AI Slop is Killing Online Communities
18 points by rmoff
18 points by rmoff
Tldr: ai good, ai slop bad.
Very snarky HN style writing
Tldr: ai good, ai slop bad.
I'm not anti-LLM: I think it's marvelous that we have natural language comprehension on a computer. I also don't think you can put that toothpaste back in the tube. But you gotta own the externalities. You know, if you like cryptocurrencies, you can't wish away the fact that some of their most enthusiastic users are state-sponsored ransomware gangs in North Korea. In the same vein, if you like LLMs, you gotta take a critical look at what makes them easier to use to harm the online ecosystem rather than enrich it.
I worry that this post is dangerously close to "my uses good, your uses bad". Many folks take this stance, but their definition of "good" ends up being extremely malleable. "I'm listening to AI-generated music but that's OK because record labels were hurting artists anyway", "I had an LLM write this design doc and I sent it out without reading, but only because I was late for my dentist appointment", "an LLM wrote 90% of this blog post, but my prompt was interesting and novel", etc.
The defining property of a principled stance is that it inconveniences you, not just lets you judge others. So what's the line that the author wanted to cross, but wouldn't?
Even worse, one of the "I want to have my cake and eat it too" people. To be fair, that's most of the LLMenablers I encounter.
It's really funny, most of these are, career-wise, a few steps above me (I'm not saying this with any judgement, I just use it as a yardstick for "can probably think further than they can spit"). Yet none seem to have lived in the same world as I have.
There are no technical fixes for social problems. What keeps bad behavior at bay is punishment (to some degree, also hard to enforce across borders) and friction (which is gone due to LLMs). Good LLM usage is, fundementally speaking, the same as bad LLM usage, but with extra steps. Therefore, there will always be more bad behavior then good, as long as any amount of money > 0 can be made that way.
Changing any of that is possible with much effort in localized scenarios. Globally speaking, we are fucked.
What keeps bad behavior at bay is punishment (to some degree, also hard to enforce across borders) and friction (which is gone due to LLMs).
No carrot all stick?
You tell me. Can't think of one, when it comes to these topics. Not a whataboutism in my mind (since the behavior is beneficial to the behaving person, but detrimental to many others): what are your carrots for countering political influence bots or spam mails? Or put another way: please, I'm all ears/eyes! I don't want this, I just see it like this.
Tldr: ai good, ai slop bad.
Pretty much. But also please don't f**k up the communities that I care about.
Very snarky HN style writing
Is that a good thing or not? :D
Is that a good thing or not? :D
Bad thing. It reads as a "I am holier and smarter than thou" vibe while presenting basic information. A lot more challenging material (for me) is often shared here while the writer retain their ego below the stratosphere.
I dislike this reply because of its snark. I thought the article was thoughtfully-written and while there’s valid discussion/criticism to be had around the stance, you’re just shutting that all down by putting it into a box. Also, fwiw i think some degree of snark is appropriate when addressing people that post ai-written blog posts to human communities
It was mentioned in another submission too, but I think one of the biggest assaults on my psyche is obviously AI written READMEs. Even if the project is actually good, an AI written README is such a strong signal for me to just close the tab. They're just painful to read.
For many years now after lobsters and HN I would also read github.com/trending over coffee in the morning, abandoned that maybe six months ago for this reason.
There are "nice", "polite" slop enthusiasts. The ones who insist they have taste and tact. They would never post bad slop, recklessly, only the very highest-quality human-refined, curated slop. Not really slop at all, they would argue, because they gave it a careful review before posting it. They insist there's a very important difference between this premium slop and the nasty kind, and that low-quality human-authored media is actually slop, too, when you think about it. They talk about how important it is for people to use slop thoughtfully, efficiently, correctly, and that we all need to learn about and discuss slop constantly because it's the inevitable future and highly relevant for everyone.
They muddy the waters. They wheedle, rules-lawyer, carve out exceptions, and talk about how important it is to have nuance in separating virtuous applications for slop from bad ones, and that focusing on the bad ones is actually very tedious and rude. We should have polite discourse about the good things about slop and stop being so mean about bad slop, which isn't even really a problem. The bad kinds of slop will be solved soon, probably, and the harms are overstated. They colonize spaces.
If moderators don't swiftly throw these slop enthusiasts out on their ass, slightly less polite ones will post slop slightly less politely. More and more of the people participating in the space will have favorable opinions toward slop, and shout down people who object to slop. In no time at all, your community is a slop bar. Who could have imagined?
Slop killed the https://digg.com reboot in a matter of months.
In almost no time, they were spammed by bots, leveraging the domain's good legacy SEO page rank, and there was no winning that fight.