No shit. All online conversations on social media shouldn't be treated as reliable information. Not all data is of the same quality. I thought this was obvious, apparently it isn't.
this post was submitted on 04 Jun 2026
382 points (98.5% liked)
Technology
86548 readers
3749 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related news or articles.
- Be excellent to each other!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
- Check for duplicates before posting, duplicates may be removed
- Accounts 7 days and younger will have their posts automatically removed.
Approved Bots
founded 3 years ago
MODERATORS
That actually makes me wonder if these things have any correlation besides "more inferences = more true". You'd hope they have things like: if on wikipedia=big mega true, if on some random dude's blog=maybe kinda true, if on the onion=not true.
My guess is they use the vote counts to classify the reliability of the data. They might even have vote identities to classify the data based on audience. That actually seems pretty likely, now that I think of it. It would be an important part of making them effective propaganda tools, which I believe is one of the end goals.
And how sarcastic replies with high vote ratios like telling someone to add glue to their pizza so pepperoni doesn't slide around get through.
Yeah, upvotes aren't strongly correlated with accuracy. Could be informative. Or informative but false. Or just a joke. Or a troll that voters enjoyed. Or a plain old aggressive comment against someone they don't like. Or positioned such that an unusual number of people accidentally tap upvote. Or brigaded from another more popular sub. Or the votes were botted or bought. Or randomly upvoted to camouflage a bot always upvoting its owner's account. Or its visibility is better than an even better comment and thus got more total votes.
That's why I think reddit is a shit source if you want an accurate bot, but might be pretty good if you want a propaganda tool because each of those things can play a role in effective propaganda.
That's why I rely on Lemmy for my information about . . . um . . about . . .
Well. Owls. I guess.
And dull things.
💪
We know a lot about crows, too.
I hope it's not just any old owl...
Fuck Reddit and Fuck Spez.
So I’ve been thinking about this: where would we post massive amounts of miss information with the sole purpose of miss leading AI?
Just dump it on Reddit, it’s that simple? Or would an instance on the feddiverse also work?
The fediverse will run into this problem: Reddit is charging AI companies for access and all other big networks are increasingly making it hard to index them in search engines or to scrape them. The fediverse is different. We will see it getting more visible in search and therefore also in AI results, because it is accessible. Which means that people will come here and try to game the system.
As long as it’s free, like the rest of the internet, I kind of want AI to work.
How could we kill commercial AI and keep Free AI?
I don’t think ai will ever be neutral/benevolent. It’s too easy for fascists and the like to try and manipulate matters through it.
I personally would rather see it go away or at the very least take a backseat.
Reddit has been trying to block crawlers, including the internet archive, so they can sell their data to commercial AI companies.
Except for legal liability, the other solution is good-enough models too cheap to run locally to justify paying hosting. Like prepackaged stand-alone AI on-site servers which are cheaper after electric costs than subscriptions.
Quora is being recommended in some circles:
https://www.quora.com/How-are-there-so-many-scammers-and-intrusive-ads-on-here-lately/answer/Franklin-Veaux
Well, the manipulation via Reddit is through paid Reddit ads.
The tragedy of this is the tragedy of our age and that is if people and companies of people put as much effort into being good as they did into trying to appear good then they would actually be good.
I think it would take significantly less effort to just be good.
Like, basically every study done on the topic agrees that treating your employees well will net a benefit on the capital investment required due to increased productivity. They're actively throwing money away.
And like employees, the way they treat customers is actively hurting them.
There's always been a short distance between SEO and manufactured opinion.
I got the new fitbit air, it came with 3 months of “AI coach”. To give it a try, I asked it how it detected my stationary bike ride, with only a sensor on my wrist.
It gave me a confident answer about vibrations, pulse, etc. I checked the linked source - it was a years old reddit conversastion about another fitbit watch, where comments were guessing about the automatic detection of workout types.
I turned off the AI feature.
I remember once I signed up to one of those things where you rank / select AI responses to various questions (i forget what it's called it was basically a scam anyway). The last question was something specific about a world of warcraft item or something. I've never played the game so I googled it and clicked on the first Reddit answer.
Someone had asked the exact question being asked to me and the top response was something like "do the damn (aiRankingThing) yourself. And then there were loads of botted responses giving intentionally false answers etc.
Kind of realised then how easy it is to create complete lies on places like Reddit.
Also upvotes / downvotes are very easy and cheap to buy. You can change the sentiment of an entire thread for $10.
And stupid reddit is hell bent on stopping me from using reddit for one upvote i have to my post using alt
As Reddit is mostly idiots and slop now, it's a tainted source of training data.
I've been doing my part by telling a believable lie when people ask questions and then if anyone doubts me I insult them.
The day AI search responds to a users reply of doubting the results with something like "You are a fucking idiot" I'll know my efforts were worth it.
Reddit Tiny Robot: "What is my purpose?"
Companies Rick Sanchez: "You serve algorithmic relevance through AI."
Reddit Tiny Robot: "Oh. My god."
The few times I asked something to gemini it gave some bs response like "one Reddit user claimed... "
Hey atleast it's transparent
"Manipulate"
I'm sure that sentence will make sense to some people.
All I do on reddit is post my "performance art", have fun menacing dorks and on extremely rare occasions offer helpful advice
If I'm providing usable data, then either their algorithm has failed miserably, or I have