AI usage in responses

[AmmoTech90]

Is there anyway we can regulate, ban, reduce the amount of AI slop that is replicated on this site. I am tired of some one saying, "I'm not sure but this what <AI bot xyz> says:". Followed by a bunch of text that can't be trusted.

No value added. You have to verify everything in that statement.

If you do know about a subject then go ahead and comment; if you aren't familiar, find reputable sources and learn about the subject and then comment, or just don't say anything.

Maybe who ever starts a thread can label it "No AI please." Responses AI with get nuked. Sorry for the extra work mods.

Anyway, I'm getting tired of seeing walls text that killed dozens of electronic trees for no benefit.

 

[PuckChaser]

Something we can look into but it likely wouldn't be something automated which makes it difficult for us to moderate. I hate it as well.

I would suggest if you find particular users constantly using AI with no independent thought you put them on ignore or don't reply to their post. The community can shame out the folks that are doing it too often better than the mod squad.

 

[Fishbone Jones]

AI is just a tool. Like any other media source out there. We know CBC, etc sometimes get it wrong, but that doesn't stop people from posting full page articles from them as proof of point. Are we going to censure every Wikipedia article? It is a much more dubious source than AI, yet we allow it.

I have no doubt Ammo Tech's complaint arose from my post.

All I'll say to that is, AT take a minute, go through it and please point out where the untruths or bullshit are. I'm interested in what you find.

As with every other thread here, if you find it too onerous, skip it. Nobody is forcing you to read it.

Do we really want to censure information, just on a single users complaint or because it makes some people uncomfortable? Sounds like our current government.

 

[The Bread Guy]

If you do know about a subject then go ahead and comment; if you aren't familiar, find reputable sources and learn about the subject and then comment, or just don't say anything.

Appreciate your frustration, but best we can do is treat it like any other source. Also, like any search engine result, showing the system used and the question asked/prompt entered shows more transparency. The more of the work shown, the better one can judge the info. If in doubt, it’s always legit to ask to “show your work” with any sources/information shared, as is done with media sources.

Also, in spite of warnings/caveats that AI can screw up, the buck stops with whoever shares the info.

 

[AmmoTech90]

I agree that attribution checking is important.

However every bit of information from AI needs to be questioned. If some called Stoker Dave posts something about maritime engines I'll treat it with trust until Dave proves himself unreliable (I'm sure he won't). With AI there is no inherent trust.
Combine that with the walls of text it produces there is less reason to engage with content on the site. Less AI means more meaningful engagement with the forum.

 

[mariomike]

in doubt, it’s always legit to ask to “show your work” with any sources/information shared, as is done with media sources.

For non-fiction , I walk to the library.

 

[Fishbone Jones]

I agree that attribution checking is important.

However every bit of information from AI needs to be questioned. If some called Stoker Dave posts something about maritime engines I'll treat it with trust until Dave proves himself unreliable (I'm sure he won't). With AI there is no inherent trust.
Combine that with the walls of text it produces there is less reason to engage with content on the site. Less AI means more meaningful engagement with the forum.

The sources AI used for my answer are all attributed after the subject of the paras.

thecanadianencyclopedia.ca
arcticreview.no
fasken.com
edmontonjournal.com
s3.documentcloud.org
ici.radio-canada.ca

One would have to assume that if AI used these references and AI is full of shit, then the sources must be as well. Unfortunate. Some of them are pretty prestigious organizations.

 

[AmmoTech90]

Do you actually check to see if AI has modified, misinterpreted, or otherwise distorted that information? That's one of major problems with AI, it can provide the answers you want; these are not necessarily the truth.

 

[Fishbone Jones]

I suppose we should probably go all the way down the rabbit hole and check the stories and sources of the journalists who are writing the references. And why stop there. Let's track down the participants in the journalists stories and make sure they are telling the truth? Sometimes, you just have to accept things until proven otherwise.

 

[The Bread Guy]

One would have to assume that if AI used these references and AI is full of shit, then the sources must be as well. Unfortunate. Some of them are pretty prestigious organizations.

Not true. One person can take the world's best quality flour, water, salt and yeast and make artisan bread, while someone else takes the same high-end ingredients and make slop. AI doesn't "read" the content, or apply logic, it sorta-kinda figures what what stuff "belongs together," and can (for one example) take accurate facts and come up with screwed up conclusions by putting them together wrong.

Also, if you don't trust "in the bag" and "bought and paid for" media (which all AI systems draw from) to get it right - and maybe I'm mistaken on how you and others feel on that front - now you trust AI looking at that very same media to get it right?

Also, what someone asks and how can change the response one gets, just like when using Google & other search engines - as well as previous search histories and questions asked of the AI wizard in question.

I suppose we should probably go all the way down the rabbit hole and check the stories and sources of the journalists who are writing the references. And why stop there. Let's track down the participants in the journalists stories and make sure they are telling the truth?

Or we can always go with the "yeah, this may not all be entirely and factually correct, but you knoooow what I mean," just like with memes on social media with incorrect information, right?

Bottom line is that consuming information, whether it's from media, from institutions (including government) or from individual members/users, relies on risk management. How much do you trust the source? How right/wrong have they been in the past? How does what they see match what you see? AI doesn't do it this way, and everyone wears different glasses when asking themselves these questions, so readers need to have grains of salt ready and posters should be able to explain/source/defend in detail what they post, AI generated or not.

 

[SeaKingTacco]

For what it is worth, when I see a post that is a wall of AI text, I won’t even bother to read it and skip to the next post.

The post is waste of time.

 

[brihard]

For what it is worth, when I see a post that is a wall of AI text, I won’t even bother to read it and skip to the next post.

The post is waste of time.

Same. I’m not interested in a wall of regurgitated AI text. I want discussion where the other person has given some time and effort to understand a subject and, where it fits the discussion, to have their own informed views. And I try to hold myself to that standard too in what I do and don’t post. But, to each their own.

 

[mariomike]

I’m not interested in a wall of regurgitated AI text.

Perhaps , as a compromise for those who are , would be a designated board to contain AI slop ?

 

[Colin Parkinson]

When I use AI, I state that this is AI and often for data gathering it does a fairly reasonable job. In a way it's informative to see what AI says, as that is what will be informing the majority of laypersons about a particular subject.

My wife uses AI daily in her job and her research into various subjects. Which AI you use and how you frame your questions matter. Chatgpt will try to please you and give you what it thinks you want to hear. Gemini and perplexity in her opinion offers better content and framing.

 

[ArmyRick]

I will use AI in some of my answers, especially when it is certain info I am looking for. I always provide a heads up and warn people to take with a grain of suspicion. I have caught AI making errors (glaring one being that it mixed up IRCHA and 1RCR in a question I asked Google AI once).

Being totally "No AI whatsoever" to me is unrealistic. Its here and will remain here for the foreseeable future. Its a tool, use it correctly. As my wife says to me sometimes "Is your spoon not sharp enough?". Use AI at the correct time and in the correct way.

 

[mariomike]

How much do you trust the source?

I trust internet chat rooms.

But, prefer to verify "facts" posted in them at the library non-fiction department.

 

[RangerRay]

I am not impressed with AI generated answers. Once I asked it about local hunting regulations. Had I not already been familiar with the Regs and followed what AI answered, I would be incurring 1000s of dollars in court costs and fines and been suspended for a long time. It has a bad habit of telling you the answers you want and lying to you to do that. I found that when I clicked on the link provided where it says it got the answer, it was the completely opposite what the original link said.

And that is with questions posed as neutrally as possible. If you asked it “Why are the Tories/Liberals/Jews terrible”, it will not say “Actually, they are not terrible at all and here are the reasons”. It will probably give you an answer based on you previous search history and algorithms.

My wife is seeing this more and more in education. Students are submitting papers with fake citations and footnotes created out of whole-cloth. This is usually from being AI generated, or asking AI for citations, with creates fake citations.

Don’t trust, and definitely verify.