In an AGI world, where are the customers of the people doing these niche jobs getting the resources to buy these products and services? Today it's people doing other, more conventionally productive work, spending some of their surplus. Is it (metaphorically) just the watchmakers who can afford to go to Taylor Swift concerts?
Turns out that this is today's SMBC comic. Which gets extra points points for the line "Humans are a group-level psychiatric catastrophe!"
Hence, I would argue that "AGI-proof" jobs are unlikely to ever provide an income basis for a significant share of the human population. To echo economist Daniel Susskind: If we think about “post-labor” economics we should not think about “a world without any work at all”, but rather “a world without enough work for everyone to do”.
And critically, even in these sorts of scenarios, the permanently unemployed humans will still be in the range of 90-99%, and I generally expect 80% as a minimum amount of permanently unemployed humans, but this post does explain well why some humans will still have jobs, and this is a fair critique of AIs making everyone losing their jobs.
this is a fair critique of AIs making everyone losing their jobs.
I have never heard anybody push the claim that there wouldn't be niche prestige jobs that got their whole value from being done by humans, so what's actually being critiqued?
... although there is some question about whether that kind of thing can actually sustain the existence of a meaningful money economy (in which humans are participants, anyway). It's especially subject to question in a world being run by ASIs that may not be inclined to permit it for one reason or another. It's hard to charge for something when your customers aren't dealing in money.
It also seems like most of the jobs that might be preserved are nonessential. Not sure what that means.
There were definitely people arguing seriously that people wouldn't do jobs at all in the age of AI, but I think that the original argument doesn't matter, as something close enough to it is true, and has most of the force of the original.
On this:
It also seems like most of the jobs that might be preserved are nonessential. Not sure what that means.
I think the correct word is that people don't have to work to get most of what they want, and that the jobs don't have to be done in the sense that if no one worked, society would essentially keep on going.
Hence, I would argue that "AGI-proof" jobs are unlikely to ever provide an income basis for a significant share of the human population.
For the categories of AGI-proof jobs that you discuss, I agree (and I much enjoyed your detailed exposition of some examples). However, in my post that you very you kindly cite, there is one AGI-proof job category that could be an exception to that, if there turned out to be sufficient demand from the AI side of the economy, my category 3:
"Giving human feedback/input/supervision to/of AI/robotic work/models/training data, in order to improve, check, or confirm its quality."
Given the progress being made in synthetic training data, and that the AIs then being trained are likely to be far smarter than any human, the demand for this as training could drop, or increase, fairly rapidly. However, if we're not actually extinct, presumably that means we solved the alignment problem, in which case AIs will be extremely interested in human values, and the only source of original new data about human values is from humans. So this is the one product that aligned AIs need that only humans can produce — and any human can produce it, not just a skilled expert. If AI demand for this was high enough, it could maintain full human employment, with basically everyone filling our surveys and being parts of focus groups, or whatever.
My guess is that this category of job is likely not to exist, both for alignment generalizing further than capabilities reasons, and the fact that I think the synthetic data pipeline for alignment is plausibly automatable as well, so I consider it a short-term solution at best.
I am very unsure about this category of job: I can see reasons for the demand for it to grow exponentially, or to go away, or to be perennial. Which of these effects dominates is on the far side of a Singularity. My prior here is to just use the uniform prior: any of these things could happen.
I had a non-uniform prior already on the category of job you're suggesting, because I had learned things that were relevant to the category of job that made me update away from it being a full-time job for most humans.
If the value of Taylor Swift concerts comes mostly from interactions between the fans, is Swift herself essential to it?
At most such concerts, the actual artist is so distant from the vast majority of the audience that you're watching on a screen and listening through speakers anyway. People show up because the artist is there, and the artist can react to the audience as a whole or to people up front, but that's about it.
The post What Other Lines of Work are Safe from AI Automation? from Roger Dearnaley examined candidate job categories "for which being an actual real human is a prerequisite". This post expands on the examples of Swiss watches, chess, and Taylor Swift in a slightly more narrative way and particularly highlights the curious economic logic of Veblen goods.
Switzerland is the global capital of mechanical watches. High-end mechanical watches are masterpieces of engineering and craftsmanship. There is no mass production of high-end mechanical watches, instead they require meticulous adjustment by skilled human watchmakers that assemble them individually.
A high-end mechanical watch can achieve remarkable precision. However, the nature of their mechanical components means they are subject to slight variations in timekeeping due to wear, temperature fluctuations, and positional changes. High-end mechanical watches that are certified as chronometers by official bodies like the COSC (“Contrôle Officiel Suisse des Chronomètres”) need to typically meet the accuracy standard of -4/+6 seconds per day, which translates to a variance of up to 3 minutes per month. Really high-end mechanical watches may have more stringent requirements. For example, the typical accuracy of Patek Philippe watches is within the range of -3/+2 seconds per day, or about 1 minute per month.
This is impressive, but if the accuracy of time-keeping were the main function of a Patek Philippe it would not be able to compete. In 1967 the first quartz wristwatch was developed. When an electric current is applied to the quartz, it vibrates at a precise frequency of 32,768 times per second. This high-frequency oscillation allows the watch to keep very accurate time, typically losing or gaining only a few seconds per month. Furthermore, quartz watches are much less complex than mechanical watches. Most parts can be produced and assembled with minimal human intervention.
Hence, you can buy a quartz wristwatch for as little as 10 USD, and it is an order of magnitude more accurate than a Patek Philippe Grandmaster Chime Ref. 6300A-010, sold for 31 million USD. Indeed, you probably have a quartz watch in your pockets, whether you know it or not. The internal real-time clock in smartphones is usually based on quartz. On top of that, smartphones connected to cellular networks and/or connected to the Internet can synchronise their time with network-provided time signals, often coming from ultra-precise atomic clocks. So, your smartphone watch is both easier to read and significantly more accurate than a mechanical watch. Yet, people still pay millions for hand-assembled mechanical watches. What is going on here?
High-end mechanical watches fall into a category called “Veblen goods”, a type of luxury good for which an individual's demand increases as the price increases. This is an apparent contradiction to the law of demand. Whereas regular products compete on quality and affordability, high prices are not a weakness but an asset for Veblen goods. This peculiar demand curve only holds within a price range that depends on the wealth of the individual. Veblen goods have three main characteristics:
So, artisanal luxury goods with artificial scarcity and a social signalling function are one way in which some human labor could persist despite being economically obsolete.
Chess has long been a symbol of human intellect and strategic thinking, making it a natural challenge for AI researchers:
Yet, despite the overwhelming dominance of AI in chess, human interest in human chess play remains stronger than ever. The game has seen a resurgence in popularity, triggered in part by the Netflix series “The Queen’s Gambit”, and driven by new online chess play platforms such as “chess.com” that allow for short chess games anywhere, and charismatic chess YouTubers. The best chess players in the world are all AI programs. However, humans are barely interested in watching Stockfish play AlphaZero. Instead they are interested in the competition between and connection with human chess players, such as Levy Rozman, Hikaru Nakamura, the Botez sisters, Magnus Carlsen, or Anna Cramling.
Machine superiority in chess and other cognitive sports is a recent phenomenon. However, if we think about it, human muscles have been outmatched by artificial energy for a long time. Humans are not competing at the frontier of what is technologically possible in most olympic disciplines:
And yet, billions of humans tune in to watch the Olympic Games, making it one of the most-watched events worldwide. So, our interest in sports does not depend as much on the absolute performance level as it depends on the human conflict, connection, spirit, perseverance, and the stories that we weave around them. Physical and mental human competitions from chess, to the Olympics, to the paralympics, to the Spelling Bee, to the Mental Calculation World Cup, to high-speed telegraphy, can all continue indefinitely under machine superiority.
In the early 20th century, the advent of recorded music sparked concerns that it would diminish the role of live musicians. The renowned composer John Philip Sousa warned that “the country band with its energetic renditions, its loyal support by local merchants, its benefit concerts, band wagon, gay uniforms, state tournaments, and the attendant pride and gayety, is apparently doomed to vanish in the general assault on personality in music.”
Yet, here we are more than a century later and Taylor Swift’s “Eras Tour” has become the first tour in history to gross over 1 billion USD. So, what brings so many to pay so much to see Taylor Swift live? It cannot be the sound quality. If you want to enjoy Taylor Swift’s songs with the highest possible audio quality, you will not find this at her concerts but at home with high-end headphones. Instead her tour seems to be a cultural phenomenon, where her mostly female fans that call themselves "Swifties" engage in elaborate preparations for the shows, including building and exchanging friendship bracelets, wearing themed outfits and sharing their experiences on social media.
Humans are social animals and like most other animals it seems that we have a special interest in other beings of our own kind. I have no doubt that anthropomorphized AIs pretending to be the friends and even romantic partners of humans will compete with content producers and influencers for parasocial relationships. Still, there is something around “authentic experiences” that puts a premium on the value of real, live human interactions and the unique atmosphere created by being physically present at an event or performance. At least for now, some emotional and social aspects of live in-person events cannot be easily replicated, such as multi-sensory experiences, unpredictability, and the ability to meet, network, and share the experience with others in-person.
The best view of a prestigious sports match is not in the stadium but at home, where you can see any goal replayed from 10 different angles. Still, many are willing to pay thousands of USD for the in-person experience. So, there is something around this that might persist in a machine-dominated economy.
The long-term persistence of some forms of human labor for artisanal luxury goods, elite sports and in-person high-end events even in a “post-labor” economy is worth noting. However, we should also acknowledge that Veblen goods only represent a very small fraction of the current world economy. For example, the personal luxury goods market, which is still significantly broader defined than Veblen goods, had a volume of about 362 billion $ in 2023, that’s roughly 0.3% of the global GDP (ca. 105 trillion $ in 2023). Veblen goods may have some room to grow. However, it would be an inherent contradiction for them to represent a significant fraction of the economy. Their entire point, and the reason why they can escape the regular logic of economics, is that they are defined by scarcity and exclusivity. Similarly, not everyone can be a chess or music superstar with a million human fans.
Hence, I would argue that "AGI-proof" jobs are unlikely to ever provide an income basis for a significant share of the human population. To echo economist Daniel Susskind[2]: If we think about “post-labor” economics we should not think about “a world without any work at all”, but rather “a world without enough work for everyone to do”.
When European explorers first encountered pineapples in the New World, they were astounded by its unique taste and appearance. The challenges of transporting pineapples over long distances without spoiling further increased their allure. So, in the 17th and 18th centuries, the pineapple became a key status symbol among European nobility. It was depicted in paintings, featured in the architecture of grand estates, and so expensive that the rich would rent pineapples to show them off at their lavish banquets without the full expense of ownership. Then came mass production, which transformed the pineapple from an exclusive luxury to an everyday commodity. So, the pineapple went from being the “fruit of the Gods” worshipped by nobility to being so ubiquitous that common “peasants” on the Internet insult it as not being worthy as their Pizza topping.
Daniel Susskind. (2020). A World Without Work: Technology, Automation, and How We Should Respond. Henry Holt and Co. pp. 5&6