Starting with the agricultural revolution ~12,000 years ago, we’ve witnessed the balance of power regularly shift between Labour and Capital throughout our entire economic history.
(Capital in this sense refers to tools & machinery that are used to produce a good or service.)
This “tug of war” ends up normalizing wages and raising living standards over time, since new innovations usually don’t end up replacing humans—but rather enabling them to be more productive per capita.
Ironically, automation actually tends to make us busier, on the whole.
However, occasionally this equilibrium can break for a while if a truly seismic disruption comes along that rewrites the whole ecosystem.
Things like the weaving machine, the printing press and the automobile were complete game-changers, whose adoption rapidly destroyed demand for a significant portion of the working population at the time; textile workers, scribes and carriage-makers, respectively.
Eventually, the net result has always been a substantial increase of jobs and productivity as the industry evolved as a result... but the adjustment periods were painful, and often led to social unrest—as with the Luddites of the early 19th century.
As I write this today, we appear to be on the cusp of another ‘Gutenberg Moment’.
There’s little question that AI is going to transform a swath of industries. Even in its nascent state, we’re seeing astonishing new capabilities surface on a daily basis; it’s genuinely hard to keep track at this point.
For instance, this week’s world-bending examples include a Youtube creator who can already clone himself and create indistinguishable, engaging content automatically—and a hot new track by Drake & The Weeknd... that they didn’t create.
And this is to say nothing of the warp-speed progress of generative platforms like CoPilot, GPT4 & MidJourney—the latter being run by a team of just 9 developers.
But let’s zoom in for a minute, and ask a few obvious, self-serving questions...
If these early prototypes are already creating engaging content, writing functional code, designing award-winning art (literally), crafting persuasive ad copy, constructing working websites from a napkin sketch, conducting insightful market research, and automating project management...
...what does this mean for people like us who make a living in the digital economy?
How much will we see AI replace humans vs. becoming a mandatory level-up for humans?
Which areas of the org chart will be most vulnerable to displacement? Which will be most resistant?
And what does it mean for the entire startup ecosystem if hiring is no longer the primary growth catalyst?
While nobody can answer these with certainty yet, what we can do is game out a few different scenarios—and evaluate whether there are any actionable risks or opportunities that surface in the process.
So for this exercise, let’s walk through 3 potential outcomes:
Let’s dive in...
The Concorde. Personal 3D Printing. The Segway. The Blockchain.
Each of these things were viewed as major innovations that would usher in a new reality for transportation, how things would get made, or an alternative monetary system.
Instead: the Concorde took to the skies for the last time in 2003. 3D printers are largely a hobby for enthusiasts and niche replacement parts. Segways were simply too weird. And the blockchain somehow turned into a cesspool of ponzi schemes and degenerate speculation.
And while it’s too early to pronounce any of these things as officially dead quite yet (well, maybe Segways), what’s clear is that for whatever reason—each of these “obvious” game-changers fell profoundly short of their projected impact.
The same could be true of AI.
In fact, prior to the recent rise of OpenAI, you could’ve easily made a similar argument that AI voice interfaces (Alexa, Cortana, Siri, etc.) are still light-years away from meaningfully replacing a human assistant.
At best, they’re conditionally useful for a limited range of activities. Anecdotally, as an iPhone user I engage with Siri perhaps 2-3 times a month, and generally only when I’m driving.
And according to the Washington Post, only 30% of smartphone users actually interact with voice assistants; even for the regular users, the “conversations” still feel stilted and unnatural.
With all of this in mind, even with the impressive capabilities that ChatGPT and generative models seem to have already—there is a chance that the current ‘AI revolution’ basically just becomes a fancy version of auto-correct.
It’s there, and it’s kind of helpful, but it basically just becomes yet another feature and life goes on.
Another potential outcome is actually that AI becomes too successful and too utilized... too quickly. And, much like ubiquitous over-usage of CGI in Hollywood during the early- to mid-2000s, what was initially “mind-blowing” soon becomes an obvious synthetic, and we simply tune out.
A user backlash is actually quite probable, as creators and publishers inevitably try to scale-up content volume during the adoption period; ironically this could push users into authentic, ultra-human content that simply can’t be faked.
(There’s a reason why shaky, selfie-stick “walkaround videos” are already more effective as an advertising medium than polished, high-production live action...)
Ultimately, in this first scenario, while AI will certainly join the productivity toolbox across a range of industries, the scheduled revolution (along with its associated job destruction) simply never comes to fruition.
While some might argue that globalization started in 1492, most economists view the post-WW2 era as the modern liftoff, which then accelerated after the resolution of the Cold War in 1991, and then moved into hyperdrive in lockstep with the information age of the early 2000s.
Why are we discussing globalization?
Because it’s arguably been the most disruptive event in economic history; the world has never before seen this scale of simultaneous opportunity creation and industry destruction. On net, global GDP and living standards have rapidly accelerated, lifting billions out of poverty.
But while the world is certainly better off on average, it’s not without its costs. You’d be hard-pressed to find many longtime residents of Detroit who view globalization favorably.
The cost of lifting the emerging world out of poverty and running ‘labor arbitrage’ to make technology and goods increasingly more affordable was the hollowing out of the West’s own manufacturing industry, and the suppression of its wages to chase productivity through Capital.
Advanced economies now largely rely on consumption and asset appreciation to fuel growth.
From this perspective, the AI revolution might well be the moment where “globalization” essentially reaches singularity.
It’s as if a new Silicon Continent has emerged; a land with an infinite population, where every ‘worker’ has memorized the totality of human knowledge, and has unlimited productive capacity. Oh—and you only need to pay them with electrons.
How much Labor do you suppose we’ll be outsourcing to this impressive new trading partner?
The answer is obvious. The question is how humans will adapt to leverage this new “partner”. And what are the boundaries and limitations of AI in terms of intuition, cultural nuance, and strategic decision-making?
Let’s explore what this looks like for internet entrepreneurs...
It seems highly probable that in the near future, almost all aspects of online content creation and digital marketing will be AI-generated, or at minimum, AI-assisted.
Even in today’s prototype phase, GPT-4 is already delivering outcomes that are within spitting distance of human marketers. For example, a recent split-test experiment by Smart Marketer compared the results from 3 direct response email campaigns; one set created entirely by a human marketer, the other created entirely by ChatGPT.
The results? ChatGPT generated almost exactly the same open rates and conversion rates as the human copywriter. The variances aren’t statistically meaningful. The only area where it fell short noticeably was in driving CTR% to a blog post, where the human email had more personality, and was easier to scan.
Their takeaway was that for transactional content (like coupon offers, signup offers, etc), ChatGPT is scarily capable, but in terms of brand voice and relationship building——it’s not there yet.
But who knows where we’ll be 6 months from now. (Or for that matter, 6 weeks).
This experiment highlights 2 commonalities that we see emerging as a best practices and early methodology for getting the most out of generative AI:
Perhaps Picasso said it best in offering his thoughts on computers—a new innovation in 1964:
"But they are useless. They can only give you answers."
Indeed. And to extend the analogy to art, AI is a proverbial automaton painter that can expertly mix colors, and execute brush strokes perfectly. It’s the ultimate, unlimited Crafter.
But it still needs an Artist to understand a culture deeply enough to know what kind of painting will impact people on an emotional level... or even on a societal level.
What AI brings to the equation is that Artists now have the ability to employ unlimited Silicon Crafters to bring their visions to life on a scale and timeframe previously unfathomable.
We might be entering the Era of the One-Man Band... in virtually every area of the creative economy. It’s not inconceivable that, soon enough, even things like AAA software, movies and games can be prompted into existence by skilled solo developers/filmmakers, and then edited and polished to perfection by a substantially smaller workforce of Crafters than is currently needed today.
This raises a number of interesting questions:
To list just a handful.
Another wild card is what happens when end-users have an active role in the overall AI ecosystem?
Think about it: Beyond running a few surveys or setting up some basic behavioral automations (eg. abandoned checkout flows), it’s very challenging and complex to truly personalize your business for each customer.
This is where AI could be really disruptive; imagine letting your users totally customize not only the types of content / offers they get from you - but also what that content says (or at least what it emphasizes), and how those offers are presented.
In this sense, you basically just define the core talking points, biases and objectives - and the AI fills in the blanks, personalizing messages for every individual user on the fly. Calling out their specific pain points, and highlighting their specific goals & aspirations on a 1:1 level.
Something like this would utterly transform digital marketing. And at this rate, the arrival of the first prototype might only be months away.
One last thought: In researching the limitations of AI in creative industries, I reached out to Filip Matous who runs a brand consultancy in London. Our discussion had a material influence on this essay, but he mentioned one thing in particular that really stood out:
“I think Art ultimately remains human and emotionally intelligent. But Craft? Yeah, that probably all gets eaten by AI. Humans will still need to know which questions to ask, which problems to solve—and how to polish AI outputs into a finished product that lands.
“But anyone in the middle—the drudgery of repetitive tasks? They’re screwed.”
Ultimately in this scenario, AI will drive almost infinite productivity, and drive the cost of certain services to zero. It transcends far beyond being a fancy auto-correct, and becomes just as indispensable as the keyboard to the digital economy.
But perhaps the biggest question is: How many “Detroits” do we end up hollowing out in the process? And how long will it take for Labor to close the gap—if ever—as Capital enters hyperdrive?
The simplest way to imagine how AI becomes permanently destructive is to modify the 2nd scenario (disruption) with a single factor: The gap between Labour & Capital grows exponentially and never recovers.
Such a scenario would eventually require a total reimagination of the global economy.
But that’s admittedly anticlimactic—at least in essay format. So let’s push the hypothetical envelope, and think about some more exciting ways to meet our economic demise... with style.
There is a scenario where generative AI becomes so capable that it’s actually more efficient for end-users to literally create their own products, content, customizations and implementations on the fly, largely bypassing any human service providers in the middle.
Think of it as a ‘Digital Replicator’ (a la Star Trek), where we simply make a few inputs into the Universal AI Constructor platform, which then spits out an end product.
Think: “UniversalGPT, create a locally deployable enterprise CRM with all of Salesforce’s same functionality, but where it also does [insert customizations here] and seamlessly connects with [your existing stack].”
Or: “UniversalGPT, take [fiction novel] and rewrite the ending so that it ends with a final twist, in the style of The Usual Suspects. Then turn it into a movie starring Tom Cruise, Nicole Kidman & Danny Devito. Also, write in a humorous cameo appearance of Steve Buscemi. And Shrek.”
I’m being ludicrous, but that’s intentional. Copyright and IP laws notwithstanding, if the current rate of progress continues on trend, this sort of capability might become commonplace this decade.
Now, throw in interoperability with real-world automation, 3D printing, etc. and you basically have a nuclear weapon that exclusively targets jobs.
A primitive example of this may already be playing out at Buzzfeed, whose CEO (Jonah Peretti) recently penned a piece called Our Way Forward. In the letter, among other things, Peretti explains their intentions to downsize their human workforce, and lean heavily on GPT-4 as part of their new content production strategy.
However, what struck me the most was Peretti’s example of AI being deployed as an involvement device, letting their users customize Buzzfeed content with a personalized output specific to them.
“UniversalGPT, write me a funny article about cats, where the 7th item in particular will shock me.”
Of all the prototypes currently emerging, fully autonomous AI is one area that currently feels the most like Pandora’s Box. Recent examples are Baby AGI and AutoGPT.
Essentially, these are applications that can put LLM’s like GPT-4 into a recursive self-improvement loop to perpetually try and fulfill an ongoing objective, or endlessly attempt to optimize for some sort of outcome.
At first glance, this is just another geeky toy to play with until next week’s shiny object surfaces; basically a novel take on robotic process automation that plugs into GPT-4.
Until you stop to think about its implications.
If these recursive loops are trained correctly, and fine-tuned around a core objective that’s meaningful, AI-actionable and AI-measurable... the impacts could be staggering, and far-reaching, because you’re letting a machine that runs at the speed of light make continual micro-decisions that cumulatively add up to a substantial, ongoing improvement.
As a simple near-term example, let’s say you run a website that makes money from display ads. The longer someone stays on a page, and the more pages they view, the more revenue your site earns per user.
It seems pretty reasonable, then, to install a sort of “Baby AGI for Websites” on your blog, and instruct it to constantly optimize every single one of your 1,000+ pages (by testing fonts, images, colors, copy, layout, etc.) to increase time-on-page and avg. pageviews per user.
It’s an obvious, low-hanging win. But what this inevitably leads to is handing over more and more decision-making autonomy, for increasingly critical areas of your business (or your personal life).
Now, what happens if you start letting an AGI literally build every major area of your business for you?
Where it constantly monitors your competition & parallel industries, picking up new ideas to test; where it’s constantly testing & improving on all aspects of your marketing & customer journey; where it’s constantly in communication with customers, solving problems autonomously wherever possible; and where it’s constantly improving / sourcing / creating better products to sell...
There comes a point where your entire business is just a big web of recursive loops that needs some occasional babysitting by a handful of human technicians.
And I fail to see how this doesn’t end up eating most of the org chart across virtually every type of digital-first business, in due time.
In 2013, a dystopian sci-fi movie starring Matt Damon called ‘Elysium’ hit theaters. It was a commercial flop, but many (including yours truly) see it as a diamond in the rough. That said, I’d be the first to agree that the premise of the film is more interesting than its plot...
The year is 2154, Earth has been ravaged by climate change and overpopulation, and all major cities have been reduced to apocalyptic shanty-towns and work camps.
The vast majority of the planet lives in third-world conditions; basic subsistence, with no chance of social mobility.
However, a small fraction of the world’s population—the elite, ruling class—lives on a large, Eden-like habitat called Elysium, orbiting the planet. Everyone onboard lives a life of opulence and privilege. They’re all in perfect health, thanks to advanced medical technology that can completely restore the human body from any injury or disease... but only if they’re an Elysian.
The movie is essentially a social commentary that hypothesizes how the final form of extreme wealth inequality might play out.
The scary part is that, Eden orbitals aside, given the rate at which AI might truly destroy the demand for Labour if things like Universal Constructors and AGI become widely adopted, an Elysium level of wealth inequality is actually plausible… if not likely.
Most of the world’s wealth would effectively flow directly to these AI platforms and their shareholders—just as it has with today’s internet gatekeepers like Google, Meta & Apple. The main difference being that the AI consolidation would be far more extreme.
If we’re not careful, the global economy could unintentionally end up orbiting around a single, monolithic entity, breaking everything else in the process.
Ultimately in this scenario, AI will blow past all but the most absurd of its expectations. Initially, this will seem impressive and exciting, likely causing a series of bubbles and busts, as has happened during every other technological revolution in the past.
But it won’t take long for us to feel the other edge of the sword, as ever more territory of The Org Chart is subsumed by automation, and at a much faster rate than the displaced masses can make their way back into the digital economy.
Inventions that have ended up changing the course of history were rarely intended to do so. In fact, usually the original purpose of the world’s most important innovations has been far smaller in scope than their ultimate impact.
Johannes Gutenberg’s printing press was no exception.
Following a financial misadventure in trying to sell polished metal mirrors, Gutenberg was desperate to recoup his losses and get back on his feet. From what we understand, the printing press was his last-ditch attempt to create thousands of religious texts to be sold as indulgences.
Whatever the original intention, his invention would go on to catalyze a paradigm shift across Europe, and the world in general.
By democratizing and decentralizing human knowledge—while simultaneously bringing down the cost of books to a small fraction of their former cost—the printing press made it possible for the masses to become literate, as well as for individuals to distribute ideas at a scale never before possible.
The rest is, literally, history.
But what made the printing press so influential wasn’t the machine itself. It was the stories it was able to distribute at scale; and what those stories said, and who said them.
For example, in just a few decades following its invention, a German priest named Martin Luther would lead a religious rebellion of sorts, powered in large part by the ability to print and distribute his thesis and other materials widely.
In turn, Lutheran’s ‘Protestant Reformation’ would end up fracturing the Catholic Church, and ultimately reshaping the balance of political power across Europe.
AI is similar to the printing press in that we can all recognize that this is a game-changer, but we don’t yet understand the full extent of these changes, nor what will be most impacted.
As with Gutenberg’s press, AI democratizes craft, process and the ability to produce things: content, software, designs, and so on.
But ultimately, it can only mimic us. It can’t (yet) replace strategy, intuition, purposeful disrespect, authentic personality or nuanced judgment. Human guidance is still critical in making something that matters.
So, ironically, our own humanity is the fuel required to run the machine.
For better or worse.
%20162x162.jpg)
%201%20(1).svg){kind=logo}
.svg)
.png)
%20162x162.jpg)
.png)
