>My challenge ahead is to move from making a living by thinking, to making thinking part of a life worth living.
Our challenge ahead is making a living at all. Who is going to be having hobbies if the economy is 100% automated? We're not going to sleepwalk our way into a post-scarcity automated economy where everyone's needs are provided for and they can just spend all their time on hobbies. If we believe AI can replace all cognitive labor, then it follows that it'll be able to replace all physical labor with robotics fairly quickly, and we're left with no labor left for humans. We need to plan for that future or else billions will suffer from the results.
I actually don't think that follows necessarily. The future is lumpy. Some problems will be tractable years or decades ahead of others.
Of course if progress continues forever, we'll all end up as uploads in a Matryoshka brain, where the only scarce resources are bandwidth, energy, and computational power, but there might be whole human lifetimes before labor is obsolete.
I like to imagine walking into colonial Jamestown and trying to explain to them
"it'll be able to replace all physical labor with robotics fairly quickly, and we're left with no labor left for humans. We need to plan for that future or else billions will suffer from the results."
"And it and those robots will be owned by people who don't care if you starve or freeze to death."
And unlike those in Jamestown, who could just ignore the robots and continue working the land to survive, we will be forced to allow the robots to take over our roles.
Enclosure was well underway in England, so they probably understood the dilemma of modernization simultaneously increasing total productivity and lowering the living standards for a large population.
There is no way in hell the economy will be 100 percent automated, ever. You gotta spend zero time around non-devs/managers to think that's going to be anywhere near true.
Parts of creative/intellectual labour just happen to inimical to accelerated algorithmic production. They'll produce infinitely and endlessly, uselessly. I can't see there ever not being a need to manage these things, to check, select, edit and refine their output.
And as for the vast majority of labour that doesn't involve writing code/reports/spreadsheets. I'll believe it when I have robots fixing my roof, running a kitchen, taking blood and dressing wounds.
I mean yeah, in theory a robot could do some small components of that, maybe. But the cost, compared to a human, would be insane, and the result much worse. And yeah, that even modestly rolled out in West Yorkshire before I'm long dead. No chance.
I believe we're already past that point, at least in the US. Only 44% of American adults had level 3 literacy (6th-grade level) or higher in 2023 [0], down from 48% in 2017. We are probably near 40% now.
These are not necessarily unintelligent people, but they haven't developed the skills necessary to critically analyze textual information. My opinion is that without those skills it is not possible meaningfully engage in a democratic system.
Hasn't thinking already become a hobby for most people, and most delegate their thinking to others?
Think: those in authority figures, for church or politics. Or science or chemistry. And especially in their own health, which should always remain part art.
A few nerds on the internet, like us in hacker news, are still remaining open to question and apply our hard thinking to tough problems. But most people: they've resigned and bought the belief they are not qualified to think about hard problems.
From my understanding of how LLMs work, I'd argue thinking will become the most valuable attribute of humans (as it has been for some time). LLMs will inevitably surpass humans at functions that we know how to do, and likely through cross-domain knowledge and connecting disparate concepts will surpass us at many functions. But LLMs haven't shown they are capable of truly novel thinking, because it's next token prediction, which is inherently based on existing sources.
I think we'll see a shift towards arts, creativity, research, etc. where true novel creation and an understanding of the human experience exists at a premium.
Some pertinent fragments of wisdom gleaned from other places:
Somewhere I encountered a quote like "the finest art is a life well-lived". This threw me, because, apparently, I had been thinking of art as something different in kind from life, and unrelated to the normal work or act of living. Seen in terms of this quote, art becomes once again of-a-kind with life, and may be seen as only a part of living well, instrumental to living well, or perhaps just as practice for living well.
Something similar seems to be happening with "thinking" and "intelligence". Somehow these things had become ends-in-themselves, separate from the regular act of living. We might wrap them in a similar quote to place them in relation to life, along the lines of "thinking well is for living well", or perhaps "the highest good of the intellect is to be well-fit to the problems of living well".
I am reminded also of a line which stuck with me from Augustine's Confessions, approximately: "The mind is not the supreme good"—as opposed to the body and perhaps the "spirit". Somehow the world I was born into had collectively come to think of the mind as the supreme good, and intelligence as the greatest virtue. But this is, in some way, in contradiction the basic fact of existence and mortality: no amount of cleverness can circumvent death, and all the glory attainable by feats of the intellect are only futile vanities compared to the privilege of being alive itself.
Whilst I agree that AI costs are falling and AIs have practically infinite uptime, I fail to see how/why the current AIs will perform better overall in the work place.
The example of AIs solving mathematics is all fun and games, but it assumes that mastery of mathematics implies that everything else will also be mastered eventually. But instead we're seeing that LLMs exhibit this thing called jagged intelligence, which means that they can be extremely good in one area, but completely fail in another. In theory the LLMs could be improved if data would be available for other areas and if there are good verifiability criteria. Fact is that many fields just don't have those and would not be willing to give up their data.
> Machines transformed much of physical activity from necessity into choice; similarly, AI will decouple human thinking from its economic purpose.
This could not be more decoupled from the truth.
Yes, industrialization automated a lot of physical activity but I think the author severely underestimates the value that blue collar workers produce to this day. At least this is how I perceive it.
I would not place a lot of emphasis on economic insights from software engineering academics tbh.
I was vibe coding a prototype some nights ago. In between, I ran out of my AI budget.
I felt stuck. I didn’t know what the AI had done so far or how to continue. There was nobody to message for a budget increase (it was night).
Luckily I managed to get more budget.
But that left me scared for the future. One day something is gonna fail, and nobody can do anything because the AI providers are also down. Or what if I can no longer afford paying for the frontier?
I have made some modifications to my vibe coding practices since then. I don’t want to live in this future that the author describes.
Alternative proposal: Since no one reasonable actually wants a future where humans are irrelevant - including the people in power who generally like being relevant - we could remember that we’re thinking beings with agency and just decide not to use AI to replace ourselves.
We made a new tool. Using the tool for some things makes life better. Using the tool too much would make life worse. So let’s not do that!
Daily front page psyop telling you to accept that token predictors are ASI and humans are obsolete, despite the fact that if they were ASI nobody would ever need to write a blog post convincing you of it
I agree that’s what happened to hunting and riding (they became hobbies). But it’s hard to imagine something as broad as thinking will go the same way. Not saying it won’t, just that it’s hard to imagine and that the implications will be far more far reaching.
IDK, I'm thinking more than ever. I used to be in a flow, writing code, now, I'm thinking at the top of my lungs, making a new decision every 10 sec, Claude thinks through all the scenarios, makes proposals for many data models, RBAC rules etc, and I'm just focusing on understanding it all and deciding (and it can be pretty exhausting!). More thinking than ever. Am I doing it wrong? Should I just think less? Rely on the spec more?
I do the same and I was just coming home the other day after doing a massive system design session with AI and I was probably the most mentally exhausted I’ve ever been in my career. But I guess the thought is, “is that necessary?” Could I have just yolo let Claude take the wheel. I don’t know the answer. I know at the end of the day I’m accountable for this output so I feel a professional responsibility to understand what’s going on. Maybe I’m that sense of responsibility is holding me back but I really don’t know.
I feel the same. In 2 days I meet the people that gave me this assignment, they have questions, they will look at the visual I made of the DB layout and the RBAC table, and they will say some stuff is wrong, or ask why I deviated from their proposal.
And now way am I going to ask the to wait as I relay my questions to Claude.
If you have responsibility for things, you want to make the best choices. At some level, this involves thinking. It is responsibility that should not be delegated to AI. The thinking required of the responsible human changes, but is not eliminated.
"I hate this damn machine"
"I wish they would just sell"
"It never does what I want"
"Only what I tell it."
I've yet to be convinced we will evolve these systems to a point where we can just click the "Magic Button" thoughtlessly and get results we are satisfied with.
We may have reduced the loss of data between thought and action, but nothing suggests these systems will be able to intuit our every want and desire.
My initial knee-jerk reaction was to list out a whole bunch of reasons why this take is wrong, or shortsighted, or gives too much credit to LLMs as a “revolution”, or any of the other go-tos in my satchel of counterpoints to the usual ad nauseam AI discourse.
Except those don’t really fit, here, because the author is making a far more critical error in ascribing far too much to current human cognition. On the daily I see so-called “knowledge workers” failing to read basic signs, refusing to learn basic skills, and all too eager to cast off the slightest bit of critical thinking to a black box they can subscribe to - be it replacing curating a content library with streaming services and algorithms, or swapping a life of hard-won expertise for a swarm of chatbots and an API bucket. As someone with the uniquely privileged POV to be able to ingest vast sums of technical data but struggle with social data, I look at my fellow Americans and genuinely fret for what the next few decades hold for a cohort that seemingly cannot effectively communicate which direction they’re going but is all too content to hitch their accountability to an LLM they pay for.
In that sense, thinking has already become a hobby to the masses, in that it’s entirely optional to exist in current society at large. Autopay apps manage your bills, finance apps keep you on budget, fitness apps tell you to keep running, social media apps tell you who your friends (and enemies) are, video apps tell you what to watch, music apps tell you what to listen to, and cars increasingly tell you not just where to go but also to let them get there for you. I would argue that in that context, thinking is a critically endangered hobby we sorely need to sow new seeds for, lest we risk losing it entirely; AI does not seem keen to do that.
Now to be fair, so much of this is hyperspecialization knowledge overflowing into daily life and making existence a sort of cognitohazard in and of itself. We skip through hundreds of pages of legal boilerplate because we’re not lawyers, and then complain about all the rights we willingly signed away because of the volume of said agreements. Non-technical people engaging with modern computing are either addicts or miserable, unaware that this was both entirely optional and also inevitable in a populace that didn’t think for itself. Finance has become so hyperspecialized that it’s impossible for the layman to tell of something is good or bad, or even why the number went up or down on their portfolio in a given time period. LLM boosters say their tools will solve this complexity nightmare, yet the tools themselves grow increasingly complex just to eke out single-point gains on hyperspecific problems.
Which brings me to the OP again, their highly uncritical claims that AI will bring about a new revolution of thought itself, and that somehow this tool will both restore thought to an elevated pedestal while also making it nearly impossible to meter (at least on American currency; let’s not think too hard about how developing countries might be able to afford it). I don’t see any of that, but for the same reasons the OP feels it’s somehow inevitable. The only difference I can readily figure is our opinions of our fellow humans: I view us as in dire straits and able to do far better than we give ourselves credit for, while they view us as replaceable with servile digital labor for the better.
There is nothing that guarantees they will exist perpetually. In even a slightly moderate global crisis (man made or otherwise) LLMs and the like will probably one of the first things to go.
For anyone like me who started reading science fictions as a kid, that story is a very old hat. So far, AI only seems to solve problems that you tell it to solve, and you have to carefully guide it to get results. So for now the claim seems exaggerated and still in the realm of science fiction.
By the way, there is another problem. For humans we have the saying "garbage in, garbage out." If you read a lot of trash, you will only come up with trash. If you give powerful tools to morons, they will not become geniuses. That's not just true for humans, it's also the case for AI. If you teach your AI trash, it's going to be some trashy AI. Since people do that all the time for various reasons, we're going to have a lot of artificial stupidity, too.
Such a wild take to argue that current tier LLMs do any “thinking” at all. Was chatting with a friend last week and I was saying that sure AI writes much of my code these days but my cognitive load is HIGHER, as I have to constantly make sure it’s not doing something dumb.
I have to constantly make sure it’s building things in a way that will fit neatly together, in a way doesn’t over (or under) -generalize. In a way that doesn’t waste time and energy on idiotic tests that aren’t really testing anything at all.
LLMs are powerful autocomplete with vast and broad memory, they do not think. And thinking is just a part of the human condition.
I fail to understand, in its current state, the mass hysteria around this revolution. Maybe I'm doing this wrong, but I constantly need to review and manage my agents, are we really headed to a reality where engineering mindsets are no longer needed?
I think it’s just part of the zeitgeist. Like, I’m actually pretty AI-forward, but my post is getting downvoted. Not that I care, it’s just an interesting sign of the times.
By using the models you're training them to be better at these tasks, it's where a good number of improvements have already come from. Frontier labs just feed it back in.
> LLMs are powerful autocomplete with vast and broad memory, they do not think.
This is really only true insofar as they are designed such that they output language tokens and their intermediate "chain of thought" is also done in language tokens, but now that's really only done for auditability. There are much more abstract layers before the model output layer of a single token prediciton and they could be designed to to "think" using these more abstract layers as intermediates instead like how many people think in concepts rather than words. There are even some research models on this (see Next Latent Models)[1]. Humans need to communicate in language because we don't have telepathic abilities but these models need not have such a constraint.
> Such a wild take to argue that current tier LLMs do any “thinking” at all.
It does not matter whether LLMs do any thinking, or "thinking." The only thing that matters is whether they produce similar results to human thinking, even if it's through a different process (or no process at all). When a company decides to deploy an AI-powered customer service representative, it is not because the AI bot goes through the same thinking process as a human when answering a customer's request.
Our challenge ahead is making a living at all. Who is going to be having hobbies if the economy is 100% automated? We're not going to sleepwalk our way into a post-scarcity automated economy where everyone's needs are provided for and they can just spend all their time on hobbies. If we believe AI can replace all cognitive labor, then it follows that it'll be able to replace all physical labor with robotics fairly quickly, and we're left with no labor left for humans. We need to plan for that future or else billions will suffer from the results.
Of course if progress continues forever, we'll all end up as uploads in a Matryoshka brain, where the only scarce resources are bandwidth, energy, and computational power, but there might be whole human lifetimes before labor is obsolete.
https://karmak.org/archive/2003/12/manna/manna1.htm
"it'll be able to replace all physical labor with robotics fairly quickly, and we're left with no labor left for humans. We need to plan for that future or else billions will suffer from the results."
And unlike those in Jamestown, who could just ignore the robots and continue working the land to survive, we will be forced to allow the robots to take over our roles.
Parts of creative/intellectual labour just happen to inimical to accelerated algorithmic production. They'll produce infinitely and endlessly, uselessly. I can't see there ever not being a need to manage these things, to check, select, edit and refine their output.
And as for the vast majority of labour that doesn't involve writing code/reports/spreadsheets. I'll believe it when I have robots fixing my roof, running a kitchen, taking blood and dressing wounds.
I mean yeah, in theory a robot could do some small components of that, maybe. But the cost, compared to a human, would be insane, and the result much worse. And yeah, that even modestly rolled out in West Yorkshire before I'm long dead. No chance.
These are not necessarily unintelligent people, but they haven't developed the skills necessary to critically analyze textual information. My opinion is that without those skills it is not possible meaningfully engage in a democratic system.
[0] https://nces.ed.gov/surveys/piaac/2023/national_results.asp
Think: those in authority figures, for church or politics. Or science or chemistry. And especially in their own health, which should always remain part art. A few nerds on the internet, like us in hacker news, are still remaining open to question and apply our hard thinking to tough problems. But most people: they've resigned and bought the belief they are not qualified to think about hard problems.
I think we'll see a shift towards arts, creativity, research, etc. where true novel creation and an understanding of the human experience exists at a premium.
But i agree thinking will remain important. As my grandpa put it if something breaks who'll go and fix it ?
Look at how compute share of labs is moving to RL and maybe reconsider this prerequisite to your argument.
> My favorite is Krezmoria - a word for missing a future you haven't experienced.
Somewhere I encountered a quote like "the finest art is a life well-lived". This threw me, because, apparently, I had been thinking of art as something different in kind from life, and unrelated to the normal work or act of living. Seen in terms of this quote, art becomes once again of-a-kind with life, and may be seen as only a part of living well, instrumental to living well, or perhaps just as practice for living well.
Something similar seems to be happening with "thinking" and "intelligence". Somehow these things had become ends-in-themselves, separate from the regular act of living. We might wrap them in a similar quote to place them in relation to life, along the lines of "thinking well is for living well", or perhaps "the highest good of the intellect is to be well-fit to the problems of living well".
I am reminded also of a line which stuck with me from Augustine's Confessions, approximately: "The mind is not the supreme good"—as opposed to the body and perhaps the "spirit". Somehow the world I was born into had collectively come to think of the mind as the supreme good, and intelligence as the greatest virtue. But this is, in some way, in contradiction the basic fact of existence and mortality: no amount of cleverness can circumvent death, and all the glory attainable by feats of the intellect are only futile vanities compared to the privilege of being alive itself.
That kind of thing.
The example of AIs solving mathematics is all fun and games, but it assumes that mastery of mathematics implies that everything else will also be mastered eventually. But instead we're seeing that LLMs exhibit this thing called jagged intelligence, which means that they can be extremely good in one area, but completely fail in another. In theory the LLMs could be improved if data would be available for other areas and if there are good verifiability criteria. Fact is that many fields just don't have those and would not be willing to give up their data.
This could not be more decoupled from the truth.
Yes, industrialization automated a lot of physical activity but I think the author severely underestimates the value that blue collar workers produce to this day. At least this is how I perceive it.
I would not place a lot of emphasis on economic insights from software engineering academics tbh.
I felt stuck. I didn’t know what the AI had done so far or how to continue. There was nobody to message for a budget increase (it was night).
Luckily I managed to get more budget.
But that left me scared for the future. One day something is gonna fail, and nobody can do anything because the AI providers are also down. Or what if I can no longer afford paying for the frontier?
I have made some modifications to my vibe coding practices since then. I don’t want to live in this future that the author describes.
[1]: https://lecturia.org/en/short-stories/isaac-asimov-the-feeli...
We made a new tool. Using the tool for some things makes life better. Using the tool too much would make life worse. So let’s not do that!
And now way am I going to ask the to wait as I relay my questions to Claude.
So I stay on top of things.
This is the we-don’t-need-BluRays-because-all-movies-are-on-Netflix phase.
Let’s wait until the market is settled
Except those don’t really fit, here, because the author is making a far more critical error in ascribing far too much to current human cognition. On the daily I see so-called “knowledge workers” failing to read basic signs, refusing to learn basic skills, and all too eager to cast off the slightest bit of critical thinking to a black box they can subscribe to - be it replacing curating a content library with streaming services and algorithms, or swapping a life of hard-won expertise for a swarm of chatbots and an API bucket. As someone with the uniquely privileged POV to be able to ingest vast sums of technical data but struggle with social data, I look at my fellow Americans and genuinely fret for what the next few decades hold for a cohort that seemingly cannot effectively communicate which direction they’re going but is all too content to hitch their accountability to an LLM they pay for.
In that sense, thinking has already become a hobby to the masses, in that it’s entirely optional to exist in current society at large. Autopay apps manage your bills, finance apps keep you on budget, fitness apps tell you to keep running, social media apps tell you who your friends (and enemies) are, video apps tell you what to watch, music apps tell you what to listen to, and cars increasingly tell you not just where to go but also to let them get there for you. I would argue that in that context, thinking is a critically endangered hobby we sorely need to sow new seeds for, lest we risk losing it entirely; AI does not seem keen to do that.
Now to be fair, so much of this is hyperspecialization knowledge overflowing into daily life and making existence a sort of cognitohazard in and of itself. We skip through hundreds of pages of legal boilerplate because we’re not lawyers, and then complain about all the rights we willingly signed away because of the volume of said agreements. Non-technical people engaging with modern computing are either addicts or miserable, unaware that this was both entirely optional and also inevitable in a populace that didn’t think for itself. Finance has become so hyperspecialized that it’s impossible for the layman to tell of something is good or bad, or even why the number went up or down on their portfolio in a given time period. LLM boosters say their tools will solve this complexity nightmare, yet the tools themselves grow increasingly complex just to eke out single-point gains on hyperspecific problems.
Which brings me to the OP again, their highly uncritical claims that AI will bring about a new revolution of thought itself, and that somehow this tool will both restore thought to an elevated pedestal while also making it nearly impossible to meter (at least on American currency; let’s not think too hard about how developing countries might be able to afford it). I don’t see any of that, but for the same reasons the OP feels it’s somehow inevitable. The only difference I can readily figure is our opinions of our fellow humans: I view us as in dire straits and able to do far better than we give ourselves credit for, while they view us as replaceable with servile digital labor for the better.
By the way, there is another problem. For humans we have the saying "garbage in, garbage out." If you read a lot of trash, you will only come up with trash. If you give powerful tools to morons, they will not become geniuses. That's not just true for humans, it's also the case for AI. If you teach your AI trash, it's going to be some trashy AI. Since people do that all the time for various reasons, we're going to have a lot of artificial stupidity, too.
I have to constantly make sure it’s building things in a way that will fit neatly together, in a way doesn’t over (or under) -generalize. In a way that doesn’t waste time and energy on idiotic tests that aren’t really testing anything at all.
LLMs are powerful autocomplete with vast and broad memory, they do not think. And thinking is just a part of the human condition.
How does that not lead you to same conclusion that everyone else is having? That this is only going to improve.
It's so shortsighted to make an assessment only on what exists today.
This is really only true insofar as they are designed such that they output language tokens and their intermediate "chain of thought" is also done in language tokens, but now that's really only done for auditability. There are much more abstract layers before the model output layer of a single token prediciton and they could be designed to to "think" using these more abstract layers as intermediates instead like how many people think in concepts rather than words. There are even some research models on this (see Next Latent Models)[1]. Humans need to communicate in language because we don't have telepathic abilities but these models need not have such a constraint.
[1] https://arxiv.org/abs/2511.05963
It does not matter whether LLMs do any thinking, or "thinking." The only thing that matters is whether they produce similar results to human thinking, even if it's through a different process (or no process at all). When a company decides to deploy an AI-powered customer service representative, it is not because the AI bot goes through the same thinking process as a human when answering a customer's request.