let's say we're no longer talking about books or writing. Now let's switch it to red meat: the number of days between us and the realization of nuclear fusion--the Shangri-La of clean energy, where unicorns float with cotton candy hair curls and Mary Poppins is always nearby.

To be explicit about our assumptions, nuclear fusion is achievable, and the Twenty-Year-Promise is true.

My thesis is that the technologies are becoming so advanced, the incentives and human biases so "irrationally predictable" (from Dan Ariely, James B. Duke Professor of Psychology and Behavioral Economics at Duke University and a founding member of the Center for Advanced Hindsight), that we appear to be governed only by the laws of physics at this point.

Cthulhu Does Selfies

Automation's creative destruction will have massive social impacts as the fantastic, soulless Cthulhu, a character brought into terrifying focus by H. P. Lovecraft. in "The Call of the Cthulhu (published in 1928).

As knowledge workers in an age where AI looms large, it's crucial to understand the inherent cognitive dissonance of this technology. In his book "AI Superpowers," Kai-Fu Lee warns of significant job displacements in fields involving repetitive tasks due to AI's rise. Conversely, Andrew Ng, co-founder of Google Brain, highlights AI's potential to create new jobs and industries, much like the internet revolution. (see footnote)

So, back to our example. With the advent of free energy from nuclear fusion, I want to know how many more days I must keep working on terms that may not always suit my current life preference.

I want to know how many more days I must wait without the promised fecundity of Beast Levels of AI Milk and Honey; we are rightly skeptical to accept.

Disclosures and Disclaimers

Before we proceed, though, let me manage your expectations. I am not great at math. I can calculate 10 to 25% of a restaurant tip with abandon (and not without some degree of petty showmanship). Save for that; load your Excel sheets and do the math. I provide the formulas for those of you biased to that level of detail.

With that disclosure completed, I should qualify my following comments that I do not make predictions. I have worked with enough people who are so much smarter than me that I find myself bumping up against the ceiling of my intelligence when we speak Scot Baldry, are you listening?). So, what I write here is for my own enjoyment and understanding. I am also happy if it benefits you, but that is secondary to me now.

Let's say that the number of days remaining to Fusion Day 1 is calculated by iteratively running an improvement model against an established baseline number to see how long it takes to go from the state Not Fusion Day 1 to Fusion Day 1. It assumes the starting point is 20 years, or 7,305 days away. It's a lot. (Super helpful from Google: for an approximate result, multiply the time value by 365.3).

The game we are playing is to see how much LESS time it might take to reach that goal if we unleash The AI Kraken on the problem.

The benefits will be realized in the aggregate, to be sure. Captured from the virtuous cycle of positive benefits from upstream AI Tool adoption (left intentionally ambiguous), we expect N% progress each day to reduce the wait for free energy to zero days.

If the thesis holds, then we may realize this memorable quote from the movie "Butch Cassidy and the Sundance Kid" when Butch says,

"I think we ought to get out of here before they kill us."

We win with free energy, accepting, minimizing, transferring, or avoiding the risks as we fumble through. We'll probably be shootin our way out at the end too, anyhow.

I recognize that this kind of trend prediction is always dumb and wrong. So many real futurist thinkers, like Ray Kurzweil, whose slide rule and graphing skills to project the Singularity have caused so much trouble that I am unlikely to add meaningful insights to their work.

Following the prompts and chatting with Bard below, you can see my approach.

But before I tell you the results, I want to tell you a story. A story about a man named Jim.

Story Beats Ideas, Every Time

Jim was an agreeable fellow, though what he lacked in looks, he dazzled in brain power.

But he could be a bit tart, sharp even. His coworkers at Acme Marketing & Co often bring him their problems, technical or not. Mostly not.

He struggled to hold his tongue because he had observed, referenced, cataloged, diagnosed, and prescribed appropriate courses of action for their presented problems.

So, if they bring him a problem and he provides a solution that is likely the best offer they will get, and then they persist in the same behavior without taking his advice, he needs...clarification.

He told them the solution, but they were still talking. The solution he suggests does NOT require further elucidation beyond the appropriate detail he has already provided. So, he remains confused and often angry.

"People are so stupid. Nobody listens."

Then Acme Marketing purchased the latest AI Tools, and he noticed a change in his coworkers, after a time. They seemed nicer. They started talking to him. Did Marcia wink at me? What is going on? His spidey sense goes on red alert because he trusts few and believes none of those he trusts. He only sees data.

But the data starts to change. The company's performance across every demographic and industry vertical was improving at double-digit rates that a few years ago would have seemed impossible. Something had happened. Gravity weakened?

Well, AI had happened. It started to infuse, inform, and infect every process. Quality, timeliness, and profit all improved. Revenue was up and to the right.

That mattered to Jim because he had a lot of stock invested in the company and had done his own analysis before joining. In that assessment, he determined that there was an excellent chance that this company would go to super-sonic in the next ten years, which was why he joined. His instincts panned out.

So, while Jim can attribute the company's performance to these improvements (and he loves how the benefits accrue to him directly through his personal balance sheet), he struggles with the people.

Marcia was looking good, though. She seems super healthy now. Her clothes fit her better. She's got a spring in her step and throws warm energy Jim's way. Now, he is intrigued.

They both work on the same floor. It is a typical modern workspace layout, with glass and steel painted in a muted, Bucolic Beige color to reinforce the message that this is NOT a prison. And while you ARE free to go, you may be judged.

In the northwest corner of the building and on the thirteenth floor sits a small cafe with a new Nespresso coffee pot. It makes the best coffee in the building; and a line often forms around 10 am every weekday.

At 9:57 on Tuesday, Marcia corners Jim at the end of the line. Jim instinctively backs up. He is unaware that as he moves back one or two inches, he lifts himself by his toes only slightly, as though startled and afraid of an attack.

"Hi Jim!" She smells good.

"Um, hi, Marcia. How are you?"

She opens her eyes wider for half a second. Then they narrow again.

"Oh, I'm great! It's great to see you!"

Hmm. "You too."

Their conversation continues, and Jim relaxes when the discussion turns to work, something he can relate to. They begin talking about the new software and what a difference it makes.

Then she said something else that catches his attention:

"And it's even better when you use the AI Tools for home stuff! It's incredible. I've completely changed my life! I've learned so much."

Jim sits and listens.

"And all those things that you always told us. You were right. We just didn't know how to do it. I mean you gave us the answer...but then we were going to have to figure out how to get there. But I don't have your brains...but the AI Tools do, so I just asked it all along the way and it has been super helpful. It's been like having you right next to me all the time."

Blink blink.

Jim's eyes widen. His pulse quickens imperceptibly. The edges of his lips both lift slightly. His laugh lines stand out. His forehead relaxes, and he turns his head to the left somewhat.

"Marcia. Would you like to go to dinner?"

"I was wondering how long it was going to take you to ask! But yes!"

-????????? Scene -

Ok, so it's not When Harry Met Sally. I warned you to lower your expectations. Hopefully, now you see why.

Back to focus

To reiterate where we stand. The problem before us is that we want to conduct a calculation that estimates how soon nuclear fusion will be ready, mainly due to the cumulative effects of AI.

Our presumed baseline is that nuclear fusion remains stubbornly 20 years, or 7,300 days, from today. But we want to do better.

In my model, our improvements--which we attribute almost entirely to the adoption and mastery of AI tools--are tested for daily changes at three levels: low, medium, or high. We seek to understand which situation is better across a wide range, if only as a method to model our uncertainty. To be clear, by "better," I mean the time gets reduced by a discrete and consistent amount each day until we reach zero. From rocket science, you can think of it like delta-V, only in reverse.

Get your rum and those little umbrellas ready. Les bon temps are about to roulet.

According to Google's Bard, the days just before Halloween in 2026 will be interesting. Because using my "mid" calculation that assumes an A% improvement each day, Fusion Day 1 will be October 11, 2026, 2 years and 269 days from the date of this writing.

Assuming a "high" rate of improvement (B%) will be an interesting Easter this year because that date is just 100 days from now, April 24, 2024. 04.24.24. (Remind me to call my broker and reach for my Sustainable Yachting Magazine's latest article, "Big Boats 4U When The Benjamins Do Rain.")

There are a few (and I identify with you) who are eager to know what the "low" date is (C%). It's not good. The date is 273 years and 286 days from now, on 2297-10-30. Good luck with that. (I'll be busy but stand by your phone; I will try to send you a TikTok from wherever I am).

As I warned you initially, I only do simple math. This may not be hard, but we likely have severe differences in defining "hard math." A latte, plus scone and 18%? I'm good. Multi-variate calculus? Well, what other coffees do they have?

This entire post started by explaining the relationships to my kids and why differences are important. How they matter, whether between people or events.

And we are talking about events. How they are measured. I explained that dispersion, variance, and standard deviation are all fancy words that just mean how many rings outside the bullseye target they are in when they shoot during their archery class. Tight clumps of holes in the center? Super consistent and predictable. Low variability. Arrows that flew off the page? Not so much.

One son mentions a kid at archery who they call 2-Rings. When I asked why they called him 2 rings, they stayed within the first two inner rings whenever he shot his arrows at the target. Always. His heritage is Native American, and he is whipsmart.

My youngest says he gets scared when he uses the bow himself. His arm shakes, and when his arm shakes, the arrows go all over the place.

I said, "That's right. You could call it a fear index. We've go something like that for the stock market. It's called the Vix."

"So it's like Fear Meter? [At this point, he makes a scary sound]

"You got it."

I'll wrap up by saying that the application of these fictional stories (and yes, they are all fictional) is that things are about to get freaky. So gird your loins and whatever other parts you may value.

The transition to these AI tools' production, adoption, and benefit realization is "likely to be unpleasant," as I imagine John Kleese muttering in my ear. This is going to get much bumpier before it gets better. Our ability to pivot mentally, physically, emotionally, and spiritually will soon be tested in ways we have never tried before. This is our wake-up call.

And so, and but, for the pain is the gain. It is what we seek. We want the world to thrive and simultaneously help reduce human suffering. Those are our twin goals. The example I use with Fusion Day 1 is a known inflection point. All bets are off once we as a society reach the point where the incremental energy cost goes to net zero.

Free resource flows mean that the Chocolate River in Mr. Wonka's factory was actual unless it's not. And who doesn't like Willy Wonka?

Monsters. That's who. You must decide which are you?

In my next post, I will discuss why our desire for certainty leads to uncertainty. Crazy makes sense. Always has. Always will.

Citations found and fetched by ChatGpt 4.0:

1. Kai-Fu Lee's View: Highlighted in "AI Superpowers: China, Silicon Valley, and the New World Order," discussing AI-induced job displacement.

2. Andrew Ng's Perspective: Reflected in his talks and interviews, focusing on AI's potential to create new job opportunities.

Disclaimer: This post and all other writings that I submit on this platform represent my personally-held views, formed from my own personal research. I am not authorized to speak on behalf of nor do my views represent the views of my current employer, T. Rowe Price.