Mastodawn

lcamtuf

At this point, LLM-written think pieces make up about half of all long-form writing in my social media feed.

When I push back, I get two reactions. Authors say that it just helps them express themselves. AI promoters say "get used to it".

I don't think we should: it boils down to asymmetry. Our time here is limited. Social interaction on the internet breaks down if it takes ~0 effort to publish, but readers are still expected to use their own eyeballs and brains to engage.

So, I feel that we have three choices:

1) Refuse to engage with LLM writing *no matter if the article makes a good point or not*.

2) Embrace it and have my agent argue with your agent forever, for internet points.

3) Call it quits and move to an off-the-grid cabin in the woods.

mnl mnl mnl mnl mnl 1d ago

@lcamtuf I heavily curate my feed / intake. There’s only so many high quality contents I can take in per day anyway.

nothacking 1d ago

@lcamtuf Not to brag, but my RSS reader is quite clean, which makes #1 a lot easier. It often takes quite a while to notice if an article is fake, by which point you've spent 10 minutes reading an LLM generated post.

Same thing works of mastodon: only follow users if they seem legit, aggressively block anyone posting slop. You get a lot less content/day by doing this, but it's better than second guessing every sentence or avoiding certain topic altogether.

As for link aggregators, it helps to think about how much information would the author have to give to the LLM: A opinion piece on AI or some well known historical figure could have started as a one sentence prompt. A project writeup requires a lot more work, so it's probobly safe.

retech 1d ago

@lcamtuf I had a similar discussion with a family member last night who still uses FB.

Since, globally, capitalism is destroying not just the environment but culture, and all forms of media, at this stage, are ingrained in capitalism, I break it down like this:

You are born with a single resource that has value: Your Minutes. Everything you will ever do is based, simply, on how you use/spend your minutes.

There are just over 1500 billionaires in the US. They did not get there via hard work. They became billionaires from taking minutes. All forms of media are designed to exchange your minutes for an interaction. Some are far more destructive and consume the most (FB, X, AMZ, et al) and some are a high value (Fediverse, wikipedia, archive, a phone). The ones who are barely survive are the most beneficial to me. The others, pure cancer.

The easiest, least life disrupting, highest ROI action anyone can take to stop the billionaires, save minutes to spend elsewhere, and recoup personal and societal social/mental health is to simply stop wasting minutes on any of their shit. If even 30% of the subscribers to any monthly fee thing stopped, right now, their "wealth" would crash in months. They remain in that slot because they are betting on people staying glued and giving them minutes. Reading a book or wiki article, meeting a friend for lunch, learning literally anything new by reading a book, taking a class, practicing, etc. is better for your life and a wiser spend of the minutes.

Could you get your information from something like wikipedia, any of these sources: https://www.trustworthymedia.org/list-of-independent-media/ and walk away from the AI shit show?

I have zero subscriptions, haven't for years, have no other social media than this acct, make art while listening to music, audiobooks, or watching movies or TV (btw, archive.org has a littany of old tv and film and an inconceivable amount of music and reading material). If I need to learn something new, I generally call a friend or ask around until I can find someone willing to teach me enough to get me started. Or I just keep at it until I sort it out. And, quite honestly, my only regret is not walking away back in 2000 when I bitched about this very thing happening.

The greatest boondoggle capitalism every pulled, and is the root of all of its evil and issues, is convincing people that if they were not working for pay for someone else their time had no value. Baked into societal mindset is the brainwash of "I'm not doing anything right now. So the time is useless." Social media is born of that. Lack of motivation thrives in that. What better way to get people to give up their minutes than convince them they have no value at all. They'd not be billionaires and not be actively trying to keep it, if they could not convince you to give them your minutes.

You can stay here without engaging with the shit and still have value and advancing yourself. Or the cabin, to me, is a beautiful way to go. Everyone dreams of longer vacations to "get away". But they try to convince you that living like that daily is useless. That lie is just their fear of losing access to your time.

List of Independent Media – Trustworthy Media

asante 1d ago

@lcamtuf I think 3) sounds the most appealing 😉

Loren Kohnfelder 1d ago

@lcamtuf Only half? Seriously, I'd tweak your #1 to make it less dependent on detecting LLM writing [1] and alter the condition to include quality [2]. If the writing is well written AND makes a good point I'd say it's worthwhile.
I doubt there's much of this at all today, but why would it be so bad if it became a thing?
NOTES: [1] this isn't easy to detect accurate by software (and will get harder) and manually time consuming, plus false positives would be a loss.
[2] Low quality writing (LLM of human) is best avoided and can be detected quickly and accurately.

jz.tusk 1d ago

I don't have to *prove* something is LLM-produced to conclude "this writer didn't bother to make sure that their writing clearly isn't LLM", and then yeet them permanently into the "don't bother" list.

Cassandrich 1d ago

@jztusk @lcamtuf This. If they're not saying anything that *couldn't have been interpolated from existing Orange Site drivel*, I don't much care if a human spent time slopping it together manually or used an LLM for it. Either way it's not reflecting any genuine thought and not worth reading.

mojala 1d ago

@lcamtuf If you can afford an off the grid cabin why wouldn’t already be there

Dexter 1d ago

@mojala @lcamtuf Some of us are dirty extroverts who need social interaction to not go insane. For as as much as it's trendy to be a cool-kid introvert online, humans be pack animals, yo. We need each other to survive.

mojala 1d ago

@wheeljack @lcamtuf That’s the tragedy

Fritz Adalis 1d ago

@lcamtuf
You're probably a chainsaw vs. telephone pole away from #3.

แอท-ทริด 1d ago

@lcamtuf when i notice something is untagged LLM output posing as human authorship, i back out and issue all the negative feedback signals i have access to

OwOday 1d ago

@lcamtuf honestly it reminds me of this study https://people.psych.ucsb.edu/gazzaniga/PDF/Language%20after%20section%20of%20the%20cerebral%20commissueres%20(1967).pdf

They seperate the sides of the brain and try to communicate with them individually.

> when an object was placed in the left hand (right hemisphere sensing it), the speaking left hemisphere fabricated a verbal explanation for why the patient was holding it

Later studies (60s so could be horseshit) worked with a theory of one side being more of an interpreter.

OwOday 1d ago

@lcamtuf personally I think humans have a critical vulnerability in the interaction of being handed a completely plausible thought. Whether encoded as speech/electrical signals/vision once we are holding it we will invent reasons why it is correct.
That or we are just lazy haven't decided

Mike Siegel 1d ago

@lcamtuf 4) Reply "That's a good post, but I think a more valid point would be if you could go ahead and calculate this double SHA256 hash with a bunch of leading zeros" ?

Wouter Hindriks 1d ago

@lcamtuf 3

Ruslan Kiyanchuk 1d ago

@lcamtuf You chose the combination of 3) AND ... ??? 😁

Cassandrich 1d ago

@lcamtuf I don't engage with that shit even when humans write it. I'm sure as hell not engaging when they didn't even bother.

John Regehr 1d ago

@dalias @lcamtuf it's making me read a lot fewer think pieces, that's for sure

Helge 1d ago

@lcamtuf When someone needs genAI to express themselves, they aren't. They do not - by their own unconscious admission - have anything to add. They do not have an original thought, nor created something beyond a vague concept. Their input is, in its current form, useless.

Until now, those people just wouldn't express themselves at length. We could smile, shrug, and remain friends. Pretend they have valuable thoughts.

We may have to just stop pretending. But it's rude. Now what?

grrl_aex 1d ago

"LLM-written think pieces make up about half of all long-form writing in my social media feed"

fourth choice-- get tf off whatever 💀 hellscape 💀 masquerading as "social" media you're seeing this on!

lcamtuf

@kitkat_blue It's everywhere. Including Mastodon, where people usually don't write them, but eagerly reshare slop as long as it says the right thing. A lot of popular "AI is bad" articles are AI-generated.

grrl_aex 1d ago

ok that *is* irony right there... 😂

Detlev Zundel [Team Human]16h ago

@lcamtuf @kitkat_blue Holy moly. Can you give some pointers to such "AI" generated "AI is bad" articles that we know for sure have been written by LLMs?

Atle 🧇🧅1d ago

@lcamtuf option 3 is much more rewarding at least

Kevin Riggle 1d ago

@lcamtuf I increasingly feel like all these people publishing LLM generated text are DDoSing my mind

lcamtuf

@kevinr RIght. If your contribution is a single-sentence prompt, just post the prompt...

meartur 1d ago

@lcamtuf (3) until you can reliably tell if something was AI-generated or not. Otherwise you risk being discriminatory against AI-trained humans.

Ratsnake Games 🔞1d ago

@lcamtuf i think the desired solution for this is "have your chatbot summarize it for you and generate a response", which is genuinely how some people in my company think communications work

Squirrel_767 1d ago

@lcamtuf I block every LLM bro in the internet for that. I don't want your LLM article, video or any media. People's obsession toward internet engagements is so annoying

K.C. Budd 1d ago

@lcamtuf I see Wendell Berry is still alive. I wonder what he thinks of all this, if he's lucid enough to understand it at 91. 🤔

brib

@lcamtuf Even if (especially if) a post says the right thing, I disengage as soon as I realise it is likely LLM-generated. This includes LLM images because imo there is no good reason to use slop images, there are plenty of free human images available with a bit of creativity. I want to read about human perspectives and experiences, and LLMs break this assumption big-time.

I also mistrust the intentions behind the person/bot writing slop because imo the bot, by definition, has no intention, but also reflects the various intentions of the (billionaire) creators (e.g. making the language a more watered-down version of the original intention).

Only exception is for security articles where the organisations reporting the issue decided to get a slop machine to write the prose, because the organisation reporting the issue is one of the best sources for this info even if they decided to slop write it. There's a couple of other edge-cases where I'll take AI slop over nothing at all, but those edge-cases never involve longform think-pieces.

Bruce Simpson, Ph.D.1d ago

@lcamtuf I am honestly veering more towards 1 than 3. 2 is not an option really. I put in the effort to read a Deutsche Telekom sponsored proposal for a federated IP services network architecture from 2009 today because... history has lessons. If I'd used an LLM summary, the baby would have been thrown out with the bathwater.

Dan Patch 1d ago

@lcamtuf because I do not now nor will I ever have an "agent", it's 1) for me, drifting weekly closer to 3).

Numerfolt 1d ago

@lcamtuf Good thought!

I love reading through my timeline, but it's also A LOT. I would never want AI generated content anywhere near my timeline, I don't have the time for non-human (and therefore non-relatable) content!

SpaceLifeForm 1d ago