• Guests may view all public nodes. However, you must be registered to post.

Suggestion; remove AI writing in the reports

Status
Not open for further replies.

Winnitude

Active member
Joined
Nov 16, 2022
The latest update reads so strongly like AI writing that it loses credibility and leads to constant questioning while reading due to the inherent issues with AI models and reporting current events. Almost reads like an AI spam site. This wasn't that bad until the end which includes this lovely bit of text:
68.jpg
Including notes from the AI for the person posting directly in the report does not help read as credible when the same kind of proofreading required to catch hallucinations is the same to catch stuff like this. It just reads like a copy paste job from a chatbot that could have any number of subtle issues in it when some people rely on the reporting on this site.
 
Winnitude said:
The latest update reads so strongly like AI writing that it loses credibility and leads to constant questioning while reading due to the inherent issues with AI models and reporting current events. Almost reads like an AI spam site. This wasn't that bad until the end which includes this lovely bit of text:
View attachment 9085
Click to expand...
Preparing and releasing these FREE reports and free videos Monthly and/or every time needed is shitload of work.

Pointing out a small and even embarrassing error politely is the right way to show appreciation for all the constant 24 a day attention and labor.

Winnitude said:
Including notes from the AI for the person posting directly in the report does not help read as credible when the same kind of proofreading required to catch hallucinations is the same to catch stuff like this. It just reads like a copy paste job from a chatbot that could have any number of subtle issues in it when some people rely on the reporting on this site.
Click to expand...
Being petty is simply petty.
 
Torch said:
Pointing out a small and even embarrassing error politely is the right way to show appreciation for all the constant 24 a day work.


Being petty is simply petty.
Click to expand...
It's not pettiness, I use AI tools pretty frequently, but on a site like this it undermines the reporting when the reports are being obviously written in the voice of a tool that is known for making errors, and pasting in blocks of text that don't even apply to the reader only adds to it by undermining any assumptions about proofreading quality.

"Preparing and releasing these FREE reports and free videos Monthly and/or every time needed is shitload of work."
This website has existed a long time before AI models existed. So its not necessary to the operation of the site. It is necessary to pump out increased quantity beyond what was possible before, but the question of quantity vs quality, and necessity of quantity, is an important one. The videos as they are looking identical to scam ads and videos immediately would make younger tech savvier audiences coming across them first write off DEFCON immediately as an AI spam site. Does this addition add to the publication or take away from it's credibility? Pumping out content should not be the express goal. Likewise with the shift in article quality and tone towards reading like any other AI generation with a DEFCON prompt. Does the increased capacity to pump out more reports make them more useful when they undermine trust at the same time?
 
Winnitude said:
It's not pettiness, I use AI tools pretty frequently, but on a site like this it undermines the reporting when the reports are being obviously written in the voice of a tool that is known for making errors, and pasting in blocks of text that don't even apply to the reader only adds to it by undermining any assumptions about proofreading quality.

"Preparing and releasing these FREE reports and free videos Monthly and/or every time needed is shitload of work."
This website has existed a long time before AI models existed. So its not necessary to the operation of the site. It is necessary to pump out increased quantity beyond what was possible before, but the question of quantity vs quality, and necessity of quantity, is an important one. The videos as they are looking identical to scam ads and videos immediately would make younger tech savvier audiences coming across them first write off DEFCON immediately as an AI spam site. Does this addition add to the publication or take away from it's credibility? Pumping out content should not be the express goal. Likewise with the shift in article quality and tone towards reading like any other AI generation with a DEFCON prompt. Does the increased capacity to pump out more reports make them more useful when they undermine trust at the same time?
Click to expand...
The only thing I would say to this is this.
The scope and volume of data and reporting has vastly expanded for the site good or bad. Personally I don’t even try to keep up with all the world events or reports.
So I won’t judge them too harshly, if you don’t use AI you just as likely opening yourself up to grammar Karens.
 
Publius said:
grammar Karens.
Click to expand...
This. AI is fine to use if you're using it just to grammar check things. I do. I'll write my tweets or reports. Then run it through AI to make grammar & punctuation corrections. Because, you know, grammar Karen's. Everything has to be held to impossibly high standards now a days.

But to use AI to make content is a different story. I do not agree with that, however...

But if they just using AI to help proofread their own material they made, than nothing wrong with it in my eyes, I do the same. Still your work. No different then having a teacher for example proof read your work for spelling and grammar issues.

After a teacher proof reads/edits your writing essay, is it not still your work/essay? Just saying.
 
If memory serves, the director puts down some bullet points of relevant news, asks staff for input on anything else that's happened, then goes off and writes it on his own, then puts a draft out for staff to read (can't remember if it happened for all of the updates but it did for the important ones I think).
Really it's up to @DEFCON Warning System what he wants to do, I don't blame him for using AI to help with the report for grammar etc... it probably saves an hour or so, increasing productivity. In this instance obviously there was a mistake, and I think that's a learning opportunity for staff to check it more thoroughly before it goes out, perhaps a 2 man rule where the update needs to be read by another staff member for errors.

As for @Winnitude 's comments, I do think when you see obviously AI written stuff it does hurt credibility. I get why it's been done but in a world where the web seems to be increasingly taken over by AI slop DWS needs to stand out from that.
 
Friendly Engineer said:
If memory serves, the director puts down some bullet points of relevant news, asks staff for input on anything else that's happened, then goes off and writes it on his own, then puts a draft out for staff to read (can't remember if it happened for all of the updates but it did for the important ones I think).
Really it's up to @DEFCON Warning System what he wants to do, I don't blame him for using AI to help with the report for grammar etc... it probably saves an hour or so, increasing productivity. In this instance obviously there was a mistake, and I think that's a learning opportunity for staff to check it more thoroughly before it goes out, perhaps a 2 man rule where the update needs to be read by another staff member for errors.

As for @Winnitude 's comments, I do think when you see obviously AI written stuff it does hurt credibility. I get why it's been done but in a world where the web seems to be increasingly taken over by AI slop DWS needs to stand out from that.
Click to expand...
As I said. It's one thing to have AI proof read your own work you made. But to have AI create your content is where I draw the line personally.

Hopefully that's how the director uses it, proofing his own work he makes, not having AI make the content.

I never do and never would use AI to create my content. Only to proof read the content I made. Because I don't want to give grammar/punctuation Karens/Nazis the light of day or derail my content I make.

If I could afford it I would hire a real person with an English major to proofread my own work/content that I make. BUT that's not realistic. So AI it is to fix my poor English writing. I hate English... So much, it's so "extra" or unnecessarily complicated.
 
👉However the AI has been teaching me how to be more proficient in my English writing. After awhile you just pick up on what is grammatically correct or not. I've recently even been using it less and less as I learn how to write better though AI proof reading. So. If you actually pay attention, you can learn to not depend on it and become better at English yourself over time. That's if you actually pay attention to what it corrects on your work and not just lazy copy and paste into AI.

Example all posts here not proofread by AI. Before years ago this would have been a very sloppy, poorly worded, insufficient grammar posts. Now I am able to articulate myself in writing much better over the years by learning via AI proof reading on how to properly write stuff in general. I used it to better myself, not be lazy. Soon probably won't need it at all as I become more comfortable with my writing.
 
I’ve read Steinbeck and Wolff and realized Grammar is mailable.
It should however at least try and be clear in intent.
Language is not mathematics
 
ALSO... AI has become extremely useful in translating extremely local dialect or foreign languages that typical translators have a difficult time with.

Also translating things in photos/videos is another benefit of AI for a OSINT p.o.v.. For example if a local in (whatever country) posts something on social media, you can have AI translate the signs on the road or what not, or translate whatever they're saying in a video if it's in a foreign dialect and you can't find someone in that dialect to translate it for you in that moment.

Important distinction though! AI can help you understand what you may be looking at faster, but serious reporting still requires source verification, context checks, and ideally native confirmation when the translation matters before posting.
 
All reports are hand crafted. AI is used to help pull it together. Otherwise I'd be spending far more time than I have writing it out.

The Staff monitor events and help with analysis. But as for the actual reports that are publicly put out, I do everything.
  • Compile
  • Edit article
  • Edit script
  • Publish article
  • Update site
  • Send announcements
  • Record video
  • Edit video
  • Upload video
I can't do all this by myself. It's too much.

It was easier back in the day when we only did one update a month, and that update was just essentially a bullet point of events that was uploaded.

Now it's grown to events and analysis, sometimes with associated sources. Plus more involved videos because some people like those. (Honestly given the number of views they get, it's not worth the effort. But when a crisis hits, they will be needed.)

So I use AI to help lighten the load. I see AI as a tool. We've been using AI for decades. That spell checker people use? That's AI. Grammar checkers? AI. Google search. AI.

It's all AI.

Rejecting AI is Luddite. It's here. It's going to be here. Use it or be left behind.

It takes a huge burden off of my workload.

The error the OP noted was a comment AI made about the quality of my analysis.
 
DEFCON Warning System said:
I see AI as a tool.
Click to expand...
DEFCON Warning System said:
Rejecting AI is Luddite. It's here. It's going to be here. Use it or be left behind.
Click to expand...
Yep, couldn't of said it better. ↕️
NuclearID said:
ALSO... AI has become extremely useful in translating extremely local dialect or foreign languages that typical translators have a difficult time with.

Also translating things in photos/videos is another benefit of AI for a OSINT p.o.v.. For example if a local in (whatever country) posts something on social media, you can have AI translate the signs on the road or what not, or translate whatever they're saying in a video if it's in a foreign dialect and you can't find someone in that dialect to translate it for you in that moment.

Important distinction though! AI can help you understand what you may be looking at faster, but serious reporting still requires source verification, context checks, and ideally native confirmation when the translation matters before posting.
Click to expand...
 
DEFCON Warning System said:
All reports are hand crafted. AI is used to help pull it together. Otherwise I'd be spending far more time than I have writing it out.

The Staff monitor events and help with analysis. But as for the actual reports that are publicly put out, I do everything.
  • Compile
  • Edit article
  • Edit script
  • Publish article
  • Update site
  • Send announcements
  • Record video
  • Edit video
  • Upload video
I can't do all this by myself. It's too much.

It was easier back in the day when we only did one update a month, and that update was just essentially a bullet point of events that was uploaded.

Now it's grown to events and analysis, sometimes with associated sources. Plus more involved videos because some people like those. (Honestly given the number of views they get, it's not worth the effort. But when a crisis hits, they will be needed.)

So I use AI to help lighten the load. I see AI as a tool. We've been using AI for decades. That spell checker people use? That's AI. Grammar checkers? AI. Google search. AI.

It's all AI.

Rejecting AI is Luddite. It's here. It's going to be here. Use it or be left behind.

It takes a huge burden off of my workload.

The error the OP noted was a comment AI made about the quality of my analysis.
Click to expand...
Right, if the AI comment gets included directly on the report, it reads like it was pasted from the AI directly, which reduces credibility. While I've been called petty and a Luddite and told I should be thanking you guys, my point was never that AI is bad, but that it being so obviously written in the voice of an LLM hurts the credibility of the site. It's a massive flood of expanded content yes but if the quality goes down and the credibility is gradually destroyed due to the obvious use of extremely error prone technology (not just to collect info and such but to pull together and then copy paste its output?) it doesnt seem worth increasing the content schedule.


Outright rejecting AI would be Luddite behavior. Pointing out that the tool is being used where it is poorest (actually writing final copy content and generating speech, videos) and most damaging to institutional credibility rather than where it is best (pulling sources and creating rough drafts that are then rewritten by human hands to ensure correctness OR massaged with prompts repeatedly to fix the obvious AI voice that makes the content read as low effort spam no matter how high effort it really is) is valid and constructive criticism.

I use AI tools all the time at work; it's both obvious and tangibly destructive to our product when they're allowed to write copy or code that makes it into shipped releases, that doesn't make the tech awful, its just not applicable in every situation despite the marketing and FOMO cycle. Like DarkNoon said, when the whole web is identical pumped out AI slop for quantity, DWS cannot stand out if it becomes the same (or looks the same even if the underlying info is better researched and we assume the AI makes no subtle changes or errors the director doesn't notice, which would be much subtler than things that already dont get noticed with the AI writing, like the above screenshot). Previously I assumed it must all be proofread only because I already trust the site, new user does not have that, and stuff like this undermines further.

It sounds like either massively expanding the content schedule because it's POSSIBLE was stretching yourself too thin and AI can't adequately fill the gaps without drastically reducing the perceived quality or credibility of the site, between the videos (especially the videos, identical format to many fake news channels using the same tech, and no, real news organizations are not going to jump to this exact format any time soon, it's not ready and acknowledging this does not a luddite maks) and the site. It sounds like it should be justified by either having staff or volunteers help with those parts. AI may have made it more efficient enough in theory to do it but in practice it seems like it produces content with negative value much of the time.
 
Last edited:
Winnitude said:
Right, if the AI comment gets included directly on the report, it reads like it was pasted from the AI directly, which reduces credibility
Click to expand...
That was an error. I was pouring over reports and script and sometimes I just miss things.

I understand your comment. I made an error. I will do my best to not repeat it.
 
I don't have an issue with using AI to help organize work or proofing. It's a useful tool to help reduce workload, especially for mundane tasks. But AI is often wrong, regularly self-contradictory, and requires human oversight/review.

AI is based on large amounts of data run through statistical models. I am reminded of an old quote (Mark Twain?): "There are lies, [darn] lies, and statistics." It's a thing that makes me go "hmmm". 🤔 AI should always have human oversight because, after all, it is a human who is accountable for the decision that is acted upon.

The way in which @DEFCON Warning System describes the use of AI seems reasonable and prudent. It's clear the director performs independent analysis and generates his own content. AI seems to be used as a secretary or administrative assistant, a fundamental service AI provides. In fact, given the rate and frequency at which the public and forum members desire analysis/updates, using AI seems justified to meet demand.

To the point @Winnitude makes, in this era of AI slop, extra proofing to clean to AI responses can help to prevent some visitors from formulating a negative impression. It's clear that @DEFCON Warning System and the staff take their mission seriously (something I value since I haven't the brains nor the resources to do this type of research myself). I wouldn't want to see that minimized by AI's sloppiness.
 
Ok so most have had a say and a rebuttal. I personally apologize for strong statements when lessor would have worked. @Winnitude unless you have a further rebuttal, Im closing this out. Please pm me and I will accommodate you.


Thank you all.
 
Status
Not open for further replies.
Back
Top Bottom