My grandad ran a small letterpress shop and when phototypesetting came in he didn't retrain, he just did smaller and smaller jobs until it was wedding invitations.
Retraining for typesetters meant becoming "prepress" guys, and going from the wide array of tools and hand skills they used to need to Photoshop, Illustrator, and digital imposers and plates. In the process, the job went from a middle-class job to a minimum-wage job, and the bosses who used to respect you (after the union that used to protect you disappeared) became fly-by-night scumbags who treated you like dirt.
As far as I can tell, the entire business for someone who was good at running a press is fancy wedding invitations and business cards with raised type. I am speaking as someone who waltzed into (and later out of) prepress because I knew the digital tools, made no money, and talked with those guys a lot about the past. This discussion would be while we were printing wedding invitations and fancy business cards.
It's simply a different, low-status job now. Retraining is no different than training the first time, because if you want to make a living you either have to somehow be better than the people who started with the new tools (and the new tools don't require as many people) and maybe get into printing packaging or manuals, or you need to do something completely different, and compete with people just starting their careers.
Going to college again is not a trivial thing. The mortgage will still be due.
reality tv star on something like real housewives. I think AI/robotics would have a hard time being that irrational and illogical consistently. That behavior is pretty unique to humans.
I feel like Neuro-sama is essentially already that, but for the Twitch / YT crowd instead of television. It's trivial to train a model to maximize engagement (as one can see gestures broadly), so I'm sure if we can get robots looking nice and human, we'll have Futurama's All My Circuits in no time.
Though I guess there was a token human on that show, so maybe you're right!
The last job is going to be hairdresser, because billionaires have wives (not to assume all billionaires are men), and those wives will want a human to gossip to.
Worse, it was just almost all horses. There are a few in Central Park pulling tourists and that’s enough for the Pollyanna’s to point to and go “See! We’ll be fine!”
Horses where the plaything of the rich. Farmers were using oxen not horses for the most part. However the modern plow (invented in the mid 1830s - John Deere being the best known inventor, but others were inventing them about the same time) needed the faster speed of a horse to work well, but if the previous farming technology worked for you then oxen (or donkeys ... - there are several other work animals used in different areas for farming) were better.
There are likely more horses today than there were in 1900 - but now they are play things for the upper middle class living in the exurbs (commuting via car to their city job). Sure there are a few Amish (and other similar groups) making a living with a horse, but most of them are toys for the wealthy.
Do you have a citation for the horse number? My research [0] shows we have a 1/5 out peak number in ~1900 and our population has 5x since then. This means per capita horses in the country is down 96%, while absolute count is down 80%. Both are huge numbers.
Good call! They have info on general population which shows about double the number of horses I originally thought (at least in 2016), and declining at a good clip of 10%+/decade)
. So drop is still large but just a bit smaller.
The general point is before cars horses were used a lot, entire industries were based around them and our now gone. The toy factor does not compensate because of cost and inconvenience of owning a horse.
>There’s no quick answers, but intuitively I think we all get that the % of jobs not doable by AI and robotics approaches 0 over time.
Can we really say that? I think that might be a prediction based on the current situation and on the assumption that jobs will neither be transformed nor created anew. We don’t know what the future holds for us in terms of jobs – or whether the problem of jobs will be a solved issue. Just as the blacksmith never gave a thought to what a mechanic actually was until he saw a car for the first time.
It is clear that at the limit, machines will be more capable in nearly every way. I suppose there might be some room for a few exceptions like working around big electromagnets that machines can't get close to... Does that sound great? Is there something else you had in mind which would be reserved for humans and fulfilling?
As someone that works in robotics/automation, it pains me to see people in creative/office verticals getting all the shit with AI these days while hard labor positions, the one that everyone could be doing a bit less because they take a toll on you (working the fields, farms, hazardous tasks) remain pretty much the same.
If you use a computer to do something, training an AI to use the computer to do the same thing just takes the computer, which was there anyway. (Yeah, plus training the model, that's not my point.) But if you do something physical, replacing you takes something physical to be there.
We may eventually manufacture the robots to do all these things. We aren't there yet.
I looked into this (or rather, had Google Deep Research do a 30-page report on it!) and it's still a very long way off. So long, that the AI, or any reference it found, would not even give me a rough timeline, but I suspect it is multiple decades away. The main blocker is a very severe crunch in the supply chain of critical materials, mostly rare earth elements which are essential to make those high power density + high precision motors feasibly.
Today we have a tiny fraction (like 2%?) of the global supply needed to make enough robots to replace all humans.
This is before we get to the other logistics like power and the rest of the materials, but we have a path to scale those up (except maybe Copper, that's another tough one.) Turns out rare earth elements are actually very abundant, but extremely hard to mine, very toxic to extract (which is partially why China has a strangehold on those) and we don't have a good path to scale those up.
All of this could be moot if we figure out alternative technologies for equally capable motors that could be manufactured at scale, but AFAIK whatever promising is on the horizon would still take a very long time to ramp up.
First, you make a robot that mines rare earth metals. Then construction robots. Then an automated factory for assembling robots. The whole premise is that the robots are faster at these tasks than humans would be.
If the robotics are feasible their rise should be exponential, not linear, since each robot built makes building the next faster and helps to remove whatever bottleneck you had.
> If the robotics are feasible their rise should be exponential, not linear
That totally does not follow. Robots can be feasible and their payback curve can still be linear or sublinear.
Take mining rare earth metals, for instance. The problem with mining rare earths isn't that it takes so much human labor. The problem is doing it without poisoning the surrounding 30 square miles. Replacing the driver of the earth mover with a robot doesn't fix the chemistry at all, nor (I suspect) does it significantly affect the economics.
Mining is not limited by human labor, we already use huge machines (and some robots) controlled by a few humans. The limitations are investment in huge machines and processing plants, and regulations on where/how you can mine/process.
You can't just have a robot start mining/processing rare earths, anywhere in the country. You need permits, and a lot of money. I guess that didn't stop AI companies ... but there isn't an "infinite" novelty/hype premium for rare-earths, like there is for LLMs.
Robots have been used in industry for decades already. Most jobs won't be worth investing in the robot doing it, since human labour is cheaper than the capital investment.
don't we already have lots of machines that do things like this though? Even if its human operated: excavators, combines, tractors, and other heavy machinery have reduced the physical burden involved in sustaining large societies tremendously.
Yes there are fewer hands on the job site but there are still hands at all. This is increasingly becoming not the case in the future. The goal is to fully eliminate all hands and all minds. There will be no more sponges of newly available labor. Then what happens exactly no one really knows or even wants to meaningfully try to answer, because the implications are too disturbing.
I would agree with you, if we taxed the ever-living fuck out of AI companies to enable UBI, so that the people replaced with robots did not end up homeless.
Ideologically I don't consider myself to be a communist or even leftist, so I'm confused by my own reaction as well, but maybe mandatory joint-stock where the government takes a large chunk of equity of any AI enterprise off the bat and manages it like a sovereign wealth fund is the right answer.
In the long run, we’re all dead anyway. (Unless something drastic changes.) How long it will take matters. We don’t know how long that will take, but driverless cars have been coming for 15 years now and they’re still unavailable in most places.
People are more adaptable than horses and tech solutions usually require people to do some of the work. Dishwashers need loading and unloading. My robot vacuum still needs maintenance. Self-checkout means you do the scanning.
Amazon wants to get into drone delivery and there are cases where that might work, but the way news articles describe it, it’s not going to get a package onto my porch.
You seem to be living under a rock. Yes driverless cars started in 2009, but in 2026 it is very much here. Waymo recently crossed a milestone of more 500k trips I think, we will have millions of driverless cars in the coming years (we’re at 10k right now)
Waymo works in a single digit number of cities with extremely rare, highly favorable climates, and still operate only on fixed pre-mapped routes. And it's taken 17 years to get that far.
We are still a long way from a self driving car being the default type of car
No. The horses were made redundant and no longer exist in such quantities
So, uh, maybe don't dwell on that too much.
Actually, dwell on that a lot. It's probably the only instructive example.
The rich and powerful only suffer us because they need us. They cannot assemble every lead painted doodad by their own hands. If they could, they never would have hired anyone. So what happens when they no longer need as many people?
Society is already far too constructed around catching some of the wealth that just spills off of the people who own it, by being in their orbit. That activity is not actually productive, but since the only people with money are the ones that have a lot of it (K shaped economy), there's just not enough wealth left in the bottom part of the market to support anything else.
There are more horses today than ever... Oxen got the short end of the deal - almost nobody uses oxen even though historically they were the backbone labor of the farm, while the horse was for the right.
The only data I can find ignores the majority of horses today - recreational riders. If you can find a better citation I'd stand corrected, but everything I can find counts race horses and horses used for farming.
So long as they can keep us fighting amongst ourselves in meaningless culture wars. If there is ever a broad alignment of populists spanning left to right, it could get spicy.
> I think we all get that the % of jobs not doable by AI and robotics approaches 0 over time.
I also think that will be the case. However, not all the jobs are worth being automated. Say, if you are a modern-day horse-riding instructor, I very much doubt the robots will come knocking the door to take your job. Not because they can't, but because the headache is not economically worth it.
It'd be uneconomical today but I don't see any reason to believe it will always be that way. Any time you want a high standard of living for people it creates an incentive to eventually make a cheaper way of doing what those people make their high standard of living with.
About the only things I can think to guarantee will have some constancy are "for the sake of being human" type things. Much like buying handmade bowls is something you can do out of pure preference/taste/ideology even though it's cheaper to buy massed produced ones.
I can see that going away too. Eventually the average horse rider is going to be someone of stature of a saudi prince at the rate horse ownership coupled with inequality is going. They'd probably be all about buying the Gemini horse teacher over learning from some lesser birth stableboy. I bet eventually all tutoring from a human is seen as some lesser ick by the rich. People already think as much about you today when you talk about say not relying on claude code. It is all about optics and virtue signalling pedigrees at the end of the day.
There are two key elements in the current AI fearmongering:
1) People think they can predict the future
2) People think humans can't adapt
The number of disruptive "game changing" technologies throughout history is immense. There's no question some people get negatively affected, but humanity as a whole has prospered because of these disruptions.
Literally every time, including during the industrial revolution, people thought it was game over for society, good manners, the youth, the planet etc.
Keep cool and try to adapt to the inevitable change. The next generation will certainly do it and leave you behind if you stay in a corner denying the new reality.
Adapt, sure, but is my life better or worse after adapting? It's a fair question and a fair worry. The goal of most AI companies seems to be to replace some amount of labor with AI. Look how that worked out for our factory workers. They adapted by getting jobs at Walmart and Dollar General, or scraping by in the gig economy.
> They adapted by getting jobs at Walmart and Dollar General, or scraping by in the gig economy.
The tech crowd used to say that only the "dumb" ones did that. They said those who were "smart like us" were much more adaptable, learning to code and whatnot to find new, often even better careers. It will be interesting to see if the emperor has clothes.
No, and obviously neither can you, else you would have explained yourself. Best to not post while under the influence of silly emotions. That is what those tech people who called those manufacturing people "dumb" did and, like this, wasn't useful.
Silly emotions? I'm not the one gleefully celebrating the ruin of others (including families). You've chosen to 'other' a broad group of people with diverse experiences and backgrounds. "Those tech people" certainly include some people who called manufacturers 'dumb', but it also includes many (most?) who didn't, and even some who used to be "manufacturing people".
You say "who cares if they suffer? They're obviously all just bad people". The people who attacked blue collar workers said "who cares if they suffer? They're obviously all just dumb people (and thus bad)." So the bulk of people who you decry are not guilty of what you claim. But you are!
> You say "who cares if they suffer? They're obviously all just bad people".
This comment aside, I do not see that text anywhere, lest my own comments. Get back to us when you return to reality.
> The people who attacked blue collar workers said "who cares if they suffer? They're obviously all just dumb people (and thus bad)."
Whether or not calling them "dumb" was justified, there no implication of "who cares if they suffer". In fact, in the context I remember, those comments were made around trying to figure out how to help those who were deemed "dumb".
> Many of us, deep down, are driven more by the joy of making things, solving problems, and seeing people use what we built than by the act of coding itself.
I've definitely always enjoyed the end-result but only for the fact that it proved I have overcome challenges and slowly but steadily inched one step closer towards mastery. Both are now gone, since mastery is not required and challenges are absent except the soul draining exhaustion of reading markdown all day. Now I'm just producing outcomes for my boss, rather than my own fulfillment. The journey has become a farce, I will ride this out and make my money but I don't care about it much anymore.
That being said in my free time, as in Hobby I am doing pentesting manually for fun. Because productivity is irrelevant for my hobbies, so I can still keep inching towards mastery and overcome challenges, at least on days where my job leaves any trace of surplus energy at the end of the day.
I used to find joy in knowing how things worked inside-out. Whether it was Postgres, or browser engines, or some detail about compilers... I used to find joy in developing the little techniques that made programming NOT building on quicksand. There was satisfation in what you had done being robust.
Pardon my lack of creativity but I'm not even sure what mastery with the new tools (LLMs) would look like.
When you are so abstracted from the details and don't have to know anything, what _is_ mastery? (I am not asking rhetorically and if people have overcome this thought I would be interested in hearing from them.)
Knowing how to ask a machine questions in the right way? Playing as a manager in control of non-human developers?
It all sounds really silly and something that is difficult to aspire to (and I can't imagine explaining my job to others with a straight face).
I think I find it even more egregious because it's all built on paying rent to the LLM operators.
It's not that I don't enjoy results in their own right sometimes but I think I am definitely in the 'this is giving me an identity crisis' camp.
When I would code by hand I often found I was concerned with the house of cards in my head, am I remembering the right variable names, did I forget something that would lead to an exploit, etc. Now I can direct the LLM to do the work and the skill is split between having an understanding of the interactions of the technologies, choosing tech that will be less of a pain to update/migrate. You can even put some LLM-bow grease into finding more efficient approaches catching corner cases etc. And just by asking I can have test frameworks setup, I can drill into some pet peeve and it takes less time. Now we still have lots of ops moments, like the radicel distributed git forge forgetting to use encryption for private repos, and I run into this class of problem in my own casual LLM use, but that's where our taste and attention to detail powers need to be spent in this modern developer world.
I think one thing that troubles me is that I so far have gotten quite far by 'caring about the little things'.
Now we live in a world where this feels, for the most part, unsustainable to keep up with.
When directing LLMs to code something up, it's still occasional useful to have the knowledge and experience of those little things earned over years, but at the same time 1) it seems like I run into fewer opportunities to add to that experience and 2) it seems like you can often just prompt LLMs to 'grill' a design a few times, it will probably surface some of the most important tradeoffs
Yes, the grilling part is where the LLM energy drain is for me, I haven't released any of the software I've been working on yet because they lack polish in my opinion. They may have nobel goals, but no point unleashing more slop on the world. Often when I ask for a correction something new breaks, or it reveals the next corner the model and I didn't think about. I've seen the models come a long way in a short time.
I'm tinkering and paying by the token for deepseek flash picked for its economy and using it to teach myself and form opinions on languages and frameworks that are new to me. It's hard to describe the worth of these opinions, but I think they're important, kinda like why I read a lot of HN looking for morsels of sentiment around the tech industry as represented here.
Well, I'm somewhat minded to agree with you, but they are fast enough, cheap enough and don't get bored of tedious work, which makes them passable at many things. Maybe.
I feel caught between two worlds, I know people both very 'for' and very 'against'.
The business is very keen for people to use them to go faster.
For non-trivial changes, output so far does not meet my quality standard and I tend to spend a lot of time cleaning up after them, probably more than I would have spent doing the work myself. This being much more boring, I hasten to add.
Their comments suck so I feel like I have to go in and rewrite them all.
How do LLMs work? Yes there's a model and there's inference, but dig deeper. How can you, today, with < $50, train a small GPT-2 class model,
how do you fine tune it, what training data do you give it, how do you give it training data, what algorithms do you use for reinforcement learning, what even is RLHF? You don't have to know any of those details to use ChatGPT, just like you don't need to know how a D flip-flop works in a CPU to write some code that runs on one, but if you're looking for feeling like you know what's going on, at their heart, LLMs are still next word prediction engines, so knowing, deeply, how that engine is created and works, would be my idea of mastery of it. Karpathy has a good set of tutorials on this.
>How do LLMs work? Yes there's a model and there's inference, but dig deeper. How can you, today, with < $50, train a small GPT-2 class model, how do you fine tune it, what training data do you give it, how do you give it training data, what algorithms do you use for reinforcement learning, what even is RLHF? You don't have to know any of those details to use ChatGPT, just like you don't need to know how a D flip-flop works in a CPU to write some code that runs on one, but if you're looking for feeling like you know what's going on, at their heart, LLMs are still next word prediction engines, so knowing, deeply, how that engine is created and works, would be my idea of mastery of it. Karpathy has a good set of tutorials on this.
Knowing how an LLM is built, is trained, inferences, isn't
"mastery" in the sense that crafts people use the word.
Mastery to a craftsmen is as much a product of mechanical performance as it is knowledge. For a software craftsmen, that means knowing how to code, debug, and architect a software application by hand.
This, plus since you can't realistically run these LLMs at home, the motivation to get invested in them is not really there (for me at least)
I mean, yes I do find the mathematical background of LLMs fascinating, but it's not going to be something that I can put to practical use (much), if that makes sense?
I've always been either entirely self-directed, choosing my own goals, or almost entirely self-directed, being given high-level goals, and getting to choose my own means to accomplish them. So my opinions are probably different from somebody who was a smaller part in a chain.
You say that you liked learning how things worked inside out, whether it was Postgres or whatever. You can still do that. You can learn it at a high level, and then it becomes like a magic word in your arsenal. You don't have to know or remember the details of implementation, but only the nature of its costs, benefits, and synergies with other systems. You just have to remember that high-level abstraction, maybe just a single vocabulary word, of a low-level reality - the exact syntax is really no longer a concern.
The better the agents get, the less there is playing manager. Really your job now is to have an execute vision. Execution is largely a function of having a map of the high level vision and low-level realities in your head.
Also, there is developing personal libraries. You may find that in your business you repeatedly tackle similar problems. You develop and refine a library that your future agents can utilize. You develop prompts that make your work more automatic and more robust. And when I say you, I mean, you direct your agents to do it.
It's not so different from what it always was, except that years get compressed into months. What was slow and tedious is now... Not.
I don't know how much advantage I have in this new world, having programmed (sometimes more, sometimes less) for 30 years. My brother has never written a line of code in his life. He's been heavy into AI for a while, and he has developed some really incredible stuff. I think he may have already outpaced me.
Did a tailor gain mastery of the sweatshop machine? Did mastery of the automated sewing machines give the same joy to the tailor or was it mind numbing meaningless activity that he would be happier to be free of.
that's a good analogy and puts into words the way i've always felt about software development. Every so often the entire industry rolls over into something new. I've been here 30 years now and, in my experience, the skill that counts is being able learn new skills. To go back to the analogy, to build something new on top of what you've built in the past before everything is swallowed by the quicksand.
I think well north of 90% of people in software engineering have job expectations that, when assisted/automated with AI, boil down to exactly that.
It's a pain in the ass to work at such a job and find challenges or things to master--your superiors don't want you to do that, there's not enough time in the day, and you're tired from LGTM-now-make-this work.
That's most software jobs, because that's what most software companies want.
Because I literally can just burn tokens instead of using any of my brain cells and the outcome is way better than what I would have achieved thinking deeply and intricately about the implementation because Opus 5.5 at xhigh implements way better than I ever could and I'm paid well and have studied for years and it's not a low-level environment at all.
Since opus 4.8 the tools have started surpassing me and now they are far gone, so implementing by hand would mean sacrificing quality and my main goal as an engineer has always been quality so even though I hate it, I must use the best tool for the job.
My grandad on one side of the family was a market gardener. He didn't become a farmer.
My great grandad on the other side was a lighterman[0]. He didn't become a ship captain.
Sometimes professions really are just gone. In both cases in my family it was due to consolidation of resources and concentration of capital/power. Technology was just a catalyst for that.
Back in the '70s my dad used to tell a joke about his dad. He said that my grandfather had a job that supported his whole large family while my dad was growing up. But unfortunately technology caught up with them, and he was replaced by a little device just this big, and he held his fingers apart to demonstrate.
Then he said the real disaster hit: "And then my mom got one."
My grandma was there when he told it more than once, and thought it was hilarious.
It usually came with another joke about the time he was caught with his pecker in the pickle slicer. He was fired, along with the pickle slicer.
In the seventies you were expected to tell jokes like that and allowed to laugh at them.
I think the punchline is that workplace romance usually gets both parties fired, regardless of who initiated. Hence, doing "things" with the pickle slicer means the pickle slicer gets fired for participating...
The first one sounds like the dad was replaced by computers. But then reveals the dad is a sex worker and being replaced by sex toys.
The second one sounds like the person was being intimate with dangerous kitchen equipment, then reveals that a "pickle slicer" is a job for a human and not a machine.
My grandpa was a shoemaker. The real change in technology left his family in poverty in Chicago during the depression. Nobody joked about that, directly.
Jewish humor is frequently about laughing at your own pain. There's a genre of Jewish jokes created in concentration camps.
Don't forget blatantly ignoring all the times something similar happened in the past and didn't turn out fine.
Sticking your fingers in your ears and singing "Lalalalalala" will in fact sometimes work for some people, and that is not a good argument for everyone to do it.
I'm not sure I can quite agree with that. I'd say that spreadsheets definitely reduced the need to manually calculate things, and calculators before that, and the abacus before that.
I basically agree with your "higher level reasoning", but not your statement that we previously only automated muscles.
Compilers automated higher level reasoning, at least to a small degree. LLMs themselves are basically just compilers with a much, much, much, much more expansive set of built-in language constructs.
Can you describe the last time you asked your compiler how you should best approach an interpersonal problem, to research an investment thesis, or to figure out an org chart by analyzing communication patterns within teams?
People aren't just using AI for writing code, but for all of their decision making. They're far from limited to writing code.
> Can you describe the last time you asked your compiler how you should best approach an interpersonal problem
Sure. I wrote out specification that described the problem. It compiled that into a description of how to deal with it.
> People aren't just using AI for writing code
Compilers have never been about producing code in particular, though, even if that is where they are most commonly used. A LaTeX compiler, for example, outputs books. To think they are about code exclusively is a grave misunderstanding of what a compiler is.
> Sure. I wrote out specification that described the problem.
The LLM usage I see from most people has the LLM write out the specification, asking clarifying questions as needed. If you're using LLMs as you describe, you're using them ineffectively.
> The LLM usage I see from most people has the LLM write out the specification
If the compiler is writing the specification, what purpose does the person serve? That is not what I'm seeing out in the wild, though. Perhaps what is happening in research labs aside, compilers, even those implemented as an LLM, don't do anything until the user specifies their intent.
In fact, you literally recognized that in your earlier comment, indicating that my action to specify my problem was required, so this change of tune now — thinking that the system recognizes my problems before I have to ask — is quite curious. There is likely a day when the tooling gets us there, but that day is not yet.
> asking clarifying questions as needed.
Sure. Even the earliest, most primitive compilers would seek clarification if there was something that it felt required more information. That's expected.
> If you're using LLMs as you describe, you're using them ineffectively.
LLMs are just compilers so they can't do anything else. Maybe you're thinking of some technology that is layered on top of a compiler like an IDE, REPL, or something like that?
> Sure. Even the earliest, most primitive compilers would seek clarification if there was something that it felt required more information. That's expected.
Can you tell me what the last question GCC asked you was, and can you rate its coherence in comparison to your colleagues questions?
I don't think I knew how to solve the problem, just like, if we want to talk about code, many coders don't know how to solve a hashmap. The compiler just figures it out for them, as it did me for my problems. I'd call that specifying the problem, but words are in the eye of the beholder, I suppose.
If we take "write the specification of how to approach the problem" at literal value, you might be narrowly thinking of a low-level compiler, or what sometimes gets called a transpiler? However, compilers encompass a much wider spectrum than that. Generally they offer higher levels of abstraction.
I think you're talking about more than just a compiler, though standard libraries are often distributed with the compiler. Take Rust for example, it's not like the compiler has special integration with hash maps, it's just that the standard library has a standard implementation and the compiler lets you refer to that implementation without having to repeat it yourself. In a way you're just including parts of another spec in yours. Specifications do this all the time, for example by using English. You don't have to define all the levels of abstraction, only the ones that matter.
Even so, I'd argue that you still have to define how to use those abstractions rather than only the result you'd like to achieve with them. Sure, maybe the abstraction has a ".achieve_the_result()" method, and calling that method is how you achieve it, but often there's a bunch of stuff you fit together to achieve a solution to your problem and the compiler doesn't necessarily do that for you.
I think there actually may be some compilers that get closer to this, for example some problems can be expressed quite elegantly in Haskell or APL and the compiler will do a lot of black magic behind the scenes to get you a performant executable (somehow). But for most traditional compilers you are still going to have to define how to approach your problem, even if only at a higher level thanks to libraries.
> Compilers have never been about producing code in particular, though, even if that is where they are most commonly used. A LaTeX compiler, for example, outputs books. To think they are about code exclusively is a grave misunderstanding of what a compiler is.
And what, pray tell, are those books output by TeX? That's right, PDF or PostScript code.
This is a tired, idiotic analogy. For one, if LLMs are compilers, they're terrible compilers because they can't be reasoned about the way one wants to reason about a compiler (but I assume you have a weird take here since you claim compilers are non-deterministic).
But more importantly, if you insist on treating LLMs like compilers, then the source language – natural prose – is the worst programming language ever invented. Coding the world in brainfuck is a better idea than where you lot is suggesting we go.
> Today, the number of people employed purely for their muscle is very small.
I also believe we underestimate how much that costs us too. Yes, it is good that fewer people are breaking their bodies on brutal physical jobs.
But there are many people in the world for whom their physical strength and dexterity are perhaps their most valuable asset. With the loss of those jobs came a loss of an avenue of meaning for them.
I could be wrong, but I have a hunch that the whole manosphere alt-right conservative thing going on in the US is driven in part by a generation of young men who don't find intellectual work to be appealing but don't see many alternatives out there either. It's telling how closely Gamergate is tangled up in that world and how many games are basically simulations of heavily physical prestigious roles.
A recent Onion headline really hits the nail on the head: "'This Here Is Probably Our Bestselling Love Seat,' Says Man Who Would Have Been Powerful, Revered Warrior 4,000 Years Ago"
Amazing. I bet this can really help thousands of programmers who are simply not needed, because companies decided to replace them with AI agents. These philosophical ponders are very comfortable to do when your ass is safe. Most of people don’t have this privilege.
I like this article, but it took decades for cars to take over: we needed to build vehicles, build roads, build up gas production and distribution... as Bill Gates wrote, one of the differences with the AI disruption is how fast it's coming, since it's almost immaterial (except data centers of course). It may be too fast for humans to adapt like they did in the previous revolutions.
Humans ourselves are a technology. We have incredibly flexible brains and dexterous hands. Up to now, we could always do things that robots couldn't. That was where our job security came from.
But what happens when a robot can do anything a human can? When a cheap machine can do anything you can do but faster and better, what makes you different from a horse in the 1920s?
This is what the article misses and what people have trouble accepting. Yea, this is a technological revolution, but unlike the Industrial Revolution, you can't just upskill your way out of it.
Less misanthropically: The technology required to replace the person is much smarter and more adaptable than that which was needed to replace the horse, so you can end up in the horse's role in the story without having to be as dumb as a literal one.
I think the analogy of moving from horses to engines breaks down when going from code to AI as AI can quite plausibly work on itself. It's a fundamentally different game we're playing this time. And there were as many engine jobs as horse jobs (maybe more). The whole point of AI is super human productivity. Even the idea of just using AI means you need less humans because one human with AI can out-produce multiple humans without. It's very difficult to imagine a future where AI can do every job and things don't turn into a sci-fi dystopia.
I don’t particularly like the horse to engine angle because it doesn’t capture the issue well.
When we automated physical activity people that were crowded out moved into knowledge work. Now we’re automating knowledge work so people move into ???
Leisure? Become a useless class? UBI? Something else entirely?
You mean my meat proxy skills while wearing a Ratatouille machine-vision head camera that tells me what wrench to pick up and how to crimp the Pex rings?
Plumbing is kind of a hard job, definitely not fun for anyone above middle age. And given it will pay minimum wage, I think I'd just take a life of leisure instead.
Sure, but that is today. When AI decimates the knowledge fields, a whole lot of people are going to be looking at different career options. There will not be any shortage of people hoping to make money in the trades, and with that surge in supply will come a collapse in prices.
> Why would it not? The economic incentives make it inevitable.
Citation needed. Again, I wanna emphasize that I'm not saying it won't happen. I'm just saying that it's not a given that it will.
There is very little sign that current models can do anything but (at best!) augment most knowledge workers (with some exceptions). And there are very reasonable open questions about whether it's economically, and ecologically, sustainable to continue development at anything near the current pace. There is, in fact, plenty of indicators that already the current SoTA is the result of a financial overexertion.
Apologies, I'm with you. I was speaking to the hypothetical, not trying to suggest it is inevitable. I think it is uncertain how this plays out, but I am an optimist and I think it ends up a net positive.
> There will not be any shortage of people hoping to make money in the trades, and with that surge in supply will come a collapse in prices.
And the worst case downward spiral will involve not only a surge in supply but a drastic drop in demand to go along with it.
All of those previous "knowledge workers" scurrying for the trades not only become more labor supply, but without their cushy jobs they will no longer be able to afford to hire trades people to prop up the demand side.
The knowledge economy was already on the way out in favour of the performance economy even before LLMs. LLMs have also shown that they can produce in the performance space, of course, so it remains to be seen how valuable humans are in that position but so far it would appear that many are willing to pay for a "human connection".
Plenty if not all people put more value into handmade things, but they can't afford to have everything handmade. An IKEA coffee table is $15. Handmade one from a woodworker is $400 - $1000.
Absolutely. Hence why manufacturing died long ago, and why the knowledge economy was already on the verge of death even before LLMs, and why we were already moving to the performance economy: because in the performance economy the consumer and the buyer are able to be disparate entities, where the buyers can have deep pockets even if the consumers don't.
In the performance economy it doesn't matter how rich the consumer is, but what the consumer wants to consume does matter. Because the buyer and consumer are disjointed in the performance economy, the consumer isn't swayed by cost. However, if the consumer prefers AI performances, which they quite easily could, then humans are in trouble. But so far it seems human performances remain what the consumer wants to consume.
I agree with this too and I find this article extremely depressing.
The AI "revolution" as described here is the wholesale outsourcing of vast swaths of thinking and decision-making to a single entity. A project coded with Claude is a project coded by Anthropic. A business whose decisions are made with ChatGPT is a business run by OpenAI.
Is that what these people really want? To become a conduit through which two or three companies get to exert control over everything in the world?
> But it makes me wonder: what if what we loved about this job was never only the code?
This exactly. Look, I "love" code but only because of what it allows me to do and there is plenty of code/coding that I don't love. Not to be too hippy-dippy but I'm a builder, I like to build things. Build things for myself, build things for others. The code was _always_ a means to an end.
Sure, it was cool I had this skill that many people didn't have but I also used to have skill in managing the windows machines and things like the registry. When I first switched to macOS I felt lost, I had this "skill" that was no longer needed but after a while I realized that was a good thing, I wasn't wasting my time fiddling with crap that was just getting in the way of _using_ my computer.
Similarly I may mourn the loss of writing code but at the end of the day code was just a way for me to build, if there are better tools to build (LLMs) then I'm all for it.
I too love building things, but by building things I mean the act of actually building it. Thinking of a problem, devising a solution and then executing the steps to solve the problem, if that doesn’t work, debugging and trying another method and so on. I am not and will never be a fan of buying things, where I just ask for something and it gets made which is what you’re a fan of if you’re honest to yourself.
> Many of us, deep down, are driven more by the joy of making things, solving problems, and seeing people use what we built than by the act of coding itself. After all, many of us decided we wanted to be developers before learning to code, just to create video games or websites. The desire came before the code.
This is a big disconnect that I think has been at the root of many different reactions to LLMs lately. Here on HN I see a lot of "I hate what LLMs did to programming" and equally a lot of "LLMs are awesome, I've never had a better time". I think the disconnect is between people who build software for the end result (having users, the joy of completing a thing, etc) versus building it for the process itself (architecting, coding, debugging).
I'm mainly the kind who enjoys the process. I think of an awesome startup job I had that solved real problems for real customers including some really big firms, but I can't honestly say I ever cared about those problems or the customers. I loved the job because it was technically challenging with a lot of freedom, and the customers were great because they were a source of new interesting details to solve, as well as confirmations that the previous solutions worked.
As such LLMs make me sad because they effectively replace a lot of what I like doing. Doing a refactor to support a new feature, digging through the code to try and understand a bug. One of my strengths as a developer, I think, is that I am fast at familiarizing myself with unfamiliar codebases. That's also a somewhat redundant skill as Claude can answer "how does feature X work in this code" in a small fraction of the time it'd take me.
A few years ago (before the LLM craze) I moved to game dev as my main job, which makes a lot of this easier because I actually care about the end product and the customers (players), so that's a whole new layer of satisfaction in addition to the usual technical one.
Which isn't to say that it's all bad. I've used Claude to get a few things done where I mainly care about the final result and not the process. There was one such project I wanted to get done, I'd had the idea and the data for five or six years, but I found that the implementation would be time consuming, often manual work and much of it involving the kind of tech I don't really enjoy. So I never worked on that but last month, I did the whole thing in a few days with Claude and got exactly the result I wanted.
I've been seeing things the exact same way. There are those that want the end product as fast as possible and there are those that love the process and the struggle of getting there. The former group is the ones telling the latter group that they'll be left behind.
sorry I think this is a false equivalency. This isn't the career path of programming getting replaced by a new technology. This is technology replacing career paths.
We're all familiar with the term "knowledge worker." What happens when that field is broadly removed? It will be a reverse-trend of telling all ex-coal miners to become software engineers. Except here's the rub: going from a software engineer to something requiring a human only for their dexterity is typically a massive earnings decrease. So, in effect, as Brian Merchant and others put it, this is not just a shift but a massive plan of wage theft.
Looking at it from the optimistic side as my bosses frame it: we're all moving up in the ranks! Well, my team already has a PM. She's already trying to vibe code her PRDs. Thankfully, right now, they suck. But they won't forever because all of us engineers are being forced to train our replacement. There can only be so many strategists, project managers, designers, etc. This isn't simply the technology changing and us adapting, this is a massive reduction in workforce requirements ACROSS industries. Hype bros thinking we'll all have jobs once the AI is good enough to replace us are simply hallucinating as bad as the frontier models of 2024.
There’s no demand for human dexterity anyway, and it’s not about money. I like doing intellectual work.
They are spending 1T to ruin the quality of work we do, also probably ruining our chances of securing a livelihood and then they want us to for some reason welcome it with open arms and openly say look how great this is and marvel at human obsolescence. Yeah no thanks.
473 comments
[ 71.6 ms ] story [ 1867 ms ] threadDon't be the old dog who can't learn new tricks.
As far as I can tell, the entire business for someone who was good at running a press is fancy wedding invitations and business cards with raised type. I am speaking as someone who waltzed into (and later out of) prepress because I knew the digital tools, made no money, and talked with those guys a lot about the past. This discussion would be while we were printing wedding invitations and fancy business cards.
It's simply a different, low-status job now. Retraining is no different than training the first time, because if you want to make a living you either have to somehow be better than the people who started with the new tools (and the new tools don't require as many people) and maybe get into printing packaging or manuals, or you need to do something completely different, and compete with people just starting their careers.
Going to college again is not a trivial thing. The mortgage will still be due.
There’s no quick answers, but intuitively I think we all get that the % of jobs not doable by AI and robotics approaches 0 over time.
Though I guess there was a token human on that show, so maybe you're right!
There are likely more horses today than there were in 1900 - but now they are play things for the upper middle class living in the exurbs (commuting via car to their city job). Sure there are a few Amish (and other similar groups) making a living with a horse, but most of them are toys for the wealthy.
0 - https://datapaddock.com/usda-horse-total-1850-2012/
The general point is before cars horses were used a lot, entire industries were based around them and our now gone. The toy factor does not compensate because of cost and inconvenience of owning a horse.
0 - https://datapaddock.com/
Can we really say that? I think that might be a prediction based on the current situation and on the assumption that jobs will neither be transformed nor created anew. We don’t know what the future holds for us in terms of jobs – or whether the problem of jobs will be a solved issue. Just as the blacksmith never gave a thought to what a mechanic actually was until he saw a car for the first time.
We may eventually manufacture the robots to do all these things. We aren't there yet.
To the end of it? Maybe 40?
Today we have a tiny fraction (like 2%?) of the global supply needed to make enough robots to replace all humans.
This is before we get to the other logistics like power and the rest of the materials, but we have a path to scale those up (except maybe Copper, that's another tough one.) Turns out rare earth elements are actually very abundant, but extremely hard to mine, very toxic to extract (which is partially why China has a strangehold on those) and we don't have a good path to scale those up.
All of this could be moot if we figure out alternative technologies for equally capable motors that could be manufactured at scale, but AFAIK whatever promising is on the horizon would still take a very long time to ramp up.
First, you make a robot that mines rare earth metals. Then construction robots. Then an automated factory for assembling robots. The whole premise is that the robots are faster at these tasks than humans would be.
If the robotics are feasible their rise should be exponential, not linear, since each robot built makes building the next faster and helps to remove whatever bottleneck you had.
That totally does not follow. Robots can be feasible and their payback curve can still be linear or sublinear.
Take mining rare earth metals, for instance. The problem with mining rare earths isn't that it takes so much human labor. The problem is doing it without poisoning the surrounding 30 square miles. Replacing the driver of the earth mover with a robot doesn't fix the chemistry at all, nor (I suspect) does it significantly affect the economics.
You can't just have a robot start mining/processing rare earths, anywhere in the country. You need permits, and a lot of money. I guess that didn't stop AI companies ... but there isn't an "infinite" novelty/hype premium for rare-earths, like there is for LLMs.
People are more adaptable than horses and tech solutions usually require people to do some of the work. Dishwashers need loading and unloading. My robot vacuum still needs maintenance. Self-checkout means you do the scanning.
Amazon wants to get into drone delivery and there are cases where that might work, but the way news articles describe it, it’s not going to get a package onto my porch.
We are still a long way from a self driving car being the default type of car
So, uh, maybe don't dwell on that too much.
Actually, dwell on that a lot. It's probably the only instructive example.
The rich and powerful only suffer us because they need us. They cannot assemble every lead painted doodad by their own hands. If they could, they never would have hired anyone. So what happens when they no longer need as many people?
Society is already far too constructed around catching some of the wealth that just spills off of the people who own it, by being in their orbit. That activity is not actually productive, but since the only people with money are the ones that have a lot of it (K shaped economy), there's just not enough wealth left in the bottom part of the market to support anything else.
The global horse population peaked in the 1910s.
Horse ownership isn't what it used to be, but the horse population seems fine.
Well that and there are more guns than people and if a million people come for you then there is no army that will be able to save you.
I also think that will be the case. However, not all the jobs are worth being automated. Say, if you are a modern-day horse-riding instructor, I very much doubt the robots will come knocking the door to take your job. Not because they can't, but because the headache is not economically worth it.
About the only things I can think to guarantee will have some constancy are "for the sake of being human" type things. Much like buying handmade bowls is something you can do out of pure preference/taste/ideology even though it's cheaper to buy massed produced ones.
1) People think they can predict the future
2) People think humans can't adapt
The number of disruptive "game changing" technologies throughout history is immense. There's no question some people get negatively affected, but humanity as a whole has prospered because of these disruptions.
Literally every time, including during the industrial revolution, people thought it was game over for society, good manners, the youth, the planet etc.
Keep cool and try to adapt to the inevitable change. The next generation will certainly do it and leave you behind if you stay in a corner denying the new reality.
The tech crowd used to say that only the "dumb" ones did that. They said those who were "smart like us" were much more adaptable, learning to code and whatnot to find new, often even better careers. It will be interesting to see if the emperor has clothes.
You say "who cares if they suffer? They're obviously all just bad people". The people who attacked blue collar workers said "who cares if they suffer? They're obviously all just dumb people (and thus bad)." So the bulk of people who you decry are not guilty of what you claim. But you are!
This comment aside, I do not see that text anywhere, lest my own comments. Get back to us when you return to reality.
> The people who attacked blue collar workers said "who cares if they suffer? They're obviously all just dumb people (and thus bad)."
Whether or not calling them "dumb" was justified, there no implication of "who cares if they suffer". In fact, in the context I remember, those comments were made around trying to figure out how to help those who were deemed "dumb".
I've definitely always enjoyed the end-result but only for the fact that it proved I have overcome challenges and slowly but steadily inched one step closer towards mastery. Both are now gone, since mastery is not required and challenges are absent except the soul draining exhaustion of reading markdown all day. Now I'm just producing outcomes for my boss, rather than my own fulfillment. The journey has become a farce, I will ride this out and make my money but I don't care about it much anymore.
That being said in my free time, as in Hobby I am doing pentesting manually for fun. Because productivity is irrelevant for my hobbies, so I can still keep inching towards mastery and overcome challenges, at least on days where my job leaves any trace of surplus energy at the end of the day.
It does feel a bit like building on quicksand, but coding has always been that way for me. Well, less so with embedded than webdev, but still.
Pardon my lack of creativity but I'm not even sure what mastery with the new tools (LLMs) would look like. When you are so abstracted from the details and don't have to know anything, what _is_ mastery? (I am not asking rhetorically and if people have overcome this thought I would be interested in hearing from them.)
Knowing how to ask a machine questions in the right way? Playing as a manager in control of non-human developers? It all sounds really silly and something that is difficult to aspire to (and I can't imagine explaining my job to others with a straight face). I think I find it even more egregious because it's all built on paying rent to the LLM operators.
It's not that I don't enjoy results in their own right sometimes but I think I am definitely in the 'this is giving me an identity crisis' camp.
I think one thing that troubles me is that I so far have gotten quite far by 'caring about the little things'.
Now we live in a world where this feels, for the most part, unsustainable to keep up with. When directing LLMs to code something up, it's still occasional useful to have the knowledge and experience of those little things earned over years, but at the same time 1) it seems like I run into fewer opportunities to add to that experience and 2) it seems like you can often just prompt LLMs to 'grill' a design a few times, it will probably surface some of the most important tradeoffs
They can build things, but it's all terrible really. Their "ideas" are terrible.
I feel caught between two worlds, I know people both very 'for' and very 'against'. The business is very keen for people to use them to go faster.
For non-trivial changes, output so far does not meet my quality standard and I tend to spend a lot of time cleaning up after them, probably more than I would have spent doing the work myself. This being much more boring, I hasten to add.
Their comments suck so I feel like I have to go in and rewrite them all.
Knowing how an LLM is built, is trained, inferences, isn't "mastery" in the sense that crafts people use the word.
Mastery to a craftsmen is as much a product of mechanical performance as it is knowledge. For a software craftsmen, that means knowing how to code, debug, and architect a software application by hand.
I mean, yes I do find the mathematical background of LLMs fascinating, but it's not going to be something that I can put to practical use (much), if that makes sense?
You say that you liked learning how things worked inside out, whether it was Postgres or whatever. You can still do that. You can learn it at a high level, and then it becomes like a magic word in your arsenal. You don't have to know or remember the details of implementation, but only the nature of its costs, benefits, and synergies with other systems. You just have to remember that high-level abstraction, maybe just a single vocabulary word, of a low-level reality - the exact syntax is really no longer a concern.
The better the agents get, the less there is playing manager. Really your job now is to have an execute vision. Execution is largely a function of having a map of the high level vision and low-level realities in your head.
Also, there is developing personal libraries. You may find that in your business you repeatedly tackle similar problems. You develop and refine a library that your future agents can utilize. You develop prompts that make your work more automatic and more robust. And when I say you, I mean, you direct your agents to do it.
It's not so different from what it always was, except that years get compressed into months. What was slow and tedious is now... Not.
I don't know how much advantage I have in this new world, having programmed (sometimes more, sometimes less) for 30 years. My brother has never written a line of code in his life. He's been heavy into AI for a while, and he has developed some really incredible stuff. I think he may have already outpaced me.
that's a good analogy and puts into words the way i've always felt about software development. Every so often the entire industry rolls over into something new. I've been here 30 years now and, in my experience, the skill that counts is being able learn new skills. To go back to the analogy, to build something new on top of what you've built in the past before everything is swallowed by the quicksand.
I think well north of 90% of people in software engineering have job expectations that, when assisted/automated with AI, boil down to exactly that.
It's a pain in the ass to work at such a job and find challenges or things to master--your superiors don't want you to do that, there's not enough time in the day, and you're tired from LGTM-now-make-this work.
That's most software jobs, because that's what most software companies want.
Since opus 4.8 the tools have started surpassing me and now they are far gone, so implementing by hand would mean sacrificing quality and my main goal as an engineer has always been quality so even though I hate it, I must use the best tool for the job.
My great grandad on the other side was a lighterman[0]. He didn't become a ship captain.
Sometimes professions really are just gone. In both cases in my family it was due to consolidation of resources and concentration of capital/power. Technology was just a catalyst for that.
[0] https://en.wikipedia.org/wiki/Lighterman
Then he said the real disaster hit: "And then my mom got one."
My grandma was there when he told it more than once, and thought it was hilarious.
It usually came with another joke about the time he was caught with his pecker in the pickle slicer. He was fired, along with the pickle slicer.
In the seventies you were expected to tell jokes like that and allowed to laugh at them.
The first one sounds like the dad was replaced by computers. But then reveals the dad is a sex worker and being replaced by sex toys.
The second one sounds like the person was being intimate with dangerous kitchen equipment, then reveals that a "pickle slicer" is a job for a human and not a machine.
Jewish humor is frequently about laughing at your own pain. There's a genre of Jewish jokes created in concentration camps.
That's how it sounds to me
Sticking your fingers in your ears and singing "Lalalalalala" will in fact sometimes work for some people, and that is not a good argument for everyone to do it.
This time is the first time we've automated higher level reasoning.
I basically agree with your "higher level reasoning", but not your statement that we previously only automated muscles.
Also relevant: https://en.wikipedia.org/wiki/List_of_obsolete_occupations
(No, compilers are not deterministic)
People aren't just using AI for writing code, but for all of their decision making. They're far from limited to writing code.
Sure. I wrote out specification that described the problem. It compiled that into a description of how to deal with it.
> People aren't just using AI for writing code
Compilers have never been about producing code in particular, though, even if that is where they are most commonly used. A LaTeX compiler, for example, outputs books. To think they are about code exclusively is a grave misunderstanding of what a compiler is.
The LLM usage I see from most people has the LLM write out the specification, asking clarifying questions as needed. If you're using LLMs as you describe, you're using them ineffectively.
If the compiler is writing the specification, what purpose does the person serve? That is not what I'm seeing out in the wild, though. Perhaps what is happening in research labs aside, compilers, even those implemented as an LLM, don't do anything until the user specifies their intent.
In fact, you literally recognized that in your earlier comment, indicating that my action to specify my problem was required, so this change of tune now — thinking that the system recognizes my problems before I have to ask — is quite curious. There is likely a day when the tooling gets us there, but that day is not yet.
> asking clarifying questions as needed.
Sure. Even the earliest, most primitive compilers would seek clarification if there was something that it felt required more information. That's expected.
> If you're using LLMs as you describe, you're using them ineffectively.
LLMs are just compilers so they can't do anything else. Maybe you're thinking of some technology that is layered on top of a compiler like an IDE, REPL, or something like that?
Can you tell me what the last question GCC asked you was, and can you rate its coherence in comparison to your colleagues questions?
I might just not know enough compilers, but didn't you write the specification of how to approach the problem, not just the problem itself?
If we take "write the specification of how to approach the problem" at literal value, you might be narrowly thinking of a low-level compiler, or what sometimes gets called a transpiler? However, compilers encompass a much wider spectrum than that. Generally they offer higher levels of abstraction.
Even so, I'd argue that you still have to define how to use those abstractions rather than only the result you'd like to achieve with them. Sure, maybe the abstraction has a ".achieve_the_result()" method, and calling that method is how you achieve it, but often there's a bunch of stuff you fit together to achieve a solution to your problem and the compiler doesn't necessarily do that for you.
I think there actually may be some compilers that get closer to this, for example some problems can be expressed quite elegantly in Haskell or APL and the compiler will do a lot of black magic behind the scenes to get you a performant executable (somehow). But for most traditional compilers you are still going to have to define how to approach your problem, even if only at a higher level thanks to libraries.
And what, pray tell, are those books output by TeX? That's right, PDF or PostScript code.
But more importantly, if you insist on treating LLMs like compilers, then the source language – natural prose – is the worst programming language ever invented. Coding the world in brainfuck is a better idea than where you lot is suggesting we go.
Having used, designed, and written compilers, this is news to me. Please, elaborate.
That’s very generous of you to call it that way.
It’s a parody at best.
I also believe we underestimate how much that costs us too. Yes, it is good that fewer people are breaking their bodies on brutal physical jobs.
But there are many people in the world for whom their physical strength and dexterity are perhaps their most valuable asset. With the loss of those jobs came a loss of an avenue of meaning for them.
I could be wrong, but I have a hunch that the whole manosphere alt-right conservative thing going on in the US is driven in part by a generation of young men who don't find intellectual work to be appealing but don't see many alternatives out there either. It's telling how closely Gamergate is tangled up in that world and how many games are basically simulations of heavily physical prestigious roles.
A recent Onion headline really hits the nail on the head: "'This Here Is Probably Our Bestselling Love Seat,' Says Man Who Would Have Been Powerful, Revered Warrior 4,000 Years Ago"
But what happens when a robot can do anything a human can? When a cheap machine can do anything you can do but faster and better, what makes you different from a horse in the 1920s?
This is what the article misses and what people have trouble accepting. Yea, this is a technological revolution, but unlike the Industrial Revolution, you can't just upskill your way out of it.
Grim.
https://www.vatican.va/content/leo-xiv/en/encyclicals/docume...
Humans are animals who throughout time have become conditioned to follow production processes for the betterment of society.
HAHAHAHAHA in a discussion full of doomers, I really appreciate the guffaw that gave me. Thank you.
When we automated physical activity people that were crowded out moved into knowledge work. Now we’re automating knowledge work so people move into ???
Leisure? Become a useless class? UBI? Something else entirely?
Citation needed. Again, I wanna emphasize that I'm not saying it won't happen. I'm just saying that it's not a given that it will.
There is very little sign that current models can do anything but (at best!) augment most knowledge workers (with some exceptions). And there are very reasonable open questions about whether it's economically, and ecologically, sustainable to continue development at anything near the current pace. There is, in fact, plenty of indicators that already the current SoTA is the result of a financial overexertion.
And the worst case downward spiral will involve not only a surge in supply but a drastic drop in demand to go along with it.
All of those previous "knowledge workers" scurrying for the trades not only become more labor supply, but without their cushy jobs they will no longer be able to afford to hire trades people to prop up the demand side.
Double whammy.
In the performance economy it doesn't matter how rich the consumer is, but what the consumer wants to consume does matter. Because the buyer and consumer are disjointed in the performance economy, the consumer isn't swayed by cost. However, if the consumer prefers AI performances, which they quite easily could, then humans are in trouble. But so far it seems human performances remain what the consumer wants to consume.
Like our capital class overlords?
The AI "revolution" as described here is the wholesale outsourcing of vast swaths of thinking and decision-making to a single entity. A project coded with Claude is a project coded by Anthropic. A business whose decisions are made with ChatGPT is a business run by OpenAI.
Is that what these people really want? To become a conduit through which two or three companies get to exert control over everything in the world?
This exactly. Look, I "love" code but only because of what it allows me to do and there is plenty of code/coding that I don't love. Not to be too hippy-dippy but I'm a builder, I like to build things. Build things for myself, build things for others. The code was _always_ a means to an end.
Sure, it was cool I had this skill that many people didn't have but I also used to have skill in managing the windows machines and things like the registry. When I first switched to macOS I felt lost, I had this "skill" that was no longer needed but after a while I realized that was a good thing, I wasn't wasting my time fiddling with crap that was just getting in the way of _using_ my computer.
Similarly I may mourn the loss of writing code but at the end of the day code was just a way for me to build, if there are better tools to build (LLMs) then I'm all for it.
https://manuel.darcemont.fr/projects/
We are the horses.
They wanted to make money, or at least be a part of history during a transformational era. No dice.
This is a big disconnect that I think has been at the root of many different reactions to LLMs lately. Here on HN I see a lot of "I hate what LLMs did to programming" and equally a lot of "LLMs are awesome, I've never had a better time". I think the disconnect is between people who build software for the end result (having users, the joy of completing a thing, etc) versus building it for the process itself (architecting, coding, debugging).
I'm mainly the kind who enjoys the process. I think of an awesome startup job I had that solved real problems for real customers including some really big firms, but I can't honestly say I ever cared about those problems or the customers. I loved the job because it was technically challenging with a lot of freedom, and the customers were great because they were a source of new interesting details to solve, as well as confirmations that the previous solutions worked.
As such LLMs make me sad because they effectively replace a lot of what I like doing. Doing a refactor to support a new feature, digging through the code to try and understand a bug. One of my strengths as a developer, I think, is that I am fast at familiarizing myself with unfamiliar codebases. That's also a somewhat redundant skill as Claude can answer "how does feature X work in this code" in a small fraction of the time it'd take me.
A few years ago (before the LLM craze) I moved to game dev as my main job, which makes a lot of this easier because I actually care about the end product and the customers (players), so that's a whole new layer of satisfaction in addition to the usual technical one.
Which isn't to say that it's all bad. I've used Claude to get a few things done where I mainly care about the final result and not the process. There was one such project I wanted to get done, I'd had the idea and the data for five or six years, but I found that the implementation would be time consuming, often manual work and much of it involving the kind of tech I don't really enjoy. So I never worked on that but last month, I did the whole thing in a few days with Claude and got exactly the result I wanted.
We're all familiar with the term "knowledge worker." What happens when that field is broadly removed? It will be a reverse-trend of telling all ex-coal miners to become software engineers. Except here's the rub: going from a software engineer to something requiring a human only for their dexterity is typically a massive earnings decrease. So, in effect, as Brian Merchant and others put it, this is not just a shift but a massive plan of wage theft.
Looking at it from the optimistic side as my bosses frame it: we're all moving up in the ranks! Well, my team already has a PM. She's already trying to vibe code her PRDs. Thankfully, right now, they suck. But they won't forever because all of us engineers are being forced to train our replacement. There can only be so many strategists, project managers, designers, etc. This isn't simply the technology changing and us adapting, this is a massive reduction in workforce requirements ACROSS industries. Hype bros thinking we'll all have jobs once the AI is good enough to replace us are simply hallucinating as bad as the frontier models of 2024.
They are spending 1T to ruin the quality of work we do, also probably ruining our chances of securing a livelihood and then they want us to for some reason welcome it with open arms and openly say look how great this is and marvel at human obsolescence. Yeah no thanks.