This weekly roundup thread is intended for all culture war posts. 'Culture war' is vaguely defined, but it basically means controversial issues that fall along set tribal lines. Arguments over culture war issues generate a lot of heat and little light, and few deeply entrenched people ever change their minds. This thread is for voicing opinions and analyzing the state of the discussion while trying to optimize for light over heat.
Optimistically, we think that engaging with people you disagree with is worth your time, and so is being nice! Pessimistically, there are many dynamics that can lead discussions on Culture War topics to become unproductive. There's a human tendency to divide along tribal lines, praising your ingroup and vilifying your outgroup - and if you think you find it easy to criticize your ingroup, then it may be that your outgroup is not who you think it is. Extremists with opposing positions can feed off each other, highlighting each other's worst points to justify their own angry rhetoric, which becomes in turn a new example of bad behavior for the other side to highlight.
We would like to avoid these negative dynamics. Accordingly, we ask that you do not use this thread for waging the Culture War. Examples of waging the Culture War:
-
Shaming.
-
Attempting to 'build consensus' or enforce ideological conformity.
-
Making sweeping generalizations to vilify a group you dislike.
-
Recruiting for a cause.
-
Posting links that could be summarized as 'Boo outgroup!' Basically, if your content is 'Can you believe what Those People did this week?' then you should either refrain from posting, or do some very patient work to contextualize and/or steel-man the relevant viewpoint.
In general, you should argue to understand, not to win. This thread is not territory to be claimed by one group or another; indeed, the aim is to have many different viewpoints represented here. Thus, we also ask that you follow some guidelines:
-
Speak plainly. Avoid sarcasm and mockery. When disagreeing with someone, state your objections explicitly.
-
Be as precise and charitable as you can. Don't paraphrase unflatteringly.
-
Don't imply that someone said something they did not say, even if you think it follows from what they said.
-
Write like everyone is reading and you want them to be included in the discussion.
On an ad hoc basis, the mods will try to compile a list of the best posts/comments from the previous week, posted in Quality Contribution threads and archived at /r/TheThread. You may nominate a comment for this list by clicking on 'report' at the bottom of the post and typing 'Actually a quality contribution' as the report reason.
Jump in the discussion.
No email address required.
Notes -
Trillions of dollars are being spent on building datacenters for inference. Amazon software engineers are inventing bullshit work for AI to inflate their internal usage scores.
I’m no expert, but isn’t there a fatal flaw here? Most of the work LLM inference is used for is essentially busywork that wouldn’t exist in an automated economy. It’s writing emails, it’s code reviews, it’s asking dumb questions, it’s transcribing or summarizing research or zoom meetings. Even in software engineering, a lot of LLM tokens are used in the kind of inference that a hypercompetent solo-coding model with limited or no human oversight just wouldn’t need.
Think of an office with 10 human employees working in, say, payroll, constantly sending each other emails, messages, having meetings, calling and speaking to each other and other people, summarizing documents, liaising with other departments, asking AI question about how to use various accounting tools, or about the company’s employee benefits package. Now say this department is automated. An AI model acts as an agent to use an already-existing software package to do all the payroll work. No emails, calls or meetings - or at least far fewer. The total inference work required goes down. And the existing software package doesn’t use AI (even if it may have been coded with it), because you don’t need AI to compute payroll data once you have sufficiently complex and customized software for your business.
In the same way, if we imagine our automated future, super high intensity / high token usage inference is actually not really universally required in a lot of occupations. It will be for some multimodal work (plumbing, surgery, domestic cleaning in complex physical environments), but for many tasks, one-and-done software coded either by AI or that already exists can just be deployed at low intensity by an agent. The AI that replaces your job might at first do a lot of coding, but as time goes on, the amount of novel inference required will diminish. Eventually, software coded in a one-and-done way by the AI may actually handle almost all the workload, and token usage for generation may be very limited to just some high level agent occasionally relaying instructions or performing oversight.
In this scenario, why would we expect inference workloads to shoot up so dramatically? Much enterprise AI usage is currently “fake” in the sense that it would not be performed in a fully automated environment. It’s a between-times thing.
AI is not in the state to do a completely automated economy yet, many tasks still have to be done (or at least directed by) humans. Thus freeing humans up from busywork is still an important gain in our current situation even if eventually this will be become redundant as well.
People shouldn’t be conflating LLM’s with AI, the way they’re imagining the future. I remain firmly convinced that the utility of LLM’s will stay relegated to that of a glorified (and occasionally useful) autocomplete at worst, and at best a work assistant. Some of the recent updates to Gemini that I’ve played with have definitely sharpened their understanding to provide accurate answers to what I’m asking it; the only problem being that they were nothing like the answers it had previously given before; meaning it’s essentially given me every answer under the sun.
It’s a cool “toy” to prompt a human driven research project or to chase down answers to problems, but even when correct information is provided, you still have to vet and validate the veracity of it.
More options
Context Copy link
People have gigantically been freed up from busywork in the office versus what things were like 30-40 years ago. Aside from an expansion of internet pornography, what exactly has been accomplished? New busywork was found, actually manufacturing things was offshored and Western economies have largely trended towards overfinancialized circlejerks where nothing actually happens.
I think it was the economist Ha-Joon Chang that argued the washing machine was more revolutionary than the Internet was, basically for reasons along this train of thought.
More options
Context Copy link
There's tons of stuff that has improved. Things that you want bigger are bigger like cars, TVs and homes. Things we want smaller are smaller like medical devices, computers and cameras. They're all typically much higher quality too. Those medical devices are saving more lives, those cameras take better pictures, those cars are less likely to kill you in a crash.
Stuff is generally cheaper now (per hours of work needed) and more accessible like how 2024 was the first year ever that >50% of Americans took at least one flight. And they did it without having to hire someone to handle bookings for them. Email/texting/etc allows for instant (and automatically stored!) correspondence with anyone I want, meaning we don't have to wait weeks to communicate back and forth. You can listen to almost any song ever recorded, watch basically any show ever made. I can keep track of my financials without having to keep meticulous and detailed records and receipts of where I went and what I spent.
Modern manufacturing is bigger!. Jobs are down because automation and robots are more efficient than people, but we make more now locally than we used to.
That article indicates that it's down since 2007 and roughly flat since 2017. A 20 year decline seems to rather contradict what you're saying, especially when mining is being used to buoy the statistics and that's a fundamentally different space. Stuff cheapening has far more to do with what's gone on with offshoring and pushing to Asia. Software is cool but reflective of the financialization of society and how actual changes in physical technology have slowed dramatically.
More options
Context Copy link
Global manufacturing has actually gone down though (for the reasons you alluded to). That’s not actually the gain people think it is however, because there’s a trade off between resilience and efficiency; especially as technological demands increase in industries like automotive.
I know satellite farming in agribusiness is one example where efficiency is really proving itself to cut down on waste and the poor industry practices of old, but not all industries are benefitting from efficiency. The social system hasn’t yet managed to adapt to rapid technological progress. Especially the government.
Case in point. A little more than a year ago I had to call the IRS to retrieve a document related to my father’s old tax return. It involved me having to send in information about myself and a few other things and they’re still requiring people to fax in paperwork to some random office, wait 2-3 days, with no direct callback number to the agent you’re talking to, to then get a single document physically mailed to me. It’s not like I could just, oh I don’t know, email them a passworded attachment of what they asked for and check it right there over the phone.
The youngest kids in my extended family don’t even know what a fax machine is. Or a VHS tape. Or a Cassette tape. Or how to write in cursive. Or know how to write a check. The IRS is still using fax machines…
Manufacturing is down in two ways. Number of jobs, and share of GDP. But output is much higher. It's down as a share of GDP because other parts of the economy in services and software grew even faster. In part thanks to the automation of factories and farms, which cleared up human labor to go into other fields. People no longer have to work out mowing the fields and picking crops or putting cars together in the assembly line, so they are now free to go do other work and that work has exploded in productivity.
Great example of how governments, without the competitive pressure to improve and outdated (sometimes even conflicting) regulations that lawmakers are too uncaring to address are unable to update themselves in the same way that private corporations generally can.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
Unfortunately, busywork is also subject to Jevon's Paradox.
Building a fully automated economy is going to require conscious effort to build systems that reduce/eliminate human participation. Otherwise the meatbags will just keep making more work for each other.
Exactly. The average office drone today has far greater capabilities than the one of the 1960s who didn't have internet or a computer and needed a manual mail room to be contacted. How much have they done with this?
The average office drone does SO much more today than they used to. What are you even talking about? Unless the bottleneck isn't tech like "the slide deck must be approved by the VP who has no time to review and approve the slide deck"
Accountants in the 1960s couldn't do shit compared to the accountants of 2026, and yet we have a fuckload of accountants in 2026? Why? Because we decided as a society that we wanted to consume significantly more complex accounting than we had in the 1960s.
If the SEC et al. decided to allow simple cash/accrual accounting and totally threw out IFRS/US GAAP we could crash accountant employment by like 80-90%. This would also be a disaster for efficient price discovery.
Hell the IASB is dead set on making accounting ever more complex and retarded with their fixation on PV-ing everything. Securing accountant jobs for future generations!
More options
Context Copy link
Economists have been making this point for a while now. The efficiency gains here haven’t meant a labor drawdown resulted in time shaving activities for workers, it’s been 2x, 3x, 4x the extractive productivity to produce larger profits.
I did see someone point out that workers have cut down on time worked, but it's time worked over their entire life. More school and leisure up front, and earlier retirements relative to death at the end. But the amount of time worked during your mid 20's through late 50's is pretty static.
More options
Context Copy link
More options
Context Copy link
IIRC there is a good amount of data suggesting that engineering teams have shrunk substantially in the last few generations: with computers (spreadsheets, CFD/FEM, digital control systems) product development from bridges to aircraft is at least abstractly more productive. Gone are the days of big rooms of draftsmen, in are a couple of CAD technicians (and they're better about answering "will it fit?" questions), and the parts themselves are getting optimized and closer-packed. Compare a car engine bay from the 60s to today, where there is almost no free space left (does make maintenance a pain sometimes, though), and efficiency is hugely up.
My dad had a 50 year career as an engineer from like mid-60s to mid-10s. When he started out, all firms had huge pools of skilled draftspeople who'd work hand in hand with the engineers all day. Plus all the attendant mail room, secretarial workforce etcetera you'd expect. There's definitely advantages to the flow and things are better, but it's hardly a stepchange.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
Isn't that loosely true of everything following from division of labor? We get more out of farmland when we ascend the technology ladder and start building cars and tractors, not when we maximize the number of field hands.
There seems to be some assumption of "big AI central planning", when adapting existing (market) distributed consensus mechanisms is a possible, and maybe even more plausible, route. Maybe we need hundreds of agents (previously human) compiling The Beige Book regularly and distributing it, not a single Five Year Plan from a hallucinating AI.
More options
Context Copy link
I don't think jevons paradox should be seen as unfortunate so long as the new usage is productive in some form. Efficiency is a key aspect of growing the pot and getting us all bigger portions.
Like this sounds good to me. New jobs getting created to meet previously underserved demands means more total demands being fulfilled and presumably better overall lives.
I don't think it's a problem that we keep coming up with new stuff for people to do (I think we will probably see more and more people employed doing things we previously would have regarded as too frivolous to professionalize).
My point is more that administrivia is somewhat self-perpetuating. Partly this is a function of Jevon's Paradox - as we get more efficient at doing paperwork, one of the biggest results is more paperwork. We now control and track and analyze stuff that would have been impractical to the point of impossibility 50 years ago. Contra some of my other respondents, I don't actually think that this work is useless (otherwise they'd get squeezed out by employers looking to cut costs), but I think it is unlikely to go away without a deliberate effort because it also a function of our prevailing employment paradigm.
Having mulled it over, Jevon's Paradox is probably the wrong conceptual reference. For the foreseeable future, you still need humans to do some stuff. This is real, valuable work, but it may not actually take up most of their time (especially if AI actually delivers on productivity improvements). However, their employer still expects them to be available full time, which means they expect to be paid full-time, which means their employer expects them work full time*, which means creating busywork. Sometimes this is merely stuff of marginal value, sometimes it is outright time wasting. Either way, getting rid of this institutional waste heat and shifting to a genuinely fully automated process would require that you both be able to fully replace human activity with machine activity (not simply augment it) and to step outside of how we currently organize work.
*Also the employees generally want full time employment and prefer employers who offer it
More options
Context Copy link
Jevons Paradox isn’t something you want to deal with, with crises like climate change looming on the horizon. When solving that you have to go to public policy, not to tech (1, 2). The problem with greater efficiency is that the effective production and precision of inputs isn’t necessarily the most optimal one when it increases fragility. That was the whole point Taleb was making when he wrote Antifragile a number of years ago. I’m all for efficiency and all that, but it doesn’t mean it’s without some massive drawbacks.
Externalities in consumption can actually be a problem, but that can be addressed in other ways such as carbon taxes.
I think there needs to be a line drawn between efficiency from cutting unnecessary things, and efficiency from removing all redundancies and backups.
If someone is making a sandwich and between every step they clap their hands for no reason, stopping them from doing that is objectively an improvement. But having another jar of peanut butter in the pantry you bought because you're running low and might need more for this sandwich is just long run efficiency, even if short term it might not be necessary.
I’ll refer you here to the episode that had “thermodynamics” in the title, if you’re interested to hear about the issues with a carbon tax.
And this is where the balance is. You saw it in the policy sphere as well after COVID struck, where people saw just how fragile shipping and supply lines were. I don’t know how many people were paying attention but within Biden’s cabinet, people were talking about the necessity of a large scale program of re-industrialization in the US; because of it.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link