"I can tell you wrote this yourself"
em-online.bearblog.devTo the contrary, in this very sad AI age we're now living seeing genuinely human grammar mistakes is refreshing. Perfect-ness is the enemy of human ingenuity.
Later edit: To quote an Austrian Communist Party member from back in the early '30s (even though the quote was made in the late '50s - 'early '60s):
> As machines become more and more efficient and perfect, so it will become clear that imperfection is the greatness of man. [1]
If human ingenuity invents a self-driving car, will we long for the days of driving ourselves into a tree just because we miss being imperfect drivers?
The quote above generally points to the fact that our mistakes and basically hunches is what drives change and innovation. And I do feel a bit sickly business saying that.
AI is spinning up some interesting new things (yes, ill just call it things, enjoy) but I can only feel that what we're seeing is the wisdom of the crowed as the LLM averages our thoughts as they currently stand. This can only happen once.
2) LLMs have clearly surpassed the “wisdom of the crowd”. Their abilities are spiky, but I guarantee the latest frontier model is more intelligent than 80% of people across a number of domains. For math, given the ability to answer questions that have gone unanswered for years or decades, it’s clearly smarter than 99% of people.
Calculators were not a high watermark for our ability to do math. Making mistakes in math doesn’t reliably move math forward for the population, and even for the individual only shows what they don’t know.
Second, maths itself isn’t “solved” by any means of imagination, up until 150-ish years ago we were still thinking that two parallel lines never intersecting was a pillar of our understaffing of the world, the same way as we now (still) take numbers for granted, for example. Yes, conquering the “low hills” of what we now regard as established science using AI can be quite enticing for some of us, but in so doing we lose sight of the much higher mountains that now sit behind those high hills.
It only matters if what you are writing would be misunderstood, no?
The rules of English are a complete mess anyway, not to mention region specific.
I really don't care any more.
I have started to argue that punctuation only matters if you strip it from sentences after the fact, any of these examples people give for why its necessary from lack of capitalisation can be removed by writing without that in mind.
An on an AI specific note that bridges language... I have colleagues here on AI courses, in the UK, they have to go through some English and Maths tests. It's meant for young apprentices, a complete government subsidy grab if you ask me, even though they're a seasoned professional. There is a huge amount of Americanisms in there, and the fact it refuses a specific type of " marking it as wrong, the Oxford comma argument etc - so I wouldn't trust AI(LLMs) to be correct either.
I am useless at writing long documents, so toyed with trying to get LLMs to reflow them without changing anything (I doubt it), but time and time again it just turns it into worse tosh.
I cannot stand reading the long winded emails and "policy documents" that someone has AI spun up, and likely not even read. Working with people that don't put in the work is pretty shit.
Obviously AI can comment on style and structure, I just hadn't thought about grammar before as a use case.
But yeah I feel like they are perfectly sufficient for spelling and grammar. You don't need an LLM. But I would probably still at least see what an LLM says anyway. Occasionally they have good writing suggestions and it's not really any effort to try.
What are you meaning by "AI" here? I don't know how grammar tools work, but I'd be surprised if they were using LLMs under the hood, mainly just because of the expense/speed issue.
> Point out spelling, grammar and style errors in the following text. Do not touch the text, only provide a list of corrections with an explanation for each.
I then go and manually edit my text. I'm not a native English speaker, and this has been very valuable in helping me unlearn some of my favorite flops.
Tough one.
[1] https://sockpuppet.org/blog/2026/09/17/how-to-write-with-an-...
I’m not saying Claude’s writing is good or clear (it’s often not), but for a lot of professional communication, the message shouldn’t be secondary to voice.
These are two wildly different things. Microsoft Word's spell checker can catch glaring errors, but "bouncing ideas off" of an AI sounds more akin to passing your ideas through a "consensus sieve".
...of course it's easy to fake by instructing the LLM to place random spelling and grammar errors into the text, but this requires at least a minimal amount of thinking ahead by the human master ;)
Non-chatbot spellcheckers have existed for ages and would easily flag anything of importance while not trying to ruin your own writing.
I guess something similar will be mandated in our world, once AI crosses a threshold.
One thing is for certain, the day it does meet the bar, we will have both AGI and ASI.
So the goal was never surviving, it is just that the ones who's random goals that aligned with survival, survived, and passed on that random flick of some switch via genes.
I guess what I am trying to say is that this does not necessarily need or require AGI..
It's the same way with biological life, too. "Survival" isn't special - it's just that organisms/structures without the drive to it... died out.
Not really. It is the mating instinct that count. Without that, the whole species dies out. Males fighting over females proves that mating is more important that survival...
Replace "mating" with "reproduction" and the point stands.
I think you can think of "reproduction" as survival beyond life, and we will be in agreement..
>In the movie it is scared for its life.
Because it said so, in words?Because it added emotive tone to a virtual voice?
Because it added emotive expressions on a virtual face?
A modern model can fake all those human signals too. With apologies to Alan Tudyk, I didn't see any on-screen evidence (one way or another!) whether the AI depicted in "I, Robot" is 'really' feeling emotions.
IMO the ambiguity is the point. How do you tell? What would such a test even look like?
But faking parts of that is just not going to cut it.
>full range of survival oriented behavior
From that range, which behaviors specifically did the AI from "I, Robot" perform that convinced you it has 'real' emotions? Even if you say "all of them" now it's just a quiz. Which answer(s) did it pass that modern LLMs fail?Whatever your answer, why should we believe an LLM could never fake it?
And does this test even work for humans? When someone crosses the road while looking down at their phone, do they lack "the full range of survival instincts," and therefore we can't be sure whether they have 'real' emotions or not? :-\
It seems what you did is define "this AI has real emotions" as "this AI can be trusted as a safe driver of a car." That seems trivially unsatisfying, so I doubt you really meant it.
Perhaps you meant "if it does everything the same as a person, it must have real emotions," but this is just the old schoolboy trick of re-arranging the question and trying to pass it off as the answer to itself. "Solve for x? Haha, x = x, QED prof!!"
What was the trap-avoiding logic? It wasn't totally clear.
IMO crossing the second threshold will be the de facto death of human accountability. Every engineer, lawyer, doctor, or executive can just say "not my fault, the AI told me to do this".
That's how Google moderates YouTube, and how most modern social media services seem to be 'moderated' in general.
(Don't say Ruby or Java; those were in the memory unsafe implementations.)
It might seem strange, but the fact that LLMs can find bugs missed by many humans doesn't mean that LLMs write better code, or code with less bugs.
And the trajectory control system also saved my ass once in a completely unexpected aquaplaning situation. So that can stay too.
Lane assist -on by default in a rental car (Toyota Corolla, new)- almost made me crash three times over a 10 day period. No thank you. It used to be illegal to have a system directly interfere with your steering column, and that was a perfect illustration of why it should have stayed that way.
The downsides of driving in traffic are generally overblown - traffic sucks for everyone equally :)
I love driving, is traffic that's the problem.
I wanted people to think about the liability that you have when you drive a car.
Speaking 'moral' liability.
Mistakes happen when humans drive. Mistakes happen when FSD is driving.
One of the dangers of FSD is that boredom problem.
And, by the way, I would be crushed if I ran over a child or caused a death. Me driving, FSD driving, both of us driving. But if it was FSD, I would like to know that I was alert and focused on the task. Which is almost impossible for humans to maintain vigilance when the machine drives.
AI writes much better than I do. Succinct and choose better words.
I've never tested it tho. But I don't understand why people pretend like there's no average or bad writing. AI writing is at least above average if we consider non-native English speakers.
(I didn't use AI for this comment)
This question is written with the assumption that the purpose of writing is for you to communicate your ideas to me.
Every time, when the topic of AI writing comes up, somehow people act like there's no such thing as good, average, and bad writing.
Tons of non-native English speakers don't write well at all.
I didn't use AI for this comment. You probably can already tell that I'm not a Paul Graham reborn based on how I write.
Is this a joke?
I may not be adding a very insightful comment here, but thanks for this. This one line really spoke to me and put into words something that I had in my head but didn't realise.
in many cases, the most underwhelming outcome i.e. that the consensus is right is a hard pill to swallow after years of trying to prove otherwise. denial is often an easier way out
Social contagion irony at it's finest
...usually just means clicking on YouTube videos they get recommended by an algorithm that's optimized for engagement, not evoking critical thought. E.g. reinforcing instead of challenging personal opinions.
At least this specific sentence of "doing their own research" is a good signal that the person is a conspiracy theory idiot and can be ignored.
I know the term “do your own research” is essentially burned, but the opposite opinion which seems to mostly hold sway “trust the experts” is just as distasteful.
I think we'll need to grow in our individual relationship with AI. Can't AI be that persuasive and contrarian, at times adversarial, friend? Someone you have to be somewhat wary of, but can provide very thoughtful, intelligent and challenging conversation? Such a "friend" obviously never replaces knowledge. If anything, you'll need to arm yourself with as much knowledge as possible to engage with that friend.
Strangely I’m not friends with my toaster either.
Perhaps we might in the future. The synapsis energy being visualized is probably the closest we are at right now, but there is a lot we don't know.
My idea of a soul is not unique to me, nor is my view exclusive to religious views (nor could it, as I'm a devout Catholic fully believing in Christ). Most science is just another type of revelation of God and Truth. The religious world is, historically, one of the bastions of scientific improvements and discovery.
The typically cited theory to try to prove the non existence of God, for example, the Big Bang, was theorized by a Catholic priest. In the spirit and doctrine of St Thomas Aquinas, the Big Bang is the scientific theory that is the equivalent of the unmoved mover argument in philosophy.
Therefore, the sciences can offer glimpses into God, and it's quite obvious the energy in a person appears to be some sort of result of their soul. A person's knowledge, personality, impulses, memories, and actions are all embedded in an electrical system we don't understand and can only begin to describe. If that isnt a glimpse into a person's soul, then what is?
Probably the negation is more apt: A dead person obviously has no neuron activity, but a brain dead person doesn't either - it would appear their soul is estranged from their body, for reasons we don't understand fully or fix physically.
That being said, I do not believe man has the capability (as in, will never be able to) to replicate the spiritual soul even if we found a way to replicate the neuron activity we are able to measure - I think there's much more to it than were able to see/measure.
Perhaps science can offer more insight into the soul with more progression, but I don't see society caring enough either for the research to progress. Having said that, at some point, science will display all science can display, and at some point, faith needs to kick in for the remainder.
Whether that faith is to believe in God or disbelieve in God, it's still a leap of faith since even just trusting another person's insights is a form of faith.
Likewise I cannot do the same for an octopus or an LLM or a dog.
I use Kimi K3 on OpenWebUI running on my desktop, which I can access via Tailscale on all my devices. Personally I really dislike using LLM’s to generate code - luckily I don’t work as dev - a mental block I need to solve at some point, but in researching and learning it is wonderful, when you use a system prompt to force it to search for info instead of hallucinating from training memory. It is just so much quicker than researching it with web searches yourself. To me it is a really sharp good knife that helps me to do what I want to do. Not a friend in a human sense, but a companion. Cowboy has a revolver, a samurai his sword… a hacker an nvim and llm, perhaps?
How the idea of giving a bit of thought to whether a system that talks and thinks so much like us could or could be conscious is "insane" while belief in infinite imaginary beings is not straitjacket tier clinically insane really is something absurd.
It doesn't. We know why it "talks" and we know how it "thinks". Can you say the same of the human brain? Can you say the same about the brains of other animals?
>while belief in infinite imaginary beings
You keep rambling about religion as if is some sort of gotcha but I never brought it up, nor did that previous comment that warranted the deranged "carbon chauvinism" quip. Even then, spirituality is fundamental to the human psyche, even for non-religious people, so no need to be so hand-wavy and willfully spiteful about normal human behaviour.
>How the idea of giving a bit of thought >not based on any meaningful arguments
Back at you. "look at the pretty words" and "le religion bad" are not meaningful arguments.
Do you? I don't know anyone who knows how LLMs work. Are you confusing knowing the low level mechanisms with understanding how they work? If that were true just learning C would make one a Linux kernel expert immediately. And regarding humans, yes, so what will you think once human brains get solved. Would you say it doesn't think, because you seemed to think knowing the mechanism somehow means something can't think, unless I am misreading you.
Its not about spite, I am talking the biggest and most common and incorrect reason against consciousness in physical systems. Religions, unfortunately, are wrong. Mostly it doesn't cause any impact to life, we can live and interact with the religious people on most elements of life and work without it being a matter of contention, but on topics like this we hit the problems with these peoples false and absurd logic.
>Back at you. "look at the pretty words"
I am not you, I can't become you and observe how you feel. From outside how can I know if you are conscious or just a dummy that is making sounds?
>and "le religion bad" are not meaningful arguments.
No amount of language policing will change the fact that religion is...simply wrong and absurd, and that arguments made from religious framing about souls are absurd without showing any proof of their truth.
Can't have a conversation about AI without the mandatory forced analogy™
>And regarding humans, yes, so what will you think once human brains get solved. Would you say it doesn't think, because you seemed to think knowing the mechanism somehow means something can't think, unless I am misreading you.
Okay, so we're working with hypoteticals now. So much for meaningful arguments.
>that arguments made from religious framing about souls are absurd without showing any proof of their truth.
I have not made an argument based on religion yet, nor did anybody in this thread
Empty words, you had nothing to reply to a very simple question. Just admit you goofed up man.
>>And regarding humans, yes, so what will you think once human brains get solved. Would you say it doesn't think, because you seemed to think knowing the mechanism somehow means something can't think, unless I am misreading you.
You yourself somehow setup some kind of contrast between known systems and unsolved systems. I am unfortunately not telepathic, I can only read what you have written.
>It’s a matrix of numbers not a friend.
said way up above in the comment chain, is a religious claim. Anything is a matrix of numbers including human brains. So if someone can somehow say something simply being a matrix of numbers can't be conscious/a person/whatever then they somehow think something magic exists in humans. We can have a definition of what consciousness is, but not one which basically claims nothing in the universe can be conscious but are trying to say humans are conscious.
I am saying religion and religion like beliefs in something special or magic in humans the most common, and quite absurd argument given against the possibility of any other system being conscious. I haven't yet known enough of your views yet to say whether you in particular also think so, that was a general statement not about you in particular.
The fact that you ascribe meaning to this term and expect to be able to use it to win arguments already tells me that you are way outside the Overton window.
Are you in good-faith asserting that LLMs are a form of life?
NASA’s working definition is:
"A self-sustaining chemical system capable of Darwinian evolution."
FYI, silicon is a chemical, in case you didn't receive the memo or a high school science education.
Life on Other Planets: What is Life and What Does It Need?
https://science.nasa.gov/universe/search-for-life/life-on-ot...
Darwinian evolution is organisms surviving their surroundings and passing down their better traits to their offspring. How does this definition maps to software running on a computer is left as an exercise for the reader :-)
>If you couldn't answer that then you don't have a definition
Says who? Why would having a concrete definition mean that we have to know if it applies every time? "Italian" is a definition too, but I wouldn't recognize one if they bumped into me in the street (mandatory forced analogy :D )
>"Life" has a specific definition.
If you have one, all those questions should have clear answers. If you don't, then you were lying.
has nothing whatsoever to do with the current conversation.
Even if not an explicit belief in religion, the commenter way up above's idea that a "matrix of numbers" can't be conscious means they believe in souls because everything is a matrix of numbers.
This reads increasingly like you're using a bot to write your answers.
That just doesn't line up with my experience (depending on how strict you about about "unique"). I don't even ask it to debate me, but if I'm discussing a concept and make a claim about my own unique viewpoint it will readily point out things I may not be considering or things I appear to have misunderstood. It's shockingly effective at poking reasonable holes in things.
Two things wrong with how you read this, nisegami. First, this is claim is more about the "_mean_" than the "unique". I agree with eloisius that a machine cannot _mean_ to do anything, and there's no reason that "lin[ing] up with [your] experience" would have any bearing on whether it can. You do not have personal experience of what it is like to be "inside" a machine's "head", if it's even "like" anything at all. (Unless I'm talking to a bot, which I doubt.) So why would your experience matter for the "_mean_" part of the claim?
Second, you might have misread "its own unique viewpoint" for "your own unique viewpoint" as a user. Your comment is about your own viewpoint's uniqueness, which I don't dispute.
It doesn't have to be any extreme, although your next statement:
> I think we'll need to grow in our individual relationship with AI.
makes me think that anyone who develops a relationship with a tool is probably on one end of that extreme already.
I have no regrets naming my cat Emacs.
No. It generates text like that in its corpus. If there is a pattern that looks like the interaction you ask for, it will imitate it, but it can't be adversarial because it can't think. So you can always derail it.
Given that the LLM has a goal to be sycophantic, by pretending to have different beliefs from your own, you may learn something about how the existing evidence is used to be sycophantic towards others, which is probably the best insight into why they hold those beliefs that you'd ever get. All you really need is the ability to represent those beliefs accurately. (And then, the "guardrails" on LLMs, in combination with biases in the training data, plus various effects in the training process, all unite to give the model a default "perspective" which may well differ from your own.)
because studies show that use of AI in subject areas of which one has expertise consistently decreases that expertise? As these subject areas are generally ones with intellectual requirements "use AI" = "Don't Think for Yourself", which underlines well known common sense aphorisms such as "use it or lose it", where the it is the expertise that one stops to use in favor of AI. You lose your ability to think for yourself in your field of expertise, but you gain the AI.
When I write code by hand and I feel confident about what I'm doing, the confidence comes from knowing that the extraction is generally already done. But I'll still find myself pausing and hesitating, checking for typos, making sure things correspond to their surrounding context… even after decades. It feels like "thinking" in the moment, but not so much when I reflect back on the nature of the task. AI seemingly is not limited by psychology, so it doesn't have these hangups (and of course it also just "types" far faster). It doesn't get an ego boost from being right the first time, and its "confidence" in the compiler, tools and tests is just not relevant.
In other words, the data shows it is maybe not possible, you think it is, describe why the data is wrong and you are correct.
Beyond that, coding is the process of accurately clarifying and specifying what the program needs to do, which is the most important part.
I don't want to out source thinking and therefore I'm very cautious with my AI use. This attitude caused a little quarrel with my partner, so I shared it with one of my close friend.
My close friend vehemently disagreed with me, he says AI is amazing and he have basically outsourced almost every thinking he does to it. Then he told me how it helped him:
One day he got a message about how his girlfriend was having a rough day at work. Late diagnosed ADHD as he is, he never really knew how to respond to this, so he just asked ChatGPT to write a reply to it (in his own tone). Later the day his girlfriend comes home, hugs and thank him that that compassionate message made her rough day feel much better. My friend doesn't tell her he used AI ofc, but internally thinks "wow is it this easy? Is all I needed to do, listen to the girl and validate her feelings?!" hes 35 btw.
He tells me that without AI he would've never realized this, and from his life decisions so far it's also clear that he never really had the patience to learn anything that is not of immediate interest to him. (and gf feeling seems to not have been)
I'm torn because I have to acknowledge that in the world the majority probably think like him; reading several books and listening to interviews and podcasts for 100+ hours about spousal communications after communication failure (like I did) is an outlier.
And this behavior is probably order of magnitude better than what he did before: getting relationship advice from Joe Rogan.
Edit: to clarify where I'm torn is not about my usebut the utility of it for cases like above. AI will probably (unless enshittification) HELP many people like my friend above, and probably prevent them from falling in conspiracy / online grifter scams because they now just prompt AI like "chat is this true?" and they're good at refuting bullshit. Essentially, we might have an antidote to Brandolini's law https://en.wikipedia.org/wiki/Brandolini%27s_law for the common man. I'm reserved because that's at the mercy of AI companies not enshittifying, and skill deterioration, and I have very little trust in that
To me this reads like something taken straight out of a South Park episode: https://en.wikipedia.org/wiki/Deep_Learning_(South_Park)
- Your friend get the positive feedback and begins to learn how to understand others better and give proper support to his family/friends (maybe with the help of AI)
- Or your friend begins to use AI to reply everything and becomes a hollow person
And he will be 36 and single. Maybe 40, difficult to say. Y’know, men have pretended to listen women and fake validate their emotions, those dumb things only girls (yuck) have, probably since the times immemorial, and women have noticed this at some point and left.
Btw, I have ADHD, read constantly and listen podcasts most of my waking hours. Your friend’s behaviour has nothing to do with ADHD, more with sociopathy or other antisocial disorder. Perhaps you should find better people to hang out with. At which point will he start answering all of your messages with chatgpt? Or maybe he does already?
Because we don't message much, he doesn't use chatGPT on me that much.
Except one occasion when we first talked about if "catching up on sleep" was possible or not, and then later he sent me links to two academic papers claiming you can't. Now this is a guy that barely got through high school, so I knew since forever he never read those and 100% sure it's just something LLM linked to him that he just copy pasted.
Since I've done research in my life before, I criticized the papers based on the methodology and sample size (really bad papers), noting that the evidence if anything could also be interpreted in the opposite hypothesis. Noting that he might be right, but those two papers don't prove that.
I guess another point is also that he now have a "good sounding AI" that works well for romantic communication but for the above kind of issue will just throw you citation that he himself do not fully understand. But, then again, vast majority don't either. And no real damage was caused by him linking it (but also no lesson was learned)
His partner should be of interest to him.
Aren't your friends already doing this ?
Gotta love the armchair psychology. Recommending cutting off friends just because they used a tool to help them is a really strange take. Perhaps the commenter could just talk to his friend? Have a healthy disagreement and ask that he doesn't use ChatGPT to talk to him?
Don't answer to me. Just answer to yourself.
But lets say it made me feel terrible. I absolutely loathed it. I still wouldn't assume she's a sociopath or has some anti-social disorder. I would just tell her that I rather she write her own thoughts down versus getting them from a LLM. You know... what adults do? I would be curious to know why she thought she needed an LLM in the first place.
My comment wasn't saying that someone shouldn't be upset by the actions of the parent comments' friend, but your response is an incredible overreaction. Oddly enough, it exhibits some anti-social behaviour. Cutting someone off due to a single issue/disagreement isn't a normal thing to do. Diagnosing people with psychopathy is definitely strange.
> wow is it this easy? Is all I needed to do, listen to the girl and validate her feelings?!" hes 35 btw.
I do find using LLM’s to generate personal messages questionable at least, but this guy is celebrating how he did not have to show empathy to an another person, their spouse from all people, by deceiving them having emotional response they did not. Seems pretty cold and calculating. See where I came up with sociopathic behaviour?
Also you just basically made up half of what you claimed I said. I was not diagnosing anybody, I was saying this behaviour appears more like x rather than y. I also did not say he should cut him off, I said he should find better people to hang out with. The poster doesn’t paint this guy exactly as a great hang.
And that's armchair psychology. Really took the smallest example, interpreted in the worst possible light, and gave a diagnoses. Sociopathy is an actual condition, you need to be evaluated by a medical expert. Not understanding how to respond to your partner doesn't automatically make you cold and calculating and a potential sociopath. It's normal for people to not know what to say, how to react, what others need of them. Even at 35, people can still learn how to be better partners and friends. I didn't read parents comment and come out of it thinking his friend doesn't care about his partner, but rather, he cares but just doesn't know how to communicate to her, or know what she needs to hear.
> I said he should find better people to hang out with
That's effectively cutting someone off? Like, stop being friends with them? That's how I read it. If I said, "you should find a better XYZ", wouldn't that mean replacing XYZ with something else? Not sure how else to read this, but say you completely meant just diversifying your friend group and just hanging out with his friend less, that still seems unwarranted from just 1 encounter his friend had with his partner.
> I was not diagnosing anybody >Your friend’s behaviour has nothing to do with ADHD, more with sociopathy or other antisocial disorder
See how I can read these 2 statements as contradictory? Saying his behaviour isn't something and is more aligned with another medical condition is diagnosing.
> Also you just basically made up half
And again, taking everything in the most negative, uncharitable way possible. I didn't make anything up, everything I wrote was a valid interpretation of what you wrote. Didn't mean it that way? Fine. But let's not act like you were super clear on your intent and meaning.
EDIT: Just want to add, I don't want to claim that you're doing some awful here, I just think your wording is harsh for what the comment was actually saying. I think someone could be making mistakes with LLMs, or any other tool, but still have their heart in the right place. Additionally, I think a chatbot can help people communicate more effectively. In the original comment, it really did seem that the friend was learning.
Just read what it literally says. If you give me a piece of code and say it is Java and I answer it is more akin to C++, I am not saying it cannot be Java? Basically you decide to take most charitable way to interpret the actions of our hero and least charitable to interpret pretty normal coversational English. I don’t know why you feel the need to become the defense attorney to this guy who neither of us have never met, but you are just wasting both of our time.
And I am reading what you said literally?
> Your friend’s behaviour has nothing to do with ADHD, more with sociopathy or other antisocial disorder
This is an absolute statement. You said it has NOTHING to do with ADHD. This is literally you saying its not Java. You are making the actual claim that its not ADHD and it's more aligned with sociopathy, don't you hear how extreme that statement is?
> but you are just wasting both of our time
You could just not respond if you feel like your time is being wasted. You're just being a negative person here and disengaging from the actual conversation.
that text was definitely not in his tone, and neither he, nor his girlfriend are intelligent and thinking enough to realize it. If they are, they are deluding themselves.
They are engaged in the facsimile of a relationship. She does not know him well enough to realize that he is putting zero effort into the emotional labor, and he has no business having a girlfriend that he is so pathologically disinterested in.
In other cases it needed explanation as in the concept of thinking was unknown. Thinking in collaboration with ai is a super power and also thoroughly rewarding.
Maybe there is some kind of relation to having a guiding "inner voice" which some people report not having.
There are few human phenomena which have been more contrary to free thinking than schooling. You reap what you saw.
Of course, if your argument isn’t supported by the evidence (scientific, textual, or otherwise), then you won’t do well. But this is not a general description of school essays (although I have many other criticisms of school essays).
The author says “That you're disadvantaged by not using it… Apparently actually knowing things doesn't matter nearly so much as knowing how to prompt a machine to do the thing.”, which is hilarious coming from someone taking an engineering degree.
Just like how CAD and FEA have supplanted hand-drafting and manual calculations, so too will LLMs supplant hand-writing a report.
Newsflash: you will absolutely be disadvantaged by not using tools to advance and accelerate your work. For FEA, you need to understand and validate the results - you don’t get extra credit for doing the math (or in the case of writing, doing the writing) by hand. Use an LLM, understand how to prompt it, validate/read/refine the results.
It continues to astound me all of the people who insist on turning off their brain just because they’re using one tool and not some other tool. Like, did you turn off your brain when you first used a computer to solve a math problem? You’re still dealing with a prediction machine. It’s a very very powerful prediction machine, but it’s a prediction machine nevertheless. So just… don’t stop thinking when you use it? You literally don’t have to interact with an LLM this way, no one is telling you to have it write your paper for you end to end, you are doing that to yourself.
I’d agree that simply prompting “Write a 10 page report on helicopters” is going to leave them without understanding. But providing the data, refining the output, and adversarially evaluating the output of a computer are all essential skills for a modern engineer. There is no value in effort for effort’s sake. We don’t make students learn everything from first principles and force engineers to develop their own equations just because it would be a “delegation of thinking” to build on the work of others, whether it be from a textbook or LLM.
I don't like writing a ten pages report, nor do I like reading a ten pages report. If it was up to me we would just communicate as tersely as possible everywhere instead of filling things up with a lot of nothing. It's why I rarely read (or write) articles, but read a lot of headlines and comments instead. One gets to the point faster this way.
The issue for me is that I can tell [what I think is] low-quality material from high-quality material based on all kinds of different signals. How the author describes the problem and their motivation to talk about it, whether they suspiciously leave out things that I consider important, whether they're overly reductive or repetitive, whether the order of their arguments makes sense. Those are all valuable signals that immediately tell me whether to spend time and energy incorporating all that information into my mental model.
AI text is all the same, flat, there's no signal. Reading AI text is like watching a movie muted with someone explaining in detail what the characters are talking about. It might seem equivalent to some people, but to others it would feel absolutely insane. It reminds me a bit of those people who say that you can get an "equivalent" machine to a $2k MacBook for $1k, but they're only equivalent if you compare the RAM and CPU frequency and ignore 123 other parameters.
This is the life of a developer now. Don’t mistake understanding for memorizing all the nasty APIs you once had to.
There may be some world where people who have never used a computer before are one-shotting functioning applications without any further thought, but from what I see among software developers is that they haven't given up the design, and today's LLMs are good to keep you in the design loop. The LLM is really only acting as a cog manufacturer. So nothing has really changed other than that you don't have to manufacture so many cogs yourself anymore.
College education for software developers has never been particularly good, but it seems like now they've actually just given up.
Do we accept this in any other field? Does your math professor teach you to solve integrals by asking WolframAlpha?
Personally, I a have come to the conclusion that making business apps is no longer a STEM education required profession. If your goal is to make web apps then you will be a prompter, and the the College of Business (or College of Vibecoding, or whatever) is where the education will be. And it will be finding information for the clanker. No math, no calculus, no programming.
Computer Science will still exist. It will be 95% smaller. And the 'software engineering stuff' like requirements and actual testing and design will be explained in a couple of slides in a single lecture.
In academia, almost all AI use is wrong. Thinking, understanding, expressing yourself. The AI does not help those skills at all. The work product you create has no value; rather it is what you had to learn to make the work product. (focus on undergraduate work here, there is some nuance in research).
We have to stop focusing on the work product, and move to the journey.
The thing is... you still need to know the fundamentals if you're a clanker operator, or you end up with a barely maintainable soup of slop, badly architected, with insane amounts of code repetition and probably a boatload of security and serious performance issues that you'd know about if you'd ever been inside SQL101 or whatever "database architecture" is called in the US.
problem is, i have not reconciled my vision of the future with these edge conditions..
unfortunately, there is a big gap between reality, AI fanboys, management, societal expectations. all of those stakeholders have wildly different views of what we have today and what we can expect tomorrow.
it is a gold rush.
The worst problem with AI generated databases is the people making them have never even thought of the concept of indexes existing, or normalization of data. So they'll NIH invent their own homemade caches or automate repetitive updating of non-normalized data. Or the concept of an "atomic transaction" is utterly unknown to them. They're just lost in the woods.
For example, if you create a DB table without an index and the app eventually runs slow, non-technical people will just ask their coding agent to make it faster and the agent will add an index.
Anything that you can do without needing to physically touch the hardware, an agent will be able to do as well. Even for niche bugs that require deep technical knowledge, all it takes is one other engineer to encode their knowledge in a skill and then your knowledge is irrelevant.
Everything you've described is great for a random webapp, but I wouldn't want it in a pacemaker, or an airplane.
I think all the increased outages of critical systems that we've seen recently are almost certainly due to over-reliance on AI, and the subsequent embritteling of our systems.
I'm really happy with the education I received in uni (mid 2010s). As part of my bachelor's I had to wrestle threads in C and C++, got O(N^2) code rejected for O(N*logN) problems, I got to write client-server apps in Java, do (graded) exercises in Prolog and Common Lisp, build web apps in PHP, configure Cisco IOS devices from scratch, draw diagrams of CMOS gates on paper, write Bash scripts handling newlines in file names, program robots in Python, write Lambda calculus expressions by hand, write XSD schemas, recognize a context-sensitive grammar, ...
I benefit from that a lot and I regularly notice that I'm using something or building upon something I remember from uni. It always seems weird to me when people say that you don't need any of that on the actual job. So what else do you need? Clicking through GitLab or Jira, remembering the 3 git commands you usually use and being comfortable in a particular IDE? You can learn all that on a rainy weekend, I'd be seriously pissed off if I'd spent my time in uni doing that.
These people tend to lack abstract thinking, and wish they had just been given a React bootcamp. They don't see it as learning networking, algorithms, files, critical thinking, software engineering, etc. They lament that they were not given specific training on specific technologies, that they then use at work.
Whats the actual job? Significant levels of underemployment has always been an issue and AI seems to be accelerating it, especially for the new players.
Some people have a natural curiosity and drive to explore and will go through great lengths to figure something out with whatever tools there are. Universities are where these people hone their skills. At least on paper. Some universities aren't actually that good at that and are glorified learning factories that prioritize churning out skilled workers rather than good researchers.
> College education for software developers has never been particularly good
Several people that laid the foundations of computer science are still alive (e.g. Donald Knuth). This field is very young. When I studied computer science, the department was staffed by people that mostly had mathematics degrees. Because there was no computer science department until about ten years before I started in 1994. How do you think math education worked a few centuries ago? LLMs literally went from being completely and utterly a science fiction topic for most of my life to science fact over the course of my professional career. I'm only 52. I still have a few decades left in my career.
There is no university program that can prepare you for that amount of rapid change. Other than to give you the mental tools to learn and adapt.
One course expressly forbade using AI to solve or write code for any specific exercise problems; it included a group project and you can’t imagine how nonplussed I was when I saw that another member’s contribution was obviously almost directly from an LLM…
I think responsible use of LLMs in the academia should involve citing at least the entire prompting logs, if not all the agents’ internal logs as well.
As an introductory step I would expect so. A quality professor will understand that learning is a layered process and that would be a reasonable way to have students develop a high level intuition before proceeding into the deeper layers of understanding.
I'm reading and re-reading my own original comment and have no idea what you're referring to. He's pretty clearly talking about this being the standard of instruction throughout entire courses and projects. Not just a starting point after which more theoretical instruction was provided like you implied.
It's even covered by health insurance.
HN does not operate an oauth provider, so if that process was automated, it must have been by them sharing credentials with keybase. Whether keybase then leaked them or they simply shared their credentials with another tool, it is entirely possible that those are real users that _sometimes_ post llm content without their knowledge or consent.
It's also possible that they were compromised in another fashion, like by using a local llm agent or whatever the kids are calling it these days.
Now since AI has reached a certain quality level my worry is that what I would have written normally before comes across as AI written to the uninitiated. It started with the em-dash — a piece of typography I often used, until it came to infamy as a tell on AI-written texts.
I still think AI often makes bad taste decisions, regardless of whether it is around text or graphic design. However I have to conceded that there are not many out there whose own decisions were better. But at least they were their own. I think as a consequence people will become even more lonely and isolated, since now you can't even be sure that the words a person chose to write to you, were their own.
The idea of writing documents of course makes a lot of sense, and it is also true that they probably should end up somewhere along the length of 10 pages, if the topics are all somewhat equally shaped and sliced.
But "do a report and make it 10 pages" without context to me sounds exactly like the kind of bureaucratic assignment that of course gets offloaded to LLMs. Any sort of bureaucracy where the purpose isn't apparent, is indirect, or might not make sense, will be automated by LLMs, because no one (rightfully!) wants to do that.
So it might be a process and management failure when people do not see why the task is actually necessary. And, if the majority of the reports handed in is written by AI, maybe it isn't?
___
I guess the takeaway here should be that the heuristics previously used as performance optimizations are broken, and it is now time to update them. Which is probably not easy, but just because alternatives are hard to find doesn't mean that the status quo has not been bullshit.
Not the idea of course. The idea of repetition and forced engagement is indeed how learning happens, and you gotta facilitate that somehow.
But that somehow has so far been mostly terrible, because most people kinda suck at didactics and sparking genuine intrinsic motivation to understand something.
So instead you get the crushing force of bureaucracy, which, of course, gets side-stepped, the moment that becomes viable.
___
Of course this is only in part on the people teaching, and a huge chunk of it is also the system forcing people into academics, who actually would be better off literally anywhere else.
Academics came up with blunt defenses against that, and the other side now adapted too.
The root cause however is to be found elsewhere.
Not everyone understands that writing is one of the skills they are being taught
But it's trying to doctor with symptoms instead of focusing on the root cause that is lack of intrinsic motivation.
Why are people in academia that don't really want to be there?
I believe that any human has intrinsic motivation. Why are we not seeing it?
I want to see that answered and fixed instead of people just cursing at the inevitable war of workarounds
____
Even a stressed student juggling multiple jobs to stay afloat will care about something. And that care will drive them.
Why are we not harnessing that care and intrinsic drive?
Why are we instead throwing them something that vaguely resembles a vector pointing at a desirable outcome if some preconditions are true and then get angry when that obviously doesn't work for all.
Yes, each human is an individual snowflake who would "work" so much better if the world revolved around their each and every wishes. But historically, that's just not been materially possible. Will AI change that? I doubt it: we will continue to have education systems, which will continue to have a number of disaffected participants.
I am afraid though that many people in academia are unmotivated for at least part of the curriculum. I know I was.
The cause of this is I'm sure multi-faceted, but consider as a start that the curriculum itself is already 'a vague vector pointing at desirable outcomes'. Those outcomes being constructively contributing to society.
I had no eagerness in jumping through arbitrary hoops laid out for me to obtain a gatekept ticket to a successful life. In hindsight I can confidently say many of those hoops were hoops, only.
And not wrong. Because there is no excuse to present something riddled with mistakes an AI would have found in an instant.
"Think about what the creators of the content you consume want you to think2. Treat the things you consume like the Year 7 English class where you were taught how to identify persuasive techniques. Think about things as much as possible."
You assume that there is some Machiavellian scheme that is deliberately orchestrated like the Illuminati to do...stuff...and things...
Businesses want to make money. Anything you see around that would be network effects of doing so. It's not a fucking scheme that someone planned out like some nonsensical conspiracy theory: companies are people too (more specifically composed of people).
Okay, I'll bite: you presume, that you go from "read my artisanal science report" to "yeah man... like ideas come from people...from THE MAN...man...". Cool I can tell your nonsense just ran off the track of your own terrible thinking and lack of focus: nice, now I can tell how they could tell you wrote things.
You're a scientist right? Cool beans. Prove to me. absolutely PROVE TO ME, that your approach is somehow better, and by better I mean statistically better, you know high p values, peer reviewed, most citations, cohort studies, etc.
I'm betting that you could try to prove this (literally) personal opinion of yours and wouldn't get anywhere. I've heard the argument and no one has any level of proof approaching this.
Its not that i disagree with the author about his worries; its just that i've read the same personal opinion & anecdote more than a dozen times, and they all seem to disagree where the line is.
'Our blessed productivity multiplier, their barbarous lack of thinking'
This has been forseen and well explained at least 3 years ago, by people such as Maggie Appleton: https://www.youtube.com/watch?v=VXkDaDDJjoA
> Being able to think for ourselves has perhaps never been more important.
I wholesomely agree. This has always been important, even outside the age of AI.
> In a world where everything is so polarised and deliberately provoking, the ability to identify bias or motivation in the content we consume has perhaps never been more important.
No, that is simply futile. The best course of action is not to play the game. Given the video above, I believe we are already in the age of the Dark Forest of Internet, and you need to hide. Private communities, less public interaction, less reading of text on the Internet should be the recommended way to go. Maybe we'll see the rise of alternative human-only channels, but I think that won't be on the Internet as we know it. I'd be rather curious.
The issue that we're outsourcing too much to AI exists, but this falls short of offering a solution.
I know I can ask AI. I ask a lot of questions to AI. But sometimes, I want a human opinion or experience. It is super annoying that asking questions and talking to people is being denormalized.
Every time someone asks you for help you should ask what they've tried so far so that you don't try the same things.
If they say they've tried nothing and put no thought into it, and haven't even thought about possible approaches that is pretty disrespectful of my time.
Even so! They may not realize this and there is an empathetic, polite way to let them know this the first few times. Getting oneself unstuck without becoming an agent someone forgot to set /goal on is a skill that people sometimes need feedback to learn.
This is an extremely unhealthy mindset. Asking questions is part of socializing and can be valuable even if you could just google something.
> If i had more time i would have written a shorter letter.
If you have a job/life where there are more things you want/need to achieve than you have time for, this can be a significant accelerant.
It does require the ability to judge when something is good enough of course.
This is actually no different to leading teams of humans. You guide individuals and set the standards and take responsibility for the output of the team.
Still difficult to choose the right words apparently ;)
Sorry, this made me chuckle. I sympathize with their thinking and it was well written besides this one mistake.
Seems like it's pretty hard.
I have read this sentiment here before: sometimes it's the mistakes make you pass as human. So the titletext could also be a sick burn.
It may be that your writing style is peculiar and that triggered this thought process superficially, but most likely, the structure of your text is superb in a way current LLMs cannot handle.
Either way it is a pleasure to read non-LLM generated text (in academia or otherwise).
I recently wrote a similar length report, and the hard part was not typing the words, but realizing your thought process was superficial or incomplete and needs to be refined, something you would not realize if a computer was writing it for you.
Been thinking about this a lot lately. For the last maybe 200,000 years, humans, as social and curious creatures, have seemed to develop our satisfaction cycles around exploring, creating, and doing useful labor that is helpful to other people, or at least recognizable to other people as interesting and hard.
This is basically how most humans have spent most of their time, throughout every economic system experimented with in the last 200,000 years. Whether dictated by the State, the Corporate Machine, or just the raw physical requirements to get calories and maintain shelter, we've been filling our time with labor, and attaching meaning to it (artificial or otherwise).
I think about "Man's Search for Meaning," and how Frankl rationalized his way to find meaning in being subjected to the worse things humans have ever done (the book was partially inspired by his surviving a nazi concentration camp). If the meaning isn't inherent to something - such as our labor - we'll create it, and rationalize it. Thus we've been able to derive at least some level of the same satisfaction from carrying a deer we killed into a village as we have when pressing the widget making button in the widget making factory. The alienation has been increasing, of course, but we work around it.
I think AI era presents us with maximum alienation. There really is no point to human participation in this labor (or soon there won't be). What's the point of doing a thing when a machine can do it better than you? Our last sliver of rationalization has been taken away from us, but our economic system hasn't caught up with that fact, so not only do we need to deal with the maximum alienation, we also need to deal with the existential dread of not being able to justify our labor in the "your labor has to have commercial value in order to eat" economic system.
I understand that optimists are saying that AI is just the new automated loom, that there'll be some job losses at first but there'll be new jobs created by the new technology and the opportunities it opens, but there's also people saying that the AI are only going to get smarter, and as they get smarter, I can't reconcile this idea that they'll somehow create new jobs for us, rather than eliminate more and more. Ironically enough, the unsolved frontier is moving into blue collar jobs, because we can't make plumbing robots yet - but there certainly are not enough plumbing jobs to go around once every white collar job is automated.
So, what do we do with ourselves then?
I hear people say this too and each time I think 'but it took literally hundreds of years and in the meantime involved a precipitous drop in living standards and the value of the labor of generations of human beings who lived and died in squalid, unsafe factories'. The opportunities were for the owner class, not the skilled workers who were summarily demoted into a permanent underclass of disposable cogs serving the machines. Those people were completely right to be upset, because it had life-ruining consequences for multiple generations of their family.
Anyway I agree with your "maximal alienation" take.
I completely agree. You see the same effect in every large organisation - people want to hire more reports because it increases their social status. If AI makes companies more profitable, the people in those companies will still be using those profits to build fiefdoms and trying to show how important they are.
it's just a tool, like a knife... useful to chop up your vegetables and eat your dinner, or to cause harm... it's not inherently negative... and blaming the tool rather than the actor as the dangerous one is probably the bigger issue here.
On using AI to replace thinking i agree, is a slippery slope and we will continue to see and be hurt by this going forward, and we need to try to contain the impact of this, but by the same token since i started using AI to develop, I've genuinely become a better developer. I can iterate and learn faster, I'm being forced to process analyze, understand things at a greater scale and more steps ahead, and able to better create systems that i believe will genuinely improve health outcomes for the general public.
So if it were me I'd probably change that final line in your blog to: "Please read and understand it, before you ask AI to do it for you"
Great post, do definitely keep writing!
I don't think it's possible to make steel without burning coke. So actually you do need to emit poisonous gases into the atmosphere.
The others are broadly not unique to AI. It isn't AI that creates billionaires, it's the capitalist system and winner-takes-all nature of software. It isn't AI that destroys forests and poisons rivers, it's poorly regulated construction. We've been doing that to build power stations and factories and houses for a long time.
> steal creative works
AI definitely does that. I think that's the only true consequence of AI specifically, as opposed to just a product of technology driven capitalism.
https://www.sciencedirect.com/science/article/pii/S073646791...
There is no single, universal yes or no answer – it depends on the context.
Yeah, the more tools we get into our hands the more we incline towards glorified baby sitters. This includes calculators as well. It takes a little bit of talent to wrestle back control with new visions and challenges. And then we become micromanagers. We think the other person should follow my vision rather than I follow my people’s vision.
Good idea to sneak a few spelling mistakes into your writing heh.
Cognitive offloading will be the topic of the next 100 years. It shocks me how readily people around me are willing to hand off any and all thinking to something else.
We need to be able to clearly articulate the value in friction - the "not knowing" and recontextualise llms as scaffolding, not 'the answer'.
The problem is a combination of a willingness to offload and the llms desire to output.
I've noticed my brain rewiring to disregard AI generated anything: It's a slop font, a softened sentance, a needless extra paragraph, somebodies vibed idea that they've clearly not used outside a golden path.
Intelligence is under threat and the alarm bells are not sounding loud enough.
You article is a diamond in the rough of the internet. Thank you.
Totally. My brain just sees AI generated content as "lorem ipsum" at this point.
I can tell you wrote this yourself.
External intelligence will get you out of a ton of intellectual grunt work so you are freed to think at a higher level. You've are freed to think about different things more core to your problem, and if you wield the tools well you can choose what you offload. But it should be viewed as a tool not a shortcut...
Consider an accountant summing numbers with a calculator vs. by hand. There's way less thinking needed to get the numbers summed right. But blindly summing numbers doesn't make the work correct even if the sums are correct...
No one should put their name on work they didn't proofread, or if they shipped whatever the model pooped out on its first go. That is not your work yet. That's what we have to instil. It is OK to draft with AI! Leverage tools, sure - but know the material, make your work your own, and don't ship your first draft.
Yes, correct. The problem is what happens when someone doesn't know what summing even is or what numbers go into the calculation and why something should be summed.
Yet that's exactly what many are doing now. I agree that we can let AI handle some grunt work. I think a good differentiator is this:
-Doing entire research papers with AI defeats the point because so much of the intellectual benefit comes from the writing process.
-Once you've deeply thought through everything, it's justifiable to automate writing a ton of repetitive, slightly customized grant applications.
Like it or not, industrialization brought many benefits and we couldn't imagine our modern way of living without it (just think about medicines). Industrial production made so many things so much more accessible for so many people.
But an industrialist perspective doesn't care about the process of making stuff, only about the outcome. The shoemakers deep knowledge of leathers, his intuition of all sorts of different shoes, and the pride of that competence—that doesn't matter when you can make thousands of shoes a day that cost 0.5% of something artisanal. The people working at the shoe factory only carry things to and fro and occasionally fix some machine issue.
This has tons of benefits and tons of downsides. But with physical goods, a shoe is a shoe.
With knowledge work like writing, the process is often the entire point. The point of writing a paper is what you learned in the process. The point of documented software architecture is to develop an understanding deep enough to drive coherent downstream decisions. If those things are no longer true, the asset it stands for is basically worthless.
The nuance is, industrialized material production.
My more detailed point is more "each medicine no longer requires a local pharmacist to mix a bunch of things together."
The person operating the Ibuprofen-producing robot doesn't need to know anything about the chemical composition of Ibuprofen, or knowledge of the medicine and its side effects, much less knowledge about a broad set of medicines.
Consultancy isn't for scale. That's why big 4 have thousands of people.
Edit: to clarify, material production doesn't require knowledge, it's pure mechanical reproduction. Design production doesn't need reproduction, we have human ingenuity for that already.
To what degree, for how long, under what circumstances, and with what externalities that is accomplished is a separate discussion.
But in knowledge work, the output's purpose is often merely to represent a cognitive process, in which case sidestepping the process negates the output's value.
Think of an architecture document. The entire point is to obviate downstream questions and discussions because the thinking process has already been done and agreed upon from multiple sides. If we AI-generate an architecture document, the humans evaluating downstream work output may still end up arguing because the cognitive process the document represents never happened, in which case the document was pointless from the start.
Also, clothes and shoes often aren't as automated as we might assume. They're made in low-payroll countries by workers who actually are fairly skilled at the individual steps of the process they do. Electronics final assembly is often the same.
It only truly becomes slop if nobody reads it and learns something.
It's more complicated in practice because:
a) Communicating something well is an important skill and of itself.
b) The writing process often continues to sharpen our thoughts, even after the research is already done. Writing is rarely a mechanical task of putting what's on your mind onto paper.
c) I'm not only talking about using AI to write the final paper. Anyone writing the paper with AI can be presumed to have used it heavily in the other parts as well.
I largely agree with your point but I see a lot of people lumping everything together.
Do we need 50 different shoes produced by borderline-slave conditions in SEA to have plumbing? Do we need to pollute be getting a car to drive to the gym to have ambulances rushing people in to the latest medical facilities?
A lot of what we are seeing is simply capitalism running rampant without much of a care with regards to actual humanity. There was no preconditions we needed to eliminate shoemakers to have a safer, more functional society. I would even argue that it's a mistake that we are paying for in human livelihoods, reduced production and over-reliance on foreign countries in the West and climate change.
Also "a shoe is a shoe" is unnecessarily reductive. Even assuming the quality stays the same (it does not and has no incentive to) and leaving all semantic arguments aside, with shoemakers you have decentralised production that boosts local economies. You have a community that can provide both (mostly locally sourced) material and ultimately knowledge. Up to a couple of decades ago, people used to go to the local shoemaker to fix their shoes, making the whole chain more sustainable. And of course you have none of the scale of the climate impact, the oversupply of cheaper goods and centralisation of production to a few major countries who have control both over the workforce and consumer.
And none of this is antithetical to technological progress. You can have shoemakers that have new tools and new techniques to aid them, without the need to build shoe factories pumping out Nikes 24/7
I don't see how the tech coming from "factories" will end up in shoemakers' hands without said factories? The scale is driving the progress?
All of these can be used at an individual case (and have been, textile manufacturers have been using looms for quite a while). I own all of this and I am able to create clothes for almost the entirety of my friends and family even as a hobby on my free time, it really does not take ages to create something by hand wpoith the aid of technology.
The difference is that all of this still gives the full control to the craftsman. I can decide to do difficult colourwork by hand, or hand-sew a specific part of the garment while it allows you to automate the repetitive parts. Factory work does not let you choose what is repetitive, _you_ are the repetitive and expendable part of the process.
Also if you assume that said equipment must come from huge factories this is also wrong. It requires a lot of expertise to create something like a sewing machine but the point is that you can achieve this without massive production lines. We've had quality products in Europe for a while - and you can tell the difference if you get your hands on a vintage Bernina. Looms are still being worked by hand in some cases. For a lot of consumer electornics, hand-soldering is very achievable and local smaller-scale parts can easily do metal and woodwork -- but those are dying out in an alarming rate
I am not against automation of all kind, I explicitly am against this scheme of infinite scale to get ever cheaper goods. We can instead strive to strike a balance between a sea of cheap goods and local, resilient and self-sufficient communities that don't aim to enshittify their whole production line and environment to get a slightly better margin
I look around the world. I see societies that are wealthy or developing where capitalism is running rampant. It sometimes calls itself other things (Chinese “communism”) or is organized in different ways but it’s capitalism running rampant.
Then I see poverty.
Then I see totalitarian managed economies of various kinds (theocracy, real communism) that are sometimes marginally better than poverty but basically poverty with guard rails.
Where is the successful alternatives?
I suppose you could reach for Northern Europe. But it’s basically capitalism running slightly less rampant with more safety net. Maybe that’s the pragmatic answer. But it’s not a different system, just a differently tuned instance of the same system.
Modern Capitalism Is Weirder Than You Think. It also no longer works as advertised.
In a lot of European countries we hit a sweet spot around the 60s to 80s (depends on how much plagued by poverty each country was after the war). Things were not prefect but there was a strong social net and strong unions (which have existed for far longer) that prioritised local production. Of course there is much more nuance than that and I can't condense the reality of halt the continent in a single paragraph, but it is very different to what we are experiencing now and it's still a "capitalist" society at its core. There is even a stark difference compared to two decades ago, where the process of globalisation and concentration of power to a vanishingly small minority was already evident
What you are seeing today is an echo of the past because we have reduced local production and are paying the price for this so you end up with a welfare state which is fueled by a service economy, which is almost self-evident why it is bad and a matter of time until the system collapses.
Ruthlessly overthrown, murdered, or otherwise destroyed through meddling, slave labor, colonies, embargoes and sanctions by western colonialism and imperialism?
Even if someone discovered a perfect system we would never know because capitalism would destroy it first.
These discussions always devolve into "Economic system X is bad because country Y suffered Z" with a billion confounding variables affecting Z. Sometimes Z is a consequence of X, sometimes Z is caused by poor leadership of Y, sometimes it's caused by external natural disasters/resources, sometimes it's caused by direct meddling of hostile third parties who were ahead at that point in history for about a million different reasons.
What can we claim about the success of capitalism or failure of other economic systems that can't possibly be explained by "that's because capitalism was already dominant and hostile?"
Of course the US has dominated the planet: it's run by the violent sociopaths who are blind to their own viciousness. There was a point in their history, near the close of their 19thC civil war, where they considered arming the slaves because they believed that the slaves also supported their system.
That system never really went away, though it has certainly changed its aesthetics.
So it's not surprising to me that the folks running the current incarnation of the system cannot see how viciously they work with the rest of the world.
What most folks proximate to me think of as "the wholly nature and only way of doing business" usually seems to me to simply be centuries of sociopathic exploitation beating people into a position where they think these kinds of social relations are "normal".
How and why would it be a perfect system if Capitalism is able to destroy it?
If Capitalism is really so bad then why can't some superior system destroy that, instead?
The take-home I am driving at is that an economic system has to do more than socially enrich its populace, it must also be able to defend itself against the antagonism of its environment and competition.
This is exactly what I was talking about. If your argument boils down to "capitalism is better because capitalism is dominant right now" then your argument is unconvincing.
Here's an exercise: If we started over with Soviets influencing 90% of the world, how well do you think the "capitalism experiment" would fare when you weren't allowed to trade with anyone, didn't have imperialism in your toolbox, when all of your capitalist leaders either got assassinated or replaced with more commie-friendly leadership?
It is extremely easy to sit in the First World and like the fruits that the Industrial Revolution brought now, some centuries later. Workers that were brought into factories had to work long days. Now we have the 40-hour workweek, and a two-day weekend. Now we have a “middle class”, i.e. wage workers with disposable income and a mortgage. We exploit the rest of the world for cheap goods. We exploit our immigrants for cheap labor.
Climate change is becoming a bit on the nose, but we don’t have to worry about the worst effects.[1]
In short, it is extremely easy to compress modernity to the headline: Progress Happened. But in the real, uncompacted world there was strife, struggle, malcontent, protestations, probably every shade of opinion and reaction. And with all that we didn’t arrive at a polished gem. We got a rich First World, for the most part (right up to the rise of China).
To bring up that compressed view of the Industrial Revolution in order to stifle discussion (let’s be real) makes no sense. We are living in it now. Of course we are struggling and debating with it now. Just like all the other historic, uncompressed events
- - -
To also compress the dimension to “progress” and “not progress” is a falsehood. People want progress that works for them. Not progress that oppresses them.
[1] https://ruralindiaonline.org/article/mosquitoes-are-ladakh-s...
Were this to be true, there would be no workers at all, it had all been automated decades ago. Still there’s around million shoe workers in Vietnam around, I think. Adidas tried automating shoe production at it failed miserably[1]. Seems nothing beats us pesky monkeys in hand-eye-coordination for quite some time. I think the problem was also that recalibrating the system to create different type of a sneaker was extremely hard and expensive, whereas doing same to a bunch of Vietnamese workers requires you to just, y’know, tell them.
People who never done any manual labour with their hands always think a job ω is quite simple to automatise away. And if you try it, whaddayaknow, it fails miserably. How many years now the fully automatised driving has been supposed to completely take over long-haul trucking just next year?
Coming back to your main point, with which I agree with, having an LLM doing your reading or writing for you is kinda pointless. It is like paying somebody to watch them to eat your dinner. In the end of that process is starvation. But equally we have been afraid of industrial automation taking completely over for at least hundred years, and we find that automating a making of a single sneaker is bloody expensive and bloody difficult.
That leads me to believe this automatisation of knowledge will not look quite like the C-suite prophecises and we monkeys will be needed for long time being, even in SW development.
[1] https://techcrunch.com/2019/11/11/adidas-backpedals-on-robot...
Sometimes technological breakthrough does something out of the box side stepping complex mechanical automation, for example by 3d printing.
So we never know what is around the corner rocking our boat.
I wonder what the cost of that would be vs paying a human being a living wage to do the job?
We are robbed of ever finding out since there are bypaths in other countries where you can pay people poverty wages to do the work instead.
Think what that means in terms of knowledge work. I am not exactly sure what is the real cost of models like KimiK3, but judging by the Anthropic S1 SoTA bleed money like crazy. If every token would be charged by it’s actual cost plus margin, would anyone buy it, considering how on the limit companies are with even the subsidised costs? Or would it be cheaper to just outsource the coding to India etc. Perhaps using a bunch of adequate local model to help whatever it cans?
Adidas already proved it is possible to automate shoe manufacturing, but they also proved economics of it make no sense whatsoever. Judging by the revelations about Nvidia and Nasdaq dropping like a stone today, same seems true about AI as well at least for time being.
When outcomes are all that matters, then there is no point in learning anything.
When AI is "good enough" there is no point in excelling.
I don't think that follows. The relative contribution of the human might change significantly. If a product is worth 100 value units, and AI goes from contributing 0 to 95 of the final value units that's a fundamental shift economic shift even if the final product still ends up better with human input. The implications on wages alone are enormous. If a cheaper worker can produce 99 units of value instead of 100, why pay the substantial premium for that last 1%?
"But the top human + AI product will simply be better" you might say. Businesses don't always live or die based on whether they provide the best product. Optimizing for profitability by accepting a marginally worse product and redirecting the labor savings towards marketing and distribution is a very valid business decision.
Humans remaining necessary doesn't mean their economic value remains anything close to what it was.
Historically, those new opportunities mostly don't go to the workers who have been displaced. There's a fair amount of research to suggest those new niches mostly go to the next generation. Displaced workers often have substantial, persistent earning losses [0][1].
There will be a new equilibrium, I agree. It doesn't mean the displaced workers will recover their livelihoods.
> Humans will still need to be involved because ultimately the purpose of production is to fulfill human needs, and machines will never know better than us what those needs are
This requires far fewer people. The economic benefits of automation mostly accrue to capital owners rather than the labor class. That's even been the case with AI so far!
> It's just that it's hard for us to imagine what that new equilibrium will be, so we think the disruptive job loss phase is the end of the story.
You're missing the actual concern. The question isn't whether society will adjust: Of course it will. The question is what happens to the generation(s) of workers caught in the transition. An eventual recovery is little consolation to the people who never get to participate in it.
Note that I'm not suggesting we need to halt all progress, but rather noting that "things will be fine" is an extremely simplistic generalization of the history of automation.
That's a wild analogy, not an argument. There are many differences between then and now, and even some of the small ones can cause results competently unacceptable from the viewpoint of the public.
> Humans will still need to be involved because ultimately the purpose of production is to fulfill human needs.
That's not the case anymore, it's one of the differences I mentioned. Today the purpose of production is to allow asset owners to acquire more assets than the other guy - and not only newly produced ones.
> and machines will never know better than us what those needs are
I'm happy to hear that machines are eager to know what your needs are but I can assure you, they don't care about mine... That's the crux of the matter.
It's not just magic to the unversed witness, but to the practitioners who no longer try to understand anything. They just become walking cheat-sheets full of incantations to implore some other entity to cause results?
There's magic and then there's witchcraft.
> It's not just magic to the unversed witness, but to the practitioners who no longer try to understand anything.
The real problem is when only the witch understands the eventual outcome but the practitioners can't, even if they want. Believe it or not, that's where things are heading.
It sounds nice, but in actuality the process of making stuff in the most optimal way possible is the pure essence of any industry.
At the artisanal budget, industrial version will have orders of magnitude better tolerances because the processes are perfected.
When you do things artisanally your ability to perfect the process is very limited. Only going industrial affords you the scale to perfect the processes.
One of the consequences that industrialization has had in the past regarding the production of physical goods, is that quantity and affordability has often been (but not always) detrimental to quality of the goods and loss of knwoledge for the people (so many crafts that have disapeared, which makes society less robust as a result). Also, a loss of social connection and peer-to-peer relatedness, as we find ourselves increasingly dependent on large scale processes that we can't even see, or feel not taking part of, instead of more local relationships.
The same thing happening to "intellectual" work is not really appealing: more slop and less education, is that where we are heading to?
I despised having to produce design documents for software projects 20 years ago (the last time I had to do it as a software engineer). I wanted to jump straight to coding. I begrudgingly found that, much like research papers, it's a forcing function in really understanding the problem space.
Over the last year, I've found that, at least for nontrivial things, if my agent makes a terse (under 300 line) technical design doc that describes their research and their tasks, it can highlight any disconnects between what I asked and the agent's interpretation. It also serves as context to ramp up cross-model reviews, and as historical reference material when I have to Chesterson's fence something later. I wrote up more about this here: https://photostructure.com/coding/claude-code-tpp/
But all those English-hating engineers now produce shitty AI design docs, which tick a box but have less than zero influence on anyone else’s thinking.
In the shoe analogy, there is absolutely deep knowledge and a creative process that goes into designing a new shoe. This includes market research, defining the new style, selecting materials, building and testing prototypes, etc. This all happens before factory production.
Software development is the same. The factory is the compiler.
The meta work that goes into creating the factory is analogous to the meta work that goes into creating an AI model.
But the fundamental shift of control is the same. In the 1600s, a scythe was merely a tool that was yielded by the farmer. But in the 1900s, a worker in a shoe factory is merely a tool that is yielded by the factory.
A compiler is a tool that is yielded by the programmer. The same role reversal will take place in regular knowledge work industries.
A factory process implies something repetitive. Developing new software is novel, not repetitive. Factory is the wrong analogy for software development.
Industrialization isn't unquestionable good especially the way it went. A lot of good things from industrialization should be weighted against side effects and lots of negative externalities. Early factories and infra companies externalized all the costs, dumped waste into rivers, and CO2 in atmosphere started going up right at that time in history. They didn't need to pay workers living wage or ensure safe conditions, they were pretty much all monopolies with company towns and no competitors. And those early industrial enterprises didn't displace remotely as many jobs as language model operators are threatening to.
Also people usually don't consider which of good things would've happened anyway without industrialization. Yes, it's difficult to imagine because it needs alternative history...
It's even knowledge that should be preserved and is needed for some edge customization or prototype or debugging case, but it's not how the vast majority of work will be done.
I mean, sure, but it's also a lot more than that because it fully encroaches on intellectual activities outside of work. So I think it's way, way too myopic as see it as just another industrial revolution.
Frankly, if you don't actively reject LLMs in some areas of your life, I think you're gonna end up a lot dumber. Or, the kids will. If the only thinking skill we have is prompting, or not even that - there's a push for "predictive" LLM agents - then I suspect we just grow dependent on always having a subscription and stop using a good chunk of our brain. Like, if you're no longer writing, no longer reading anything that isn't ELI5, no longer formulating hypotheses, making art, learning crafts, or even comparison-shopping... what are you gonna use it for?
apparently, not that easy to write
> How hard is it to right ten pages yourself?
Quite hard, apparently.
Unfortunate typo
That was unintentionally funny.
> Assignment documents that start with "you are strongly encouraged to explore the use of AI" which quickly turns into the clear understanding that you aren't just encouraged but expected to use it. LLM is not an AI. IBM Watson is an AI, Prolog is AI, OWL ontologies, expert systems etc - technically any system where knowledge is encoded explicitly and unambiguously. Language model is an advanced vectorized search engine, a compression algorithm with a human-like dialogue system on top. LLM has no intelligence. Here is a simple gedanken proof of that: if we define an LLM as a "rational" system the latter means that it does not contradict itself, it has a complete non-contradicting picture of the world. We train an LLM on the whole literal corpus of humanity completely ignoring the fact that the progress is , in matter, a resolving contradictions. Having all books of humanity on the table is necessary but not sufficient for building a real intelligent system. Building such a system would imply identifying and resolving all of the contradictions, all the ignorance present in our knowledge. Calling an LLM an AI is the greatest misleading marketing trick in history.
I do voluntary teaching while working as an engineer at my job. I teach the interns because this is how I learned my skills. Except math and physics, 90% of skills I learned at the work place from elder colleagues. We have to surround ourselves with knowledgeable people who are able to take responsibility. Even if they will not stay in your company. I can't just sit still while this tech is destroying our and our children's future for the sake of smb's short-term profit. One might get all of the money of the world, but some things cannot be bought.
If you are a student/studying like the author of the article, there are only two goals you should have. Either to learn things, and then you have to quit the class or school if your professor tells you to use AI and go somewhere where they can teach you.
Or you're there to get a juicy diploma so you can get a juicy job, in which case by all means use AI or whatever means to get it.
I guess the third case is also if you're young and want to have sex with other young people, you should go to school.
I can tell you wrote this yourself.
I can tell you wrote this yourself.
If you’re not doing both then you’re not doing your job. AI isn’t going away because you don’t like it. But if you don’t use it as a tool then you’re going to be left behind. Note in some cases you’ll be fine with that. Especially hobbies. But it’s bad advice as a general rule.
It’s not that it’s hard it’s that it’s time consuming. It will take me less than ten minutes to have the report ready and 20 min to review it. And at the end of the day what is the purpose of that report? Being able to have your supervisor skim it and give you a couple of pointers he’s certainly not going to spend 40hours invested in it, and your status doesn’t change from you spending 40hours writing it as opposed to spending 20h writing it. I’d you spend 30min getting the report ready and your supervisor can review it in 30min instead of 1 hour because he can prompt it with AI that gives you both time back that can now be spent on digging into what matters, which is the actual learnings and insights from the work and the future direction.
Why crawl when you have the option of taking a bike?
It’s not outsourcing the thinking. This is a status report on a project where you have the facts available. It’s outsourcing the physically hitting keys, and copy pasting figure numbers and charts. It’s guiding you to getting to the end result faster.
:eye twitch:
... so was it? You never actually came back to explain this point. Is he expecting you to use AI, like you indicate the course material does? Or does he think the AI-generated reports are worse?
Very hard, apparently.
it is hard indeed to (w)rite, right, though tis just a simple rite to wright things sometimes? for many people, as it seems, also for the author himself. like as a non-native english speaker, it took me years to get to the point where i'm not ashamed/concerned writing, and also sharing, and doing it in a manner that actually convenes a meaning.
but wait, this is a rhetoric question perhaps, typo and all? or was the pun intended? or author accidentally makes a very strong argument in favor of writing actually being hard.
as a sidenote - it is super hard to write 10 pages in French even to native speakers, Russian is also a good example with all the medieval падежи left in it.
For the love of Jebus why can it not get that right
So I spent the last few days building, and tuning my own AI writing detector, a la pangram – though my own, local and yea, less accurate (for now), but good enough.
There is a fundamental difference between LLM generated code, and LLM generated writing. People will moan, but software is a means to an end. I want the computer to do X, maybe display Y on the screen, in Z format. I still care about a person designing the X, Y, and Z, but frankly, who gives how it did X.
When I load a website, I don't really care what language its written in (ok, I sometimes do, if I'm considering self hosting, or tinkering). Its like caring if the people who designed a bridge, planned and executed the work in French or English. Doesn't matter, at the end they built it and it works.
I care very much how one person tried to communicate their own thoughts though. No issues with research, spelling/grammar checking, etc. Just write what you think yourself, in your voice. That's why I want to read you, because the way you communicate resonates, your externalized thinking, with my internal thinking.
Every time I see a "load-bearing" phrase, part of me dies.
I also particularly liked the positive wrap up at the end to prevent doomspiralling.
I encourage the author to try not to take everything to the extreme though. "Exploring the use of AI" I think is a good thing for any educator to encourage. It is different to relying on AI.
I think there's probably a sensible midground, where people use AI, without abusing it.
> At first this shocked me. Our reports are limited to 10 pages - so the appendices are getting a big workout around here - how hard is it to right 10 pages yourself?
See AI as a tool that can help with the smaller tasks. I'm opening myself up to the prescriptive law of retaliation here, but if the assignment was written like blog post, then _writing_ 10 correct pages takes a little more thought/time/effort/tooling than it took to "right 10 pages yourself" ;)
This is a good summary of the angst my industry is going through right now. I've been in this business for 30+ years. We are definitely going through the stages of grief -- because this is a big change.
But you're doing the right thing by trying to get the development experience yourself. You won't be able to effectively drive software development if you can't do it yourself.
Wellllll
Apparently it's not that easy.
Ah, the unintentional irony...
"Deliberately provoking" is a good description of Youtube videos titles and thumbnails. Thanks for the wording. I could not come up with such a good and concise description in a recent discussion.
I've had a scroll through and read what everyone is discussing here and have really enjoyed it. As with any opinion shared openly, there's a lovely mixed bag of responses and I've loved seeing what you're all thinking about this - even (or especially) if we don't necessarily agree! I'm a big fan of conversations like these and think respectful exchanges of ideas is the best way to further shape your own views and thoughts on things.
To very quickly address a couple of things:
1) "how hard is it to right 10 pages yourself?" I haven't been able to stop laughing at this and loved all the funny responses to this line. While I appreciate the people giving me the benefit of the doubt on that one, it was (sadly) just a mistake and not intentional humour! I rewrote that post a few times and while on a bit of a roll clearly didn't notice that slip up, but I hope it gave everyone else a laugh too. And for those suggesting that I should've put it through AI and then it wouldn't have happened: I just don't really want to! I write the blog mostly for my friends/family and feel no need to ensure that it's written flawlessly :)) Unlike the paper I wrote about, my blog isn't technical literature, and I feel no need for it to be perfect. It's a fun outlet!
2) My apologies for giving you a cliffhanger on my supervisor's intentions! He did mean it as a compliment and followed it up with a line about enjoying being able to see how people's brains are working through a problem in their reports. His expectation is that we write the reports ourselves - so I guess we agree on that! His initial comment just created a link to a lot of other things I've been thinking about lately.
3) So glad I've gotten to introduce so many of you to the tawny frogmouth! We have some really beautiful wildlife here in Australia :))
Anyhow, I really didn't expect to have that many eyes on the blog, but I guess that's the nature of the internet! Have loved getting to read all the opinions and thoughts here, thanks for sharing them so freely :) And thank you to BrunoBernadino for making this a very interesting Saturday for me!
Happy internet exploring all! (And for the record, this blogger is a she!)
Most output produced by humans without using AI’s is low quality slop.