Post

i think a core beef that i have with a lot of this "automation is bad" sentiment lately is that there is no inherent moral good in laboring the story of humanity is one of invention, where we improve our conditions by building things that help us do more things more easily

This post is unavailable.

65 reposts463 likes

I'm not sure how you can be against UBI, it's literally how the wealthy live their life. The 1% and higher basically can never be poor unless they start taking out insane loans because their base wealth generates so much interest.

UBI doesn’t fix the deeper issues of ownership and power. If we don’t fix that, 200% of your UBI will go towards your landlord/mortgage and thus you’ll still end up on the streets without a job.

tbh i find it deeply irritating and disingenuous to call this "automation", as if we'd just invented the dishwasher. it's not a contraption that performs a rote task; it's a box that spits out Whatever

the story of humanity is one of telling stories, and now i am constantly advertised the box as "What if you didn't have to tell stories. What if you could automate talking to your children. What if there were a TikTok nobody had to make. Just endless banal sludge, forever"

that's this part: bsky.app/profile/stev...

Steve Klabnik@steveklabnik.com

sort of along my recent repost, the issue at hand has been the same since the dawn of capitalism: when automation increases productivity, who gets to capture the value of that productivity increase? the beef is not with the productivity. it's with the value capture

i don't think this actually contradicts what steve is saying? automating away human creativity and connection is bad, automating away merge conflicts and meeting notes is fine. yes the most powerful people are sociopaths but that's always been the case

the same box does both and he is funding it. i don't find "the spam machine may be destroying the internet and chewing at the edges of what makes life worth living, but it helps me resolve a merge conflict ten seconds faster, so it's impossible to say if it's bad or not" especially compelling

i think that this is a *little* bit of a waffles situation. i don't disagree that, for example, Sora is boring, bad sludge. but i wasn't talking about that in this thread i have a lot of beefs and concerns with llms! they just aren't what i was talking about here

you opened by characterizing the feelings of other programmers as "waffles are bad"

My problem with AI right now, and sure it may go down over time, is people not realizing that the code working is only part of the story. When your architecture and design docs that you also had ai make refer to files and services that don't exist, it makes the job harder for everyone around you.

Automation isn't bad. Trying to do it with expensive tech in fields that have in some cases been efficiently automated for 100 years, is stupid and bad. See: all the crypto idiots that think a humanoid robot makes sense for harvesting grains. When we already had self driving combines for a while.

There's even a ton of cases where we've automated stuff extremely efficiently WITHOUT getting rid of humans, actually most cases. The general purpose robo worker of the future has never made sense and never will, whether it's a humanoid or a box of GPUs that hacks stolen data together.

I definitely have never had that Protestant work ethic but lately I am discovering that internally I do value accomplishing hard things. Not all hard things are valuable (to me); but valuable things (to me) require hard work.

That's a strawman argument. People don't want AI slop, AI making porn images out of peoples' baby photos, AI spying on them, AI producing flawed records, AI breaking the internet, AI hogging a community's water/electric, etc. AI isn't out there digging ditches, man. 🙄

AI is objectively not improving the human condition or building anything that's helping people do things more easily. It's used for mass theft of intellectual property, spewing lies & bullshit, & creating porn based on phone shots of the lady who lives next door.

That’s just literally untrue. Like there’s a super easy argument that the bad massively outweighs the good but “AI is not helping people do things more easily” is just trivially factually untrue. It was maybe true a year or two ago but it ain’t true today.

I think the profound sadness of this technology is that it was developed for profit, by corporations, using our, in a way that sounds like it'll steal jobs. If this was made by well funded public universities, to publish openly, even if it used copyright data, the public would love it.

OpenAI being de facto a publicly traded corporation, that are openly glib about copyright, and framing their tech as messianic and scary, it just feels like a cruel scam. This poisoned the well for open source ai, and probably models strictly trained on public domain data, as well.

And I hate it! AI has been of legit use in my life, building my own agents. For example, I set up a presence sensor in my room and a bunch of esp32 buttons around the house, and now it tells me to go to bed and wake up on time and reminds me to do my hobbies, making me go to the rooms at least

And it's stupid, but for my constellation of AI, I'm more sensitive to embarrassment and annoyance than following a regular todo list, and whether or not that's healthy for me, it gets me to make healthy choices. I'm motivated by a realistic human voice that can persuade me more than a buzzer.

oomfie you're so correct and cool

oliver@eikopf.com

“do you think we should automate <THING>?” YES! WE SHOULD AUTOMATE EVERYTHING. THERE IS NO VIRTUE IN LABOR

My 2c: we're automating for the wrong reasons. I see *tons* of code written to satisfy the gods of unnecessary complexity, when using a different approach would literally eradicate much of it. To mitigate this burden, we resort to LLMs to do the grunt work for us. It's all self-inflicted.

"building things that help us do more things more easily" once meant e.g. inventing Rails to remediate to the Java XML crazyness. Today it means delegating to a machine what we can barely comprehend, because mastering the full stack has become next to impossible. Which is fine for some, I guess.

The thing I keep coming back to is the question of whether its good to automate away the jobs of junior engineers, which is largely how LLMs are being deployed currently. I'm worried we will retire and bequeath an empire of shit to a generation who had no opportunity to learn.

I'm still not sure how much of the job market is really end-of-zirp and Biden-tax-writeoff correction. But I am real worried about the empire of shit. I'm convinced these things are useful now, but they still empower bad actors, copycats, and slop. Legacy software stamping on a human face forever.

No one is publicly talking about it because they fear for their jobs but my sense is that big tech is having more incidents at higher sev criteria than before and they are more difficult to solve due to brain drain

yes, there's a huge question of "do you trust the ceos aren't lying when they say ai is why they're doing this?" i'm not sure i do. not any consolation to those laid off though, ultimately

i worry a lot about this too. without getting too detailed, someone very close to me who is junior has been looking for work for like three years now. it's fuckin rough out there

i am a little unsure on the learning thing. like, i would hope that llms can make it easier, not harder to learn. there's a weird thing with the sycophancy where it's the inverse of the IRC rooms i cut my teeth on, a sympathetic teacher may help people learn, but of course they can also be incorrect

No one learns anything from a sycophant. Real teachers challenge their pupils and none of the interaction design of LLMs index towards that.

i simultaneously worry that these changes will make it tough for learners while also being like, i dunno i didn't have google let alone stack overflow when i learned, yet the kids have turned out okay so far.

"We" would like to automate the things that "we" would like to automate. The problem is "we" are not doing the automating, "they" are. They are forcing it on workers specifically to destroy fun and well paying jobs, and on consumers specifically to provide them even more worse lack of "service."

One argument I've seen in game dev is that "automating tasks like proc gen and QA takes away jobs that people could have used to enter the industry", which doesn't pass the smell test.

It will undoubtedly change, the question is whether it will be for the better. Losing entry points for juniors won't be better and will make creative industries especially even harder to get into for working class folks.

one of my best friends is a QA person, we talk about this stuff all the time he hasn't spent enough time with the tools to have a strong opinion yet but he's at least curious in how it can help him focus on the interesting parts of his job while automating the not interesting parts

I know you bring this up later in the thread, but as badly-targeted as the unnaunced anti-automation sentiment is, it's hard not to be pessimistic about the future of the working class without some serious reform to society that ensure that productivity gains aren't just hoarded by billionaires

While their (former) employees are just SOL when they lose their jobs. There's only so many times workers can Just Get Another Job until there aren't enough jobs to sustain the working class, and any economic system that wants to sustain itself needs to recognize this fact and adjust accordingly.

It may sound foolish, but despite how dismal the world is right now, I'm still optimistic we can solve *that* problem before the contradictions get too bad and hopeless.

So, the social context we are in is one in which people's access to the essentials of life are contingent upon their labor. If you automate someone's job, they might not have a roof over their heads. This is not because there's an inherent moral good to laboring, but because our society is like that

and our society is very much not well-positioned to change that right now, because reactionaries are ascendant. Automation in this context is therefore more likely to lead to a number of people becoming non-people in the eyes of the economy, and less likely to lead to easier lives.

but some people have an object-level vs meta-level issue and think this is because automation is inherently bad in a universal way that isn't contingent upon anything, which is a bit silly. It's a problem entirely because of where and when we are.

yes. i do not fault anyone who is upset because their job may no longer exist due to changes in how society works. it's up to us as a society to handle this appropriately. but we also can't use those instances as a reason to hold the rest of humanity back

I would also say that the leverage workers have on bosses is our ability to withhold labor from them; certain types of automation hinder our ability to exert that leverage and aren't morally neutral

Yeah something I wish could be more easily distinguished is whether people are building tools to leverage LLMs on their work, or chasing the mirage of the "AI coworker." I think the latter is both more harmful and less likely to actually do something useful!

(saying this as someone who is not opposed to using AI for software dev but also is very interested in the class character of how these tools are deployed)

Automation of tedium=good Automation of creativity=bad The problem is that there's no clean break between creativity and tedium, the extremes are obvious but the border is fuzzy. One person's tedium can be another's creativity. Make no mistake, there CAN be creativity in tedium, but not always.

this framing sounds like there's a fixed amount of creative activity that can be portioned into automated or 'done manually' parts. but automated 'creative activity' doesn't diminish the amount of non automated creativity that's possible, & can allow that part to shift to higher leverage efforts.

Labeling AI as “automation” or “improving conditions” is such a grand euphemism. It’s not like LLMs are fixing our HVAC, mining the rare earths, or sewing garments. It’s keeping people with easy jobs from having to think as much and types fast. It’s not improving conditions anywhere.

i think your view of these is far too narrow and, it's true that i'm only talking about a very narrow thing here. there's more to it! that's not the part i'm talking about right now though

I think there is a lot to say about applications of generative AI, but specifically for coding what I find ethically problematic is the massive laundering of intellectual output, where clearly licensed code goes in and unlicensed code comes out.

It's entirely clear-cut to me that this is a breach of both the spirit and the wording in licensing. This is particularly so for GPL-licensed code, but also for MIT, where attribution and license reproduction is necessary for substantial portions.

What is substantial? I think those who set arbitrary numbers on line counts and percentages of the code base are missing the point. If someone spends 6 hours debugging and perfecting the 50-line implementation of an algorithm that they were only able to develop because they ...

had studied the problem space for years, is this not substantial? Of course it is. So the LLM essentially copies little bits and pieces from here and there and remixes it so that it doesn't look so recognizable any more. Were any of those bits substantial? Maybe, maybe not. The user doesn't know.

what is bad is yknow, creating a machine that can only grow by plagiarizing work. build a model out of things you specifically have permission from the author to use, or are entirely public domain (open source does not mean free to use!) and I'll stop complaining then.

who is saying that to the extent that you can point to a pattern like you're trying to do here. all i see is people saying they don't like generative AI

yea i think ppl do a labor = exertion/engagement-with-life trick even when they think they aren't or other kinda mental substitutions like arguing tool bad bcz particular impact in current social setting rather than like, critique the relevant aspects of social setting as much

You can’t build just inventions from slave labor and stuff you stole without consent! I need you to actually try and understand why we are mad about this shit.

There's no *ongoing* "slavery" or "theft" in the LLM development right now so this is a moot point. Even if you think the first ones absolutely needed it, at this point they're perfectly capable of improving based on synthetic data.

I don’t excuse a product that was originally built on slavery and theft and retains the roots of slavery and theft at its core. If you’re a good person, you don’t either.

I live in America. The whole country is stolen land, the roads were built by slave labor, the capital that funded every single corporation comes in significant part from money earned through slavery or exploitation of said land. I still would like to buy food and use computers and drive on roads.

I don’t want you to use my works to build a CSAM generator, I want you to remove my works from the CSAM generator training set, and I want you to respect consent whenever you think about building something based on the works of others. LLM slop companies don’t do any of this.

This thing you think is just “automation” is a nonconsensual way to take what other people built and make the world worse while enriching the wealthy, and I’m so sick of seeing you pretend that this isn’t the problem when we have all repeatedly expressed this to you and others

i mean, all of these things are simply not what this thread is about. but if you want to get into it, i think that this attitude is the one that makes the world worse and enriches the wealthy. and framing it as consent is gross. private property is the problem, strengthening it is not the solution

I keep coming back to, is the plow inherently bad? What about the horse-driven plough? The tractor? Is the right level of automation just "whatever was around when I was child" or do we have some principled theory here

everyone who automates less than me is an idiot, and everyone who automates more than me is a maniac

I get what you both are saying, but I think what has people alarmed is for one, the scale at which things can now be automated, and two, it isn't just "blue collar" jobs being replace now. Knowledge workers are concerned because they see it as a threat to their jobs.

sort of along my recent repost, the issue at hand has been the same since the dawn of capitalism: when automation increases productivity, who gets to capture the value of that productivity increase? the beef is not with the productivity. it's with the value capture

"when automation increases productivity, who gets to capture the value of that productivity increase?" i think i need to get that on a shirt or something

This is where I am too. I'm somewhere between agnostic and positive about individuals using AI tools to do what they otherwise would just have to say no to. I'm rabidly against decisions like "We're going to lay off a bunch of people in favor of AI tools". But the two are closely/complexly linked.

Absolutely. When folks tell me about how much more productive AI has made them, my conversation with them goes like this: "Oh that's great! I'm sure you've gotten a raise then right? No? Oh well at least you must not work as many hours now right? No? What's that? Your co-worker got laid off?"

As it happens, labour is currently the primary means of non-upper-class folks getting the money they need to live, so any automation which threatens labour directly threatens their way of life. In order for automation to thrive it's important to first establish better systems for sharing value

Sure—but the folks pushing this are capturing the value. Not the workers. This is the whole historical Luddite argument. It’s about political economy. Why push the tools that are going to further impoverish people who have to make a living through labor?

I think also negative externalities as well not being compensated for by those who are capturing that value e.g. xAI and Gas Turbines being turned on in black neighborhoods. I think it's good to be mad at that and the people making those decisions.

Degradation of quality matters, too. I see a lot of parallels to what happened with textiles. Yeah we got more "productive", but we went from a world where pieces of clothing routinely lasted an entire lifetime to fast fashion and disposable clothing that you're lucky to have last a year

On the flip side, I think programmers of the 2010s have been building too much “forever tech” that was really just a ship on top of a lower layer. Some tech can be disposable and while others need to be foundational. Making the disposable tech as easy to throw away as it was to create is important.

i think this romanticizes the past a bit. yes, there are issues with fast fashion, but clothing lasting a lifetime isn't inherently a good thing. clothing being more accessible does matter actually

yeah, the memphis data center should result in prison time that doesn't mean all data centers are bad it's important that it's pointed in the right direction

This exactly, and the whole AI is bad is so intertwined with capitalism value capture because the tech CEOs have been pushing it. If AI as a tool was a bottom up approach then we’d be fine with it. They just pushed too hard too fast from the top.

Also the luddite movement smashed the machines in factories because they hated the *concept* of the factory and the exploitation factory labor inherently represented. They didn't _want_ to control the factories; they wanted to own their own labor and direct it as they pleased. Distributed capital.

I have seen some comparisons with luddites and people that have an extreme take on generative AI, but I'm interested where this comparison stops. Like the scale of how certain internet things impact people is not something that I've read about before, like social media and LLMs.

They smashed the machines *in factories where the owners did not share the increased profits* and not in the ones where they did. They wanted to share the economic benefit, not break machines

(They themselves used machines! They opposed concentrated capital power busting working folk down to the level of serf or indenture.)

a thing i've been thinking about posting for a long time is "back in those days, the workers didn't say "eww factories were created by capitalists to enrich capitalists, we should abandon factories" they said "we should seize them and direct their output for the good of humanity"

I think the difference is despite the flaws it is easy to see how factories produce things that benefit people. A textile factory is going to make a blanket that keeps you warm, meat processing will keep you fed. Modern tech's benefits are niche

I understand where you're coming from, but I also think it's worth thinking about the fact that the split in tone here is along the lines of people who are currently financially stable and the people whose livelihoods are *presently* being eroded by these technologies with no upside in sight.

verily

Ted Underwood@tedunderwood.com

Imagine you time travel back to 1847 and you find the left response to industrialization is a) machines will never be as good as human weavers or b) we need to copyright loom patterns or c) it’s a speculative bubble. You’d say “Y’all. Not helping. What you need is obviously a labor movement.”

and i'm not the first to say this, but it is a grave mistake that folks want to cede an important technology to the right we fight for technology so that we can direct it towards good ends. we don't just let the other side take good tools because we don't like how they use those tools

I honestly don't know how you're getting code of acceptable quality out of these things. When I use Cursor to generate code, it gets things obviously wrong, struggles to factor things in a way that is extensible or extends existing code, and makes changes unrelated to my requests.

Like, I can see using this to create a single-use disposable application that only needs to *mostly work,* but not something that I'm interested in maintaining. The automation isn't bad (morally), it's bad (low quality).

Loved the thread. I think it's naive to ask people not to use a piece of technology that makes their lives better. Technological change has always happened and people have always chosen the path of least resistance. The question for me is more like "How do we use this without losing ourselves?"