Without speculating here on how likely this is to happen, suppose that GPT-4 (or some other LLM or AI) speeds up, streamlines or improves quite a lot of things. What then?
The Dilemma
Samo and Ben’s dilemma: To the extent that the economy is dominated by make-work, automating it away won’t work because more make-work will be created, and any automated real work gets replaced by new make-work.
Consider homework assignments. ChatGPT lets students skip make-work. System responds by modifying conditions to force students to return to make-work. NYC schools banned ChatGPT.
Consider a bullshit office job. You send emails and make calls and take meetings and network to support inter-managerial struggles and fulfill paperwork requirements and perform class signaling to make clients and partners feel appreciated. You were hired in part to fill out someone’s budget. ChatGPT lets you compose your emails faster. They (who are they?) assign you to more in person meetings and have you make more phone calls and ramp up paperwork requirements.
The point of a bullshit job is to be a bullshit job.
There is a theory that states that if you automate away a bullshit job, it will be instantly replaced by something even more bizarre and inexplicable.
There is another theory that states this has already happened.
Automating a real job can even replace it with a bullshit job.
This argument applies beyond automation. It is a full Malthusian economic trap: Nothing can increase real productivity.
Bullshit eats all.
Eventually.
Two Models of the Growth of Bullshit
- Samo’s Law of Bullshit: Bullshit rapidly expands to fill the slack available.
- Law of Marginal Bullshit: There is consistent pressure in favor of marginally more bullshit. Resistance is inversely proportional to slack.
In both cases, the lack of slack eventually collapses the system.
In the second model, increased productivity buys time, and can do so indefinitely.
Notice how good economic growth feels to people. This is strong evidence for lags, and for the ability of growth and good times to outpace the problems.
A Theory of Escape Velocity
We escaped the original Malthusian trap with the Industrial Revolution, expanding capacity faster than the population could grow. A sufficient lead altered underlying conditions to the point where we should worry more about declining population than rising population in most places.
Consider the same scenario for a potential AI Revolution via GPT-4.
Presume GPT-4 allows partial or complete automation of a large percentage of existing bullshit jobs. What happens?
My model says this depends on the speed of adaptation.
Shoveling Bullshit
Can improvements outpace the bullshit growth rate?
A gradual change over decades likely gets eaten up by gradual ramping up of requirements and regulations. A change that happens overnight is more interesting.
How fast can bullshit requirements adapt?
The nightmare is ‘instantaneously.’ A famous disputed claim is that the NRC defined a ‘safe’ nuclear power plant as one no cheaper than alternative plants. Cheaper meant you could afford to Do More Safety. Advancements are useless.
Most regulatory rules are not like that. Suppose the IRS requires 100 pages of paperwork per employee. This used to take 10 hours. Now with GPT-4, as a thought experiment, let’s say it takes 1 hour.
The long run result might be 500 pages of more complicated paperwork that takes 10 hours even with GPT-4, while accomplishing nothing. That still will take time. It is not so easy or fast to come up with 400 more pages. I’d assume that would take at least a decade. It likely would need to wait until widespread adaptation of AI powered tools, or it would bury those without them.
Meanwhile, GPT-5 comes out. Gains compound. It seems highly plausible this can outpace the paperwork monster.
This applies generally to places where a specified technical requirement, or paperwork, is needed. Or in places where otherwise the task is well-specified and graded on a pass/fail basis. Yes, the bar can and will be raised. No, if AI delivers the goods in full, Power’s requirements can’t and won’t be raised high enough or fast enough to keep pace.
Replacing Bullshit
If the bullshit and make-work needs to keep pace, it has options.
- Ban or regulate the AI, or use of the AI.
- Find different bullshit requirements that the AI can’t automate.
- Impose relative bullshit requirements, as in the nuclear power case.
Option 1 does not seem promising. Overall AI access likely can’t be policed.
Option 3 works in some situations and not others, as considered below.
Option 2 seems promising. Would likely be in person. Phones won’t work.
New in person face time bullshit tasks could replace old bullshit tasks. This ensures bullshit is performed, bullshit jobs are maintained, costly signals are measured and intentionally imposed frictions are preserved.
I expect this would increasingly be the primary way we impose relative bullshit requirements. When there is a relative requirement, things can’t improve. Making positional goods generally more efficient does not work.
Same goes for intentional cost impositions. Costs imposed in person are much harder to pay via AI.
Thus, such costs move more directly towards pure deadweight losses.
When things are not competitive, intentional or positional, I would not expect requirements to ramp up quickly enough to keep pace. Where this is attempted, the gap between the bullshit crippled versions and the freed versions will be very large. Legal coercion would be required, and might not work. If escape is achieved briefly, it will be hard to put that genie back in the bottle.
One tactic will be to restrict use of AI to duplicate work to those licensed to do the work. This will be partly effective at slowing down such work, but the work of professionals will still accelerate, shrinking the pool of such professionals will be a slow process at best, and it is hard to restrict people doing the job for themselves where AI enables that.
Practicalities, plausibility and the story behind requirements all matter. Saying ‘humans prefer interacting with humans’ is not good enough, as callers to tech support know well. Only elite service and competition can pull off these levels of inefficiency.
Passing Versus Winning
It will get easier to pass a class or task the AI can help automate, unless the barrier for passing can be raised via introducing newly required bullshit in ways that stick.
Notice that the main thing you do to pass in school is to show up and watch your life end one minute at a time. Expect more of that, in more places.
It won’t get easier to be head of the class. To be the best.
Harvard is going to take the same number of students. If the ability of applicants to look good is supercharged for everyone, what happens? Some aspects are screened off, making others more important, requiring more red queen’s races. Other aspects have standards that go way up to compensate for the new technology.
Does this make students invest more or less time in the whole process? If returns to time in some places declines, less time gets invested in those places. Then there are clear time sinks that would remain, like putting in more volunteer hours, to eat up any slack. My guess is no large change.
What about a local university? What if the concern is ‘are you good enough?’ not ‘are you the best?’ If it now takes less human time to get close enough to the best application one can offer, this could indeed be highly welfare improving. The expectations and requirements for students will rise, but not enough to keep pace.
Attending the local university could get worse. If what cannot be faked is physical time in the classroom, such requirements will become increasingly obnoxious, and increasingly verified.
The same applies to bullshit jobs. For those stuck with such jobs, ninety percent of life might again be showing up. By making much remote work too easy, it risks ceasing to do its real task.
Speed Premium
It comes down to: If they do happen, can the shifts described above happen fast enough, before they are seen as absurd, the alternative models become too fully developed and acclimated to to be shut down and growth becomes self-sustaining?
If this all happens at the speed its advocates claim, then the answer is clearly yes.
Do I believe it? I mostly want to keep that question outside scope, but my core answer so far, based on my own experiences and models, is no. I am deeply skeptical of those claims, especially for the speed thereof. Nostalgebraist’s post here illustrates a lot of the problems. Also see this thread.
Still, I can’t rule out thing developing fast enough. We shall see.
Don't Worry About the Vase 
I am completely and totally confused by both this and the school post. Have you made the argument elsewhere that bullshit jobs actually exist? In my experience I see many people who don’t understand their job’s value but hardly any that are actually valueless.
Similarly, with school, despite the occasional autodidact, I see a lot of value to most students, enough that I’m convinced there’s a baby in that bathwater.
I understand that you can’t define your terms every post, but at the moment your assumptions are so far removed from my experience that I feel I might as well be debating you about your Thetan level.
Agree that bullshit jobs are not something I have run into in business. I don’t think my >5k employee company has any, and I have never run into anyone in one while visiting clients.
Disagree about school. “Go be an apprentice/do physical labor” would be better for most boys who can’t be autodidacts. “Help your mother and prepare to be a good wife” would be better for 95+% of girls. There’s a pretty narrow band of midwits that school makes sense for.
I mean, there’s parasite jobs, where an outside movement is aggressively trying to subvert your organization by creating positions that solely try to create more parasite positions. But that’s not the same thing.
Elon Musk took over twitter and fired >90% of staff (exact figures differ) and the website works as well as ever, and is actually coming out with more service innovation than it has in the year prior. At least some of them must have been “strategically aligning priorities” or “special projects” or “creating vision statements” or whatever Twitter called its useless tasks.
The argument for bullshit jobs is pretty precise in the actual book, but firing most of Twitter will still be the canonical demonstration.
Hell, Galileo had great arguments but he had to tell the leaning tower story to have street cred. Same here.
It is way too early to tell whether Elon Twitter is a success. Take all of Zvi’s Bounded Distrust post and turn it up to 11: many of those involved do consider this worth a One Time. Maybe if, in a year, they are paying all their bills without further cash injection, and haven’t hired anyone to replace those fired, that would be evidence for your position being true about private equity targets.
What value do most students get out of school (particularly homework)?
– Learning. Most adults can read, largely can do basic math (but see https://www.youtube.com/watch?v=6egeUxIEQnM), aaand that’s about it. The set of knowledge & skills ubiquitous among adults is an upper limit on the usefulness of schools.
– Daycare. Yes, this is what schools are for, but it doesn’t imply homework.
– Training personality traits. No, shared environment effects are a bust.
– Filtering on preexisting personality traits, including performance under pressure sustained for a few months. Although a version of the test only a few months long contains most of the information-value of a longer version of the same, unfortunately the willingness to participate in the test for a longer duration is correlated with performance on the test itself.
Reading and math are actually really important! I also believe that school provides a lot more opportunity for learning even on subjects that aren’t universally successful; I wouldn’t have ever taught myself to write without the external discipline, and an engineer who can’t write is actually pretty useless.
I won’t argue that schooling has been optimized, there are definitely aspects that could be improved. But the vibe I get from zvi is abolition, not reform.
Yes, literacy (I shortened “reading and writing” to “reading”) and math are valuable skills. Some instruction in them is even valuable-over-alternative. But instruction in them has negligible marginal benefit — less than the marginal opportunity cost — for most of the 12 years they are taught. A large fraction of children would learn them without any school (hence “upper limit”); it is only modestly uncommon for entering 1st graders to already be semi-literate! (And a smallish fraction of students leaves high school semi-literate, a.k.a. functionally illiterate.) Same for math. Doing nothing but cutting school down to, say, two years, would keep most of its benefits (in terms of learning) while greatly reducing its costs.
My current guess is that the crux concept is “normative” and “social model of disability”. Reverse-designing schools, a notional design assumption of a quite narrow window of anticipated personality trait combinations becomes visible. Which window would not only be too narrow even in the best case, it is not even particularly “centered” on the actual distribution. Thus a large proportion of children, with most of whom “there is nothing wrong” by other standards — they will be perfectly functional adults, they are in a densely populated part of the distribution, close to the overall center of mass — are outside the window by some distance. This can be conceptualized in a number of ways:
– “social model of disability”: school is “designed” wrong, burn it down and build something better on the ashes;
– “medical model”: they have ADHD, here’s some Adderall;
– “skill model”: try to invent or look for practical tips and skills to alleviate the problem;
– “philosophy model”: yes, school is mostly pointless, if you are able to, put your emotions elsewhere and treat it as a day job. Nothing personal. Feel free to do thought-experiments with the self-concept.
Reading this is like watching the first scouts butt up against an Outside Context Problem.
There’s a relevant historical pattern, and I’d say burden of proof is on people to demonstrate that this *won’t* happen:
A bizarre civilization level pressure grows and grows and grows and grows and seems like it will never stop growing until it leads to a cataclysmic war, and after the war, raw power writes a new peace with new terms.
So: dismiss, discuss, or defend against: the bullshit jobs economy will not be eternal, but somewhere between the bullshit jobs economy of today and the no bullshit jobs economy of the far future, there will be a cataclysmic war about the bullshit jobs economy.
In particular, “No, if AI delivers the goods in full, Power’s requirements can’t and won’t be raised high enough or fast enough to keep pace.” isn’t good for stability. Power wouldn’t react to that gap by saying “oh, we’ll just let this gap exist while we slowly backfill it”. Neither would Would-Be-Power. You’re talking about creating a huge vacuum, which has an extremely predictable consequence.
I think the no bullshit jobs economy is beyond the
event horizonSingularity. Humans don’t work like that and never did (transhumans might).If Zvi and you don’t mind the comparison, a large part of this reminds me of 19th c. socialism/communism. After the Revolution, there will be no:
– alienation (people not finding their work meaningful, because humans are adapted to the Stone Age);
– people trying to screw each other over while ostensibly cooperating (adversarial contract negotiation, suit for breach of contract, theft, laziness, “fusi” work (using company equipment, and sometimes materials, for a side gig), NIMBYism, — and after companies grew large enough, departments squabbling);
– otherwise real work that turns out to have been blatantly wasted because the context changed;
– largely necessary work of getting a large number of people to do a large number of different tasks in a sequence that accomplishes a Thing, especially if for most of them it isn’t routine to do that Thing (people can build a house without an entrepreneur if they already built several other houses) (I mention this because socialists notoriously miss(ed) this concept);
– pink-collar work and its extensions into various directions;
– school homework, which none of the above justifies. Some stuff is pointless. (Or rather, I would say politicians and “educators” imagine(d) that after the Revolution/Reforms, all children would enthusiastically do their homework, be proficient in everything taught, and generally above average.) It’s tempting to put this under the expanded-pink-collar and/or adversarial categories, but no, most teachers don’t assign homework just to watch students squirm. Even boot camp drill sergeants mostly don’t do that, despite it serving an actual purpose there, namely to forge the recruits into a cohesive unit.
“If the bullshit and make-work needs to keep pace, it has options.”
4. Use the AI to generate more make-work rapidly.
It isn’t obvious why GPT preferentially replaces bullshit jobs – writing CRUD apps as line-of-business software for non-tech companies is a real job that is GPT-vulnerable. So is low-tier graphic design like book covers.
Graeber’s original categorisation of bullshit jobs was (per Wikipedia):
-flunkies, who serve to make their superiors feel important, e.g., receptionists, administrative assistants, door attendants, store greeters, makers of websites whose sites neglect ease of use and speed for looks;
-goons, who act to harm or deceive others on behalf of their employer, or to prevent other goons from doing so, e.g., lobbyists, corporate lawyers, telemarketers, public relations specialists, community managers;
-duct tapers, who temporarily fix problems that could be fixed permanently, e.g., programmers repairing bloated code, airline desk staff who calm passengers whose bags do not arrive;
-box tickers, who create the appearance that something useful is being done when it is not, e.g., survey administrators, in-house magazine journalists, corporate compliance officers, quality service managers;
-taskmasters, who create extra work for those who do not need it, e.g., middle management, leadership professionals.
Flunkies can’t be replaced by GPT – the whole point is for a human being to perform subordination. Goons and duct tapers are doing a wide variety of real work that shouldn’t be done – I don’t see why they would be any more replaceable than people doing the same work when it should be done. Only box tickers are unusually GPT-replaceable.
I think the issue is that GPT doesn’t replace bullshit jobs. Most of what GPT replaces is real jobs producing real bullshit in response to a real demand for bullshit – which is actually the least central example of a bullshit job.
Maybe Graeber was the real bullshit all along.
“Box tickers” are the closest thing on that list to a bullshit job. The others are either “person doing real job poorly” (most of the “box tickers” fall into that category too honestly), “zero/negative sum warfare in the firm’s interests”, or even just outright value-adding jobs.
> “Consider a bullshit office job. […] perform class signaling to make clients and partners feel appreciated.”
If this weren’t an unwritten (unsaid (illegal and/or saying-it-would-defeat-the-point-of-having-it)) aspect of the job, but officially the central part of the job†, would it still count as a bullshit job, or would it be a partly or perfectly legitimate production of a positional good? In other words, in your view how bullshit a job is “prostitute”? “Trophy wife”?
†: “job” loosely interpreted on the Monarchy Inc. model: there may not be a clear line between “long-term hired entertainer (paid from the collective budget managed by the lord)” and “part of the retinue for reasons unrelated to its ostensible purpose”.
Prostitute is not a bullshit job – you are providing something your customers value inherently, whether or not you think providing it is a good idea.
Trophy wife I think can go either way, depending on details.
These “bullshit job” claims are frustrating because we have actual big examples of bullshit make-work that cause so much suffering, and they are school/government/nonprofits. People, do not project the crimes of your own social class on the nice clean industries we escape you by entering.
There was a village, and in this village was a blacksmith and apprentice. Every year, just before spring, there would be a great work undertaken, to get all of the villager’s tools ready; and throughout spring, they would toil to repair the tools that had broken; the blacksmith doing the labor, and the apprentice working the bellows.
Come summer, as the great work lulled, the blacksmith would have the apprentice hammer out nails; but as fall came again, once more they would turn to the work of tools. Winter again had the apprentice making nails.
Time passes, as it tends to do, and the apprentice became the blacksmith, and took on an apprentice. And that apprentice would work on tools in the spring and fall – but the blacksmith couldn’t figure out what exactly the prior master had been doing with all the nails, for the apprentice made far more nails than the village required – and so instead the blacksmith let the apprentice have time off in the summer and winter.
Time yet passed, but when it came time to pass the hammer, the blacksmith was displeased, for the apprentice never developed a true mastery of the craft.
————————————————-
James took over Nail Corp after the passing of his father; the first thing he did was look around at all the employees not doing productive labor, instead filling their time with pointless busywork. So he took away the busywork, and, observing how many employees now sat idle, fired half of them, that the number of employees matched the amount of work to be done. Things went well for about half a year, until the snow thaw began, and the demand for nails surged as construction season began – and James was unable to meet the demand with his reduced workforce, and lost many of his contracts.
He entered a downward spiral, firing employees to meet the demand, and then losing business, such that the demand fell further. His company eventually went bankrupt.
————————————————-
Jamie took over Hammer Corp after the passing of her mother; the first thing she did was look around at all the employees not doing productive labor, instead filling their time with pointless busywork. So she took away the busywork, and, observing how many employees now sat idle, cut the workday in half. Things went well for about half a year, until the snow thaw began, and the demand for hammers surged as construction season began – and she called in her workforce to work full days for the boom period. It was a rocky year, and she lost a contract or two, as her employees had reorganized their lives around the shorter workdays.
Her company never fully recovered, although it did limp along for many years until her death. Half her employees mostly remembered her for ruining family events by demanding they work longer hours during the summer.
————————————————-
John took over Plank and Board Corp after a hostile takeover. He didn’t change anything substantive, and continued making money.
I think Ben’s point is that “Bullshit job” is an exaggerated or homiletic claim.
I notice I am confused.
You speak as though “everything is wonderful” is a semi-plausible outcome of this race, which could happen if not for sclerotic processes, but TAI and X-risk AI are AFAICS essentially the same thing, and AFAIK you’re in the same “neural net alignment is not a thing” camp as I am.
So I’m not seeing how “GPT-4 makes everything wonderful” is even on the table to consider here. Either it’s TAI/XAI and it has a bloody-good go at killing us all, or it’s not TAI and it doesn’t make everything wonderful. The cake is a lie.
Yeah, it’s already kind of scarily good at programming. I don’t think we’re very far from a foom.
I always took the zero rather than copy or cheat on homework. That pressure to defect ramps way up when your classmates can knock out 20 page papers in 5 minutes and not get caught.
> There is consistent pressure in favor of marginally more bullshit. Resistance is inversely proportional to slack.
Directly proportional, yeah? The more slack you have the better your positioned to resist additional bullshit.
“better you’re positioned”. Voice to text, sigh.
No, inversely. If you, say, have a construction project with little slack i.e. a tight margin, then when NIMBYs try to extract surplus from it, you either cancel the whole thing or you fight them. However if you have a huge margin, you do the CDT thing and throw money at them to go away because you still have a large margin after paying the bribe/ransom. To the extent the scarce resource is your attention rather than money (or things like it), you have better opportunities than bullshit-fighting on which to spend it.
I can see that perspective, but I was assuming that Zvi is using ‘slack’ in the sense that he defined it some years ago (https://www.lesswrong.com/posts/yLLkWMDbC9ZNKbjDG/slack), where slack provides freedom that enables you to fight back against the pressures that would otherwise lead to the proliferation of bullshit jobs.