I know this overall battle is essentially lost on these boards but it’s continuously wild to me that some folks are seemingly unable to understand why “stuff a human member of our community wrote” is a more desirable thing to read than “stuff an LLM output” when that sentiment is expressed.
Who failed to understand that? That’s a straw man or a non sequitur. The OP wanted to propose a backdoor limitation on AI by using an existing rule that didn’t really apply. Were people supposed to pretend that the rule made sense because they supported the general direction of the outcome? If you want to restrict using AI on the boards in some way, go ahead and propose your own rule that makes sense rather than trying to hijack one that doesn’t apply.
I also hardly see any quoting from LLMs on the boards nor do I see any obvious signs that posters have replaced their own organic output with LLM text. It’s mostly a hypothetical problem and a moral panic.
Sure, there are plenty of reasons to prefer human answers. It’s just that “AIs are plagiarism” isn’t one of them.
AI slop, AI schlop. Modern generative AI is only a few years old. It’s gotten this good in less time than between the Apple II and the Macintosh.
We don’t know if it will get exponentially better in the future, given that it is now sometimes feeding itself its own output, but expectations for some improvement are high. I’d already rather post an AI summary answer to a question than read 90% of the slop that sometimes gets posted here by humans.
The true underlying issue is that the SDMB long ago became a community, in which people would rather converse and argue with one another at great length or go off on random hijacks at great length than get accurate and concise answers to those questions.
I’m not saying that’s necessarily a bad thing. It’s indeed totally human. Yet the preference for human-written slop over AI correctness is ironic and somewhat hypocritical. “Better to make our own mistakes than accept the superiority of correct Others” is a summary of human history that an AI wouldn’t give you; it’s too hallucinatory to be real. Yeah, and climate change is just as real and look what the status of that is.
Because that’s not a valid criterion.
The appropriate question is “is this information rich in useful content and verifiably accurate?”. If the answer is “yes”, I don’t care whether the information came from silicon or from meat – it’s still useful information.
I concur- for facts. But not opinions.
Yes, I understand that you feel that way. But if I say that I’d prefer to not read, here in this social space, information created whole-cloth by AI that raises all sorts of hackles about “moral panics” and hypocrisy. It’s unfortunate and it’s genuinely making me value this place less, which is a shame. But see the first line of my previous post.
I didn’t really read it as calling it plagiarism, just extrapolating from the idea that copy/pasting large amounts of text isn’t good practice. Presumably I’d be modded if I started doing that with huge swaths of Wikipedia articles in lieu of writing anything of my own and that would not be breaking any copyright at all.
I am curious about something but it might not be exactly right for this thread.
I’ve noticed many posts on the SDMB saying some variation of “I asked Google and…” I am never sure how to parse these because the casual meaning of “asked Google” has changed.
In the olden days of a decade ago, before widespread AI usage, I would interpret someone saying that as saying they did a search on Google and located some links that said this or that.
Back then the “I asked Google” would be usually be followed by a link or cite from some webpage, e.g. “I asked Google and I found this from the NYTimes (or from the Bible or from Fox News).” Often this included a link.
But nowadays in 2026 I see people saying they “asked Google this-or-that” followed by “and it said (e.g.) that Trump is a the 47th president of the US” or maybe “and it said that (e.g.) Trump is a Poopy-head.”
But that’s it… No indication what the cite was from, no links to any cite. So I’m curious: in 2026, when someone here says “I asked Google” can we assume they mean they asked Google’s AI and the answer they are repeating is from AI, (not an actual cite discovered via Google search), and that this needs to be clearly indicated as AI by the poster.
TLDR–
When folks on the SDMB say “I asked Google” in 2026 is it a safe assumption that they’re essentially saying “I asked Google AI”?
Are people posting AI slop to the Board with the implication that it came from a cite from a Google search when really it’s just some answer of questionable veracity from AI?
Even the “A.I. search” is supposed to retrieve results traceable to one or several links, otherwise it is worthless, as you say. The links themselves may contain bullshit, but that is a separate issue. There should be an actual reference.
Yes, but what if the Doper doesn’t include any of the AI offered links (which seems to be the usual case)?
Same as any other claim: you are well within your rights to ask for a tangible citation. A couple of independent sources, even.
For me, “I asked Google” is a marker of epistemic status. It’s saying I did not possess any innate mastery of this topic before this conversation but I’m capable of using digital tools to rapidly gain conditional mastery. I’m explicitly bracketing that my claims come from my ability to use digital tools vs pre-acquired knowledge. To that end, I’ll mention the specifics of the tool if it matters to how you interpret the question and obscure it if it feels irrelevant to the task. ie: Sometimes, I’ll provide the exact search query I used. Sometimes I’ll say this took me multiple googles and provide each false avenue before the real avenue. Other times, I’ll omit all that and just provide the results.
I will still always explicitly call out if the answer was from a Search of pre-written Human knowledge, assembled into links or a AI query that synthesized the knowledge. I’d say “Google’s Gemini said X” instead of “I Googled and got X” but that’s my distinction.
In other words:
from xkcd: Extended Mind from a good 20 years ago. IOW, this is not a new problem. If it really is a problem.
There are people who insist on posting to topics they know little about using new or old google / wiki as their sole source of knowledge. And it usually shows. How it shows is different from the results of an old fashioned search and follow-up versus some posting some AI generated content. But show it does.
But far more often we all do things like check google / wiki for some half-remembered detail. For sure when I don’t do that, and post from fallible memory about something I read 20 years ago, that’s when I say something stupid. And feel foolish for having scrimped on the teeny effort needed to confirm that, e.g. Obama was president from 2009-2017, not 2005-2013.
And IMO the board is a better place for folks who have the diligence to confirm their memories, not just run with it.
IMO this is the nub right here:
To some degree FQ is still treated as a legit source of hard fact answers to hard fact questions. The old fashioned manual personal expertise talking version of google or wiki if you will.
But the whole rest of the 'Dope is about yakking with friends. Ideally friends who’re a little less prone to raw BSing than most, but still friends.
And if I was sitting around a table with y’all with some beers, and somebody had their pocket encyclopedia with them and kept up their end of the conversation only by quoting from it, well, they’d be a right bore to hang out with. And would probably not be invited back next week.
Do we need a cite for something like that? or even a “I asked Google”?
Generally when I say that, it is to note how many cites there are, e.g.- “I checked Google and i couldnt find an answer” or " I checked google and there are several dozen cites".
Yeah…it sucks when someone can call you on your bullshit on the spot and can prove it in a moment. Drinking with suck-ups and yes-men is so much more fun.
Many people hate that. I dont mind.
My understanding is that mods will remove large chunks of a Wikipedia article, just as they would of a copyrighted piece. (Technically, Wikipedia is copyrighted but operates under a Creative Commons License in which it allows free and full use as long as credit is given. Works can be put into the public domain, though, as shown on that page. Copyright is tricky all the way down.)
The Straight Dope worked wonderfully as a column in the 70s and 80s when hunting down even basic information was difficult and time consuming. When the AOL version started in 1999 that was still largely true, although the 99ers favored creating a stew of personalities rather than answering questions from the bits I’ve seen. FQ/GQ started declining after Google became an instant source. The mods had to ban “why don’t you just Google it” posts? People want their community to respond, accuracy being a lesser priority.
AI is terrible at almost everything now done on the board. Non-factual discussions encourage posting of statements based on individual experiences and circumstances providing foundations for opinions, analyses, and speculations. AI doesn’t have cultural likes and dislikes, preferences in food, reasons for avoiding an artist’s work, memories of childhood, or hopes for a future. AI may hallucinate, although certainly less and less as time goes on, but it will never post drunk or high, in an emotional fury over a personal situation, or in reaction to comments made about one’s religion, politics, or sexual predelictions.
A message board may be the best internet ammunition against the takeover by AI. We’re idiosyncratically and obstinately human. Lots of posters have left because of that fundamental truth, others stay on because or despite that reality. AI doesn’t stand a chance.
That seems to me like a long and convoluted way of describing a marker of chronological status – that is, knowledge that I had long ago versus knowledge that I’ve just acquired via Google or my favourite AI engine. What makes one more valid than the other?
The knowledge that one has from in-depth training and experience is certainly more valuable and dependable than the result of a Google query, but I see no qualitative difference between some dimly remembered factoid that may or may not be accurate and a Google response that may or may not be accurate.
No, it’s not mainly about time, it’s about the relation to my other knowledge. In areas I’m familiar with, I’m able to vet links, figure out if there’s discrepancies, notice what are the popular-but-incorrect framings and winkle out the hard to find specialist view that differs/adds nuance to the popular view.
With knowledge out of my wheelhouse, I’m explicitly disclaiming this ability and putting the onus on you. I might say “a Google reveals that Iran’s major exports were X” with the full knowledge that export data has all sorts of idiosyncratic quirks without a knowledge of what those quirks are. I can say “The US military claims the top speed of this missile is X mph” with the full knowledge that the military lies and understates those figured but not the texture of that understatement.
VS areas I have expert knowledge on, I can share my own personally derived wisdom alongside the data. I don’t need you to agree with the wisdom, but the wisdom is coming from a defined perspective that should be taken into account when understanding my arguments.
IMHO, where it becomes an issue is with posts where the only poster-created content is the “I asked Google” followed by several sentences or even paragraphs taken straight from the AI. Again, we explicitly don’t want an AI response to be the entirety of the post.
There’s also the community aspect involved again. If one of our Aviation experts expound on something, I’m going to take their word on it, because they have experience on it. Yes, I don’t know that said expert is remembering accurately, or BS’ing, but that’s back to the community aspect of earned trust.
If someone quotes AI, I want a link or three. I want to see where the knowledge is coming from. With the requirements in the TOS, I at least can use the same AI and create a similar question to see if I can get the cites.
I personally only use AI responses for something quick I don’t care much about, or as an advanced search engine (duh) to get the links I need to evaluate.
Not that it doesn’t do a pretty good job as a search engine, but in my personal experience, it’s just… wrong so often, on simple, factual answers that I can’t trust it.
Yesterday afternoon I was playing Starfield, a recent-ish video game, and I asked a question about specific ship-building elements. It hallucinated three different conflicting answers, and when I tried to correct it, it gave out another one.
That’s for a very low consequence bit of research. And I don’t “blame” it for being wrong (it’s not a person) but if I want to discuss a subject of consequence, I darn well want to know where it’s getting it’s information from.