Why We Don't Use AI
92 points by chris-evelyn
92 points by chris-evelyn
Are you zealots or luddites who just hate AI?
I think we should remove the negative connotation from "luddite". I read a bit of the book Blood in the Machine, and I think the luddites were a lot more principled and relatable than we give them credit for. They had the same concern with the Industrial Revolution this article has with AI today--new technology destroying livelihoods to make rich people richer. They weren't opposed to all new technology either! It was very focused on the ones that replaced humans instead of helping them.
It was very focused on the ones that replaced humans instead of helping them.
But we did replace them, and the world is a far more prosperous place now. If we'd listened to the luddites, then absolutely, some textile workers would have had better job security for a few years, but in the long run it would just make clothing much more expensive and require far more people to be working in textiles, stopping them from doing other stuff.
But we did replace them, and the world is a far more prosperous place now.
Sure, but the average clothing people have now is worse than it used to be. Clothing got cheaper, but it did not get better. Modern clothing falls apart pretty quickly.
"Cheaper" vs "better" is a tradeoff, and I'd like it if the tradeoffs weren't decided by what makes rich people the most rich.
But you don’t need to buy the cheap clothing that falls apart - you can absolutely buy high quality handmade expensive clothing that will last years. You now can choose!
The tradeoff isn’t decided “”the rich””, it’s decided by you when you decide how much money to spend on clothing.
Some quick napkin math suggests a very basic pre-Luddite linen shirt might last about 15 years with repair and care, and would cost 5-10 shillings (~3 weeks of wages).
Let’s say a modern shirt lasts 2 years. It needs to be 7.5x cheaper than the Luddite shirt. 3 weeks of wages at the average UK salary is around £1500 post tax….. and an equivalent basic modern shirt only costs about £30. So for the old style shirt you’d be paying £1500 every 15 years, compared to £225 every 15 years now. That’s a ~7x improvement!
you can absolutely buy high quality handmade expensive clothing that will last years. You now can choose!
The number of suppliers of that clothing is now almost non-existent, far more niche, and much harder to come by. The population of customers is diminished, so fewer suppliers can exist, and what suppliers that can exist now have significantly higher production costs, since the entire supply chain is altered all the way down to the farmers.
Before the invention of the car, you had a horse-drawn carriage. Now that there are cars, you can now choose between a horse-drawn carriage and a car. You can still choose to drive a horse-drawn carriage. Rather obviously, fewer people today know how to fix a wagon wheel and so finding someone to fix a wagon wheel is harder now than in the past. Also rather obviously, your ability to stable a horse is not today what it once was. The existence of an alternative does, in fact, have an effect on an option. You could also choose to speak Latin.
As it turns out, people live in societies.
The number of suppliers of that clothing is now almost non-existent, far more niche, and much harder to come by.
That there are fewer suppliers of fine clothing or that it's much harder to come by is not true. There are many brands that produce high quality clothing with natural materials that last quite a long time. Their products are also very cheap compared to how much something would cost in ages past. And because of the internet and global supply chains you can very easily acquire it without ever leaving your home.
However, a cognitive cost assorted with buying anything these days is unfortunately true and difficult to combat, but I imagine AI would be actually helpful here. Before you can buy a higher quality shirt or trousers you actually need to know why the cheap stuff made from polyester is not really that good and that there is an alternative. You need to be aware of the niche. Perhaps in the case of clothes it's not that difficult or time-consuming, but that you need to do this for almost everything in your life adds up very quickly.
Sure, but a car is several orders of magnitude better at transporting stuff than a horse. The same is true for textiles, and there are markets for machine-crafted high quality stuff, it's just not "brand clothes", but specialized use cases from medicine to civil engineering.
The number of suppliers of that clothing is now almost non-existent, far more niche, and much harder to come by. The population of customers is diminished, so fewer suppliers can exist, and what suppliers that can exist now have significantly higher production costs, since the entire supply chain is altered all the way down to the farmers.
The higher production costs of handmade clothing is due to labor being more expensive than raw materials these days.
The ONLY way to return to a world where labor intensive stuff is cheaper, is to massively devalue labor.
Also, people in the past had much smaller wardrobes. If you're OK with only having, say, four sets of regular everyday clothes, you can spend a lot more of your clothing budget per item.
The tradeoff isn’t decided “”the rich””, it’s decided by you when you decide how much money to spend on clothing.
Except when you don't have the wealth to buy these high quality handmade expensive clothing. Then, you can only buy what the the richcapitalists decide to produce and which is affordable with the amount of the money they decide to pay you.
Yes, and without all this automation those people wouldn’t be able to afford clothes at all!
You complain about hand-made clothing being too expensive You complain about cheap clothing not being as high quality as expensive clothing You complain about labourers not earning enough
The only way you can have cheap hand-made clothing is to pay labourers less or to automate things
without all this automation those people wouldn’t be able to afford clothes at all!
what percentage of people do you think ran around naked in 1810 because they lacked the automation to be able to afford clothes?
My understanding is that many working class people in 1810 may have had only two complete outfits of clothing (heavily dependent on undergarments), and that many women spent a lot of time making and repairing clothing at home. The baseline living standard is different today.
I do think we’ve gone way too far into the other direction with the “fast fashion” deluge of essentially disposable clothing. I remember there being a more reasonable middle ground when I was growing up.
I guess personally I’m closer to somebody from 1850 than 2025 in a way — I prefer to buy a few good quality pieces and keep them as long as possible (right now I’m wearing an Italian wool sweater I probably paid $400 for in 2010 and it’s still in good shape). But I also don't like clothes shopping!
Bret Devereaux has a good series on this. tl;dr it took about 2500 hours of labour to keep a family of six clothed at a "subsistence level" for one year. This is just for making and mending one set of clothes per person, not cleaning them or making any spares. That's roughly the equivalent of one woman working for 50 hours a week on top of childrearing and farmwork.
Also I don't know if "clothing lasted much longer back then" is true or not. Those studies seem to indicate that people would go through a change of clothes a year.
People would own 1-2 outfits, spend vast amounts of money on them relative to salary, and use a lot of unpaid reproductive labour (primarily from women) to repair and clean them. Do you really yearn for that time?
You know, in 1867 some German guy wrote quite a lot about the economics of weaving and tailoring.
I also suspect that if you have a more expensive piece of clothing you're far more likely to want to repair it rather than buying a new one, especially if you don't already know how to repair clothes (or have a wife that can do it for you). So even if the quality stayed the same (I don't think it did), cheaper clothes will be replaced faster.
edit: My understanding is that machine washing, and especially machine drying, is also rougher on the clothes than doing it by hand, which is another complicating factor.
Any places you recommend that sell clothing that lasts years? I've heard of the Swedish brand Asket claiming this, but I haven't heard any experiences as to whether it checks out.
Low quality is strictly due to the company ordering it in such a quality.
Machines can do far better textiles, there are better materials, better patterns to use, etc, like not even close to what someone could hand-craft.
Aren't most clothes we wear made by children in 3rd world sweatshops though?
Absolutely not too knowledgeable about the topic, but I'm fairly sure they are (ab)used to do only hard to automate, trivial stuff like putting the insole into a shoe and the like. The fabric itself is definitely done by some kind of machine.
The industry employs approximately 60 million factory workers worldwide yet less than 2% earn a living wage. Garment workers endure unsafe conditions, wage theft, exhausting hours, minimal pay, and gender-based harassment. The global exploitation in the garment industry was thrust into the spotlight in 2013, following the collapse of the Rana Plaza factory in Bangladesh.
Beneath the Seams: The Human Toll of Fast Fashion
Sewing is hard and expensive to automate. So human workers will be used as long as they're cheaper.
And why do you think "they are (ab)used to do only hard to automate, trivial stuff like putting the insole into a shoe and the like" somehow makes that OK?
And why do you think "they are (ab)used to do only hard to automate, trivial stuff like putting the insole into a shoe and the like" somehow makes that OK?
That's your own addition.
Sorry, maybe that's an ESL thing, but your phrasing read like a defense of sweatshops to me.
the world is a far more prosperous place now
This is an often repeated claim but it falls apart as soon as you try to analyze it logically. To begin with, do you have any evidence to back it up? Or should we ignore the mind boggling and endless list of problems caused by technology and declare the net impact of technological progress as a success based on our vibes? How are your vibes about the increase in prosperity of the world when you read about climate change, factory farming of livestock, nuclear contamination etc.?
do you have any evidence to back it up?
Yes
Life expectancy is up.
Infant mortality is down.
Quality of life for almost all disabilities is far higher.
Insulin exists and diabetics don’t simply die.
Access to clean water is nearly universal in developed nations.
Famines have become rare rather than routine.
Antibiotics exist & infections aren’t death sentences.
Anaesthesia makes surgery survivable and humane.
Vaccines have eliminated smallpox, polio, measles, etc.
Refrigeration means food doesn’t spoil within days.
Central heating and air conditioning make extreme heat/cold survivable.
Literacy rates exceed 85% globally versus 10-15% in 1800.
Hearing aids exist.
Prosthetics are functional rather than crude wooden pegs.
Travel that took weeks takes hours.
Communication that took months is instantaneous.
Child labour in developed nations is illegal rather than normal.
The work week has dropped from 60-70+ hours to roughly 40.
Retirement exists as a concept for ordinary people.
Absolute poverty has fallen from over 80% to under 10% globally.
Many of these are true and on point, though not all. Nevertheless, you ignored the crux of my post - which was not that technology has no benefits, but that it is not at all clear that the world as a whole is more prosperous net due to technology, because you are not accounting for all the downsides of technology. A list of benefits of technology does not mean that technology has been a net benefit to the world.
Honestly they had me convinced with just the first two bullet points about life expectancy and infant mortality.
My position has been consistent since I first started writing about this stuff four years ago. I do not think this technology (LLMs and generative AI) is going to go away. Given that, I think the ethical thing to do is help people understand how to use it in a positive way that adds value, while arguing against negative uses.
I feel comfortable with my decisions here. I have thought very hard about the ethics of what I chose to write about.
You and I clearly have a very different idea of what "evil" is.
EDIT: Actually you know what? I can't let this stand.
If you're going to call my position "evil" on a public forum you could at least have the decency to explain why you think that.
Is this about environmental impact, ethics of training, impact on society, impact on the jobs market or something else?
No, you're right, evil is too strong a word and I apologize.
I let my personal frustration get the better of me; long-winded justifications don't change that. I strongly disagree with you. Let's leave it there.
Thanks, very happy to leave this as a strong disagreement. There are a lot of very real downsides to this stuff, and some of them (the mental health ones in particular) are genuinely chilling. Reasonable people can come to very different conclusions as to if the harms outweigh the positives.
But not recognizing that significant portions of humanity have worse outlooks on many metrics under the violent, extractive imperial system that supports our high tech society
Besides the absolutely uncalled fact that you just called someone "evil", I'm not even convinced this is true. Like who do you think of here? Really, who? Like I have a hard time imagining a country where this general trend would fail to apply, thanks to technology.
Again, there are obviously huge societal problems. But you absolutely fail to recognize that there were far larger problems, some of it got solved (e.g. labor laws are significantly better than they ever were. People dying were absolutely normal during constructions and was simply part of the deal), some just simply transformed from one problem to another.
I absolutely agree that in many places a lot of things are a lot better than they used to be. But it is true that industrialization has caused and is causing huge problems for many people that wouldn't exist otherwise, mostly outside of Europe and the US/Canada. Climate change is a huge driver of this, but not the only one.
It sure has its own negatives (like the west simply dumping their waste on developing countries, among many others), but even still, technology made so many things cheaper that a general trend of lower infant mortality, etc absolutely "trickle down" to poorer African countries as well (and they may not even have to go through every step of the industrial era, so less net pollution).
So I don't think we are in disagreement, and I absolutely don't believe that we should take our eyes off of the countless suffering in the modern world. I'm just not convinced that arguing over LLM could sway the boat in any way, much larger forces are at play behind it.
Also, one way or another humanity will end up in a situation where there will be not enough jobs for everyone, LLMs are just an "attack" on white-collar jobs which was previously unseen/rare. But we have seen much larger "layoffs" "thanks" to technology e.g. in agriculture, we could just create new jobs for new fields at the time.
So we will inevitably have to face a future where only a couple of people have jobs, but still everyone have to eat. It either becomes an utopia where the rich tries to keep the rest in poorness, or we go for something like UBI.
an utopia where the rich tries to keep the rest in poornes
An utopia, you say...
Sorry, I assumed utopia can mean both positive and negative for some reason. Dystopia would that be.
I think it was originally value-neutral, but it strongly correlates to describing a positive view of a planned society nowadays. From the wiki
The word utopia was coined in 1516 from Ancient Greek by the Englishman Sir Thomas More for his Latin text Utopia. It literally translates as "no place", coming from the Greek: οὐ ("not") and τόπος ("place"), and meant any non-existent society, when 'described in considerable detail'. However, in standard usage, the word's meaning has shifted and now usually describes a non-existent society that is intended to be viewed as considerably better than contemporary society.
Edit Sir Thomas More would have had no time for LLMs, or even for machines:
There is no private property on Utopia, with goods being stored in warehouses and people requesting what they need. There are also no locks on the doors of the houses, and the houses are rotated between the citizens every ten years. Agriculture provides the most important occupation on the island. Every person is taught it and must live in the countryside, farming for two years at a time, with women doing the same work as men. Similarly, every citizen must learn at least one of the other essential trades: weaving (mainly done by the women), carpentry, metalsmithing and masonry. There is deliberate simplicity about the trades; for instance, all people wear the same types of simple clothes, and there are no dressmakers making fine apparel. All able-bodied citizens must work; thus, unemployment is eradicated, and the length of the working day can be minimized.
I disagree on some finer points but I think over all we're aligned in the same direction. I think I need to stop engaging with this topic online for the moment, but thank you for taking this conversation seriously.
because you are not accounting for all the downsides of technology
I think these upsides outweigh almost all downsides. Yeah, I wouldn’t sentence almost all disabled people to death, let millions of babies die, and millions of adults starve, just so that people can feel like they’ve got a better village community or whatever. I think the number of lives saved really outweighs almost anything else
I'm sorry, but I get "but what have the Romans ever done for us" vibes from this conversation.
I think it's absolutely a hard fact that technology has significantly made the world far more prosperous. That's another thing that this didn't come for free and it has created its own set of very real problems, no question there.
technology has significantly made the world far more prosperous. That's another thing that this didn't come for free and it has created its own set of very real problems
Why is it another thing? I don't see how it can be another thing at all in this context. It's very much part of the same thing. If you claim that X is a net benefit, then you can't also claim that the downsides of X are "another thing" - "net" means you account for the good and the bad together.
I am not claiming that I know for sure that technology has been a net negative. In fact, yes purely based on vibes, I also do think it has been a net positive.
But it's incredibly dishonest to use this as a "fact", even more so when it is an argument often used to shutdown and preempt criticisms and cautious attitudes towards new technologies. Can you or anyone in this thread lay an honest claim (@simonw, @john-h-k) to having done a serious and balanced accounting of the positives and negatives of the impacts of technology through history and the risks to future generations? Or have you read (or even know of) any studies which aim to do so? It's a massive undertaking. Forgive me if I'm not convinced by a list of benefits seemingly written to appeal to emotion. One could probably do the same for the concept of militia/war as it seems to be a massive source of technological innovation.
having done a serious and balanced accounting of the positives and negatives of the impacts of technology through history and the risks to future generations?
Well, have you done so, before you decided to use the computer, internet, electricity, a car, anything modern? Isn't it a bit dishonest of you to call out one use of technology while enjoying the benefits of countless others that have questionable negative tradeoffs?
Do you think @simonw is as impactful, that he alone materialized the whole LLM technology into the world? It's out of the bottle, so his singular usage of it doesn't have a significant impact one way or another.
I haven't. I am not the one making the claim that technology has been a net positive (or negative). I am the one questioning the validity of this claim by pointing out the gaps in reasoning that leads the masses to uncritically accept this claim. And I am concerned about this claim being regularly used to mock and dismiss reasonable criticisms about new technologies.
Given how challenging it is to actually account for the net impact of technology and categorize it as positive or negative, maybe it's best that we stop saying "technology good, luddites bad" when debating technology? There are plenty of other nuanced arguments to be had.
Great seeing companies with this attitude instead of the usual AI startups' empty promises.
I've written a blogpost that's a little similar to what they say, even though my reasons for not using AI are different (but coherent with theirs). It's here: "Stop Using Big, Bloatsd AI Tools" https://zoug.fr/stop-using-big-bloated-ai/
When you boil down everything they say and strip it right back, what they make are tools to either fire people or demand more work without hiring anyone new to help
This is a pretty disingenuous way to describe what is normally just called "productivity". It also applies to Excel, diggers, forklifts, high-level languages, and almost every other invention! These frees people up to do other things. If we lived in a world where we refused technology that might lead to "more work getting done without hiring anyone new", we would be in a very very primitive society.
The problem here of course is that the benefits of increased productivity most usually accrue to capital, not labor. The solution is an exercise for the reader.
Yeah and this is a complex problem to get right.
But I’d still prefer we make the progress! My life today is much preferable to that of 100 or 200 years ago. I’d rather have some incredibly rich people + my current life than have a very equal society with 1800s technology.
In 2010 ~5% of the world owned a smartphone. Now around 70% of the world do. Why? Because productivity increases meant creating a smartphone took far less labour. And this is clearly a benefit for people who are primarily not capital owners
You prefer your today life not because of the increased productivity of the society. But because there are labour laws preventing your employer to treat as a slave. (And because of science progress, but has nothing or very little to do with productivity.) As you already know that's not the case everywhere, if you were working in a mine in Congo, do you think you would be happier than people in 1800 because you own a smartphone or because people in Europe and America (and some other places) are wealthier?
And a note that the shares of productivity gains going to capital vs those going to labor have sharply diverged since the 1970s, with the proportion going to capital ever increasing while wages and salaries largely remain stagnant.
Hm. Still, it used to be that programmers to a great degree "owned the means of production", ie. their brains. Sure, organizations might buy a fancy IDE or pay for cloud subscription, but workers were still essential and to a great degree hard to replace due to institutional and expert knowledge.
The argument of "you will be left behind" never makes sense to me.
You could argue that knowing where it started vs. where is is will help you master a technology. That's true for fundamental knowledge, i.e. knowing how it works under the hood. But it doesn't apply to using tools.
Who's punching cards here as a developer? Yet knowing how memory works at the metal level is very helpful.
(edit: what I mean here is at we might be still at the "punching cards" state for ai-assisted coding, so there's no rush to adopt it. Waiting a few more months won't put you out business. )
1-b. It's a risky time investment
The bleeding edge is interesting to experiment with and get exposed to new way of working. It's also the natural place for those who have reached the edge of a knowledge domain, and wish to push further.
For the common folk (which I'm part of) the chances to invest efforts in the wrong place is high. On the contrary, investing in fundamentals is proven to have a great return on investment.
When we pick technologies, we value picking stable, battle-tested technologies. Sometimes at the expense of not having the latest and greatest for the sake of stability.
We value deep knowledge, stablity and experience.
But when it comes to AI, it seems that we tell people to throw everything out the window and put all their time into an unproven technology.
If we hit AGI, then coding knowledge becomes obsolete.
If that's the case, why don't we invest more in (again) fundamental knowledge? Business knowledge? Sleep? Family time?
It feels to me that AI assisted coding, if it end being as impactful as some say, it should promote slowness. It will give us time to study, understand deeply a subject, think and experiment. Implementation would become insignificant.
Long rant to say that while some argument have legs, the FOMO inducing "quick, you're missing out!" is unfounded at best, or even harmful.
Note: I don't believe at this stage that we will hit that "AGI" milestone, but I'm willing to entertain it for argument sake.
Note: I used interchangeably AI, AI-assisted coding and LLM. I'm refering to using LLM to code.
Edit: added "---" to make it more readable..
Since it came across my feed today, this seems relevant to this conversation (but off topic as a standalone story here). 15 coal plants that would otherwise have been retired soon instead won't be, and at least one operator explicitly cites AI, not generic datacenter energy usage, as the reason.