When Every Job Is Automated
A utopia or a dystopia?
App developer. Social media manager. Personal trainer. No person on Earth could have predicted these jobs a few decades ago. These are but a few odd jobs that exist today, which have replaced jobs of the past. Throughout modern civilization, jobs have come and gone. Not too many decades ago, about ten perhaps, Americans and Europeans predominantly worked in agriculture. Today, however, mere percentages do. Yes, with the birth of new machinery that allowed one lone person to accomplish what previously required a plethora of coworkers, came the death of agricultural jobs in industrialized countries. In their stead came new jobs.
This has been a consistent pattern throughout the history of modern civilization. As old jobs become obsolete to technology, new jobs appear to rise in their stead.
Despite this historical trend, fears of jobs disappearing have nevertheless reemerged with the surge of popularity for machine learning. Indeed, does this pattern still exist? Are we still seeing a surplus of jobs? Yes, says many studies like this Gartner study and this Oxford paper.
For now. These studies mostly focus on the near future into consideration.
What makes AI unique is its theoretical ability to automate any job, no matter how cognitively challenging. In whatever role that you can find clear routines and rules, a machine learning AI could more efficiently accomplish the work than a human. It should not be unfeasible to imagine a potential future wherein virtually every job, spare a few decision-making leadership roles, are automated.
In this highly subjective and hypothetical article, I’d like to argue why and how we might reach that point, and what comes after. For that reason, I will be jumping over some of the dilemmas of the present day. If you’re curious about how value creation works in AI, you can read more about that here, and the severe ethical implications of these developments here.
The Death of Driving
From a process perspective, vehicular automation is what I would call the perfect case for automation. Autonomous AI thrives in environments with clearly defined rules, and there sure are plenty of rules in traffic. As there is a correct course of action for every conceivable scenario, machines are blatantly superior to humans when it comes to driving vehicles. Self-driving cars do not get tired. They do not drive over the speed limit. They drive energy-efficiently. They drive smoothly and comfortably for the passengers. And to top it all of, they don’t ask for a salary.
The technology for completely autonomous vehicles is almost complete. Elon Musk, CEO of Tesla, believes that their company will have fully autonomous cars by the end of this year, which would be 2020, in case you’ve arrived at this article from the future. After that, it will take a few years for the legal and judicial landscapes to pave the way for the technology, though inevitably, machines will be allowed to drive fully autonomously. I believe that machines will replace all jobs in industrialized countries that revolve around driving. This is a substantial amount of jobs. Truck drivers, bus drivers, taxi drivers, train conductors. AI is simply superior in these tasks.
Again, what makes driving unique is that, while the processing of the data might be complex (i.e., the ability to observe and comprehend all manners of visual information such as road signs and other vehicles), from a process perspective, it’s rather straight forward. This is certainly not the case for many other tasks, however. Ask a banker, a programmer, or a doctor to explain in minute detail how they make every decision, every task, every objective of their day, and they will surely struggle to do so. Finding clear routines and rules for every job is challenging.
Alas, task by task, machine learning AI is slowly but surely automating even complex roles. The world’s best flu vaccine was created by an AI in 2019. When a jury was asked to vote for their favorite perfume, an overwhelming majority chose a perfume developed 100% autonomously by an AI. Music curation? Automated. Text editing? Automated. Stock investing? Automated. And all of this by 2020.
The Death of Jobs
If our planet had existed for a day, humanity would have existed for 10 minutes, the industrial era for two seconds, and machine learning for some mere millisecond. Imagine how far machine learning will have come in a decade, in two decades, in five, in ten, in a hundred decades.
All AI applications that exist today are examples of narrow AI, sometimes called modular AI. These are applications that have been built to accomplish one specific task. The AI that defeated the world’s best Go player cannot drive a car, and the AI that perfectly drives cars cannot play Go, while a human may be able to do an adequate job at both of those activities.
Imagine how far machine learning will have come in a decade, in two decades, in five, in ten, in a hundred decades.
In contrast, artificial general intelligence (AGI) is a theoretical AI that can accomplish every task that a human can do just as well as a human. This is the kind of AI typically seen in sci-fi films. If humanity does discover AGI, then most jobs could, by definition of AGI, be replaced by computers.
However, whether or not researchers discover how to build AGI matters not, I would argue, as jobs could become replaced nevertheless. You don’t need AGI to replace even highly skilled positions. In time, modular AI applications could replace job after job.
There are plenty of obstacles to overcome, however. Besides technical challenges, there are ethical dilemmas such as overreliance in machines and algorithmic discrimination, judicial and legal difficulties, the need for human connections, and more. Mass automation of jobs will take a very long time. But I do believe it to be inevitable.
The Birth of… What?
With the death of jobs comes new jobs, or so I’ve been arguing all along. But what happens at this theoretical point where (almost) every job is automated? Where machines can do everything just as well as humans, and the few jobs that haven’t been replaced, such as those of political nature, instead use AI for decision-making support?
Erik Brynjolfsson of MIT coined the phrase “digital Athens”. Two millennia ago, Athenians enjoyed democracy, arts, philosophical debates, and Olympic games. They were free to fulfill their passions and made astonishing progress in science, arts, politics, philosophy, and rhetorics as a result. This was made possible primarily because Athenians had slaves do all the tedious work. Since then, nations around the world have agreed that slavery is horrifying and certainly not something we should be doing in 2020, but what about having machines as metaphorical slaves? Much like a self-driving car could be compared to having a personal slave driver, so too could AI slaves do all of your work.
The way we tend to look at life today is that it revolves around working. We are either employed or looking for work. We spend at least eight hours a day working. Our lives are built around having jobs. It might be hard to imagine a norm wherein most people do not have jobs and are, in fact, not looking for work either.
How would the economy work if (almost) every job is automated? Machines (both physical and digital) can do the job of humans significantly cheaper. If that weren’t true, there wouldn’t be much point in creating machines in the first place. This next sentence might upset some readers, but at this point in society, corporations can be taxed higher, and a basic universal income can be established. I know that many people today may be against the idea of having higher taxes and the idea of the government giving away free salaries to citizens for doing nothing, but bear with me.
Here’s why. Machines can do the work of humans significantly cheaper. Thus, governments can tax companies greater, and corporations would still make a healthy profit. This extra money can be spent on a basic income, eliminating the need to work. In this theoretical future, I argue that this is not only a suggestion but a requirement for societies to function.
But if people don’t work, what will they do? This, I believe, is the most beautiful part. Humans would be free. Much like the Athenians were free to explore any creative arts they desired without monetary worries, so too would the modern humans be. Free to pursue happiness in any form they find fulfilling. No longer would they be a struggling artist, barely making enough to pay their bills. They would be an artist with no need to worry about making enough to pay their bills. People would be free to pursue their passions. And even if AI could also create art, songs, and films, humans would nonetheless be able to express creativity themselves. Our entire education system would be rebuilt, where creative arts would be at the top of the educational pyramid, rather than the bottom.
But if people don’t work, what will they do? This, I believe, is the most beautiful part.
All children are born creative. After their school years, their creativity has been systematically erased. In the schools of this hypothetical future, children would be taught to retain their creativity. Our way of looking at life would be altered from developing skills that will pay the bills, to developing skills that will lead to fulfillment in life.
Sounds beautiful, Jacob, but get real
Indeed, this is an optimistic view of what may happen when all jobs are automated, yet looking at societal trends in recent years will provide anything but an optimistic outlook. While the wealth of 99.9% of the population hasn’t increased much over the last decade, the wealth of the top 0.01% has exploded.
If one were to take a more cynical approach, one might ask themself why the 0.01% would care to give the other 99.9% a universal income. Yes, why even allow the majority of the population to continue to live? This idea comes from another MIT professor, this time the brilliant mind of Max Tegmark, who shared these ideas in his book Life 3.0 (2017, ISBN 978–91–88659–67–5). While discussing AGI, Tegmark argues that the wealthiest of society might simply opt to eliminate the redundant civilians who no longer contribute.
I agree; it’s a possible future. If virtually every job becomes automated, humans become unnecessary. With a twisted enough mindset, one might even argue that the mass elimination of humanity is for the good of Earth, as climate change and overpopulation would be issues no longer. The elite could wipe out the world with ease. If an AI could create the world’s best flu vaccine in 2019, there is surely a future wherein an AI can create the perfect virus. A virus that spreads incredibly fast, shows no symptoms to the bearer for the first thirty days, and then kills its bearer instantly. Naturally, the wealthiest will be genetically engineered to be immune to the virus.
This way, the 0.01% of the population could enjoy a perfect world wherein they never have to work, never have to worry about climate change, and never have to worry about another revolution sparked by unemployment.
Death or Birth?
Whether or not AGI is discovered, I believe that there will be a point of time in the future where a majority of jobs will have been automated, and not enough new jobs will have been invented. I also believe that the most likely scenario in this event is that universal basic income will become commonplace, but there are naturally plenty of other possibilities, and I wanted to provide an example of a cynical such.
Before I end this argumentative piece, I want to stress that what I have described is a potential future that likely will not exist for many decades, probably not for many centuries. I do not believe that we are anywhere close to mass automation. However, I also want to point out that technological progress is unpredictable and has proven to be generally faster than one would imagine. A single breakthrough in a single field can be a game-changer that disrupts life as we know it. Inventions such as the internet and the smartphone are mere decades old and have, in such a short time-span, fundamentally reshaped our lives and disrupted countless industries.
Whether it turns out to be a utopia or a dystopia: mass automation of jobs is coming.