Member request: please report any suspected AI Posts.

splitimageview

Administrator
Local time
4:21 AM
Joined
Aug 30, 2007
Messages
10,248
Location
Austin
Members: if you come across posts that appear to be AI-generated, please report them.

Mods have a difficult enough time with spam posts that are not related to photography, but it appears the spammers are now using AI for relevance, in order to infiltrate.

We want to rid the forum of these accounts.

I'm leaving this post in place (see link below) as an example.

(The Report button link appears below every post.)

Thank you!

 
Can you please explain how this post ("Japanese Summilux") is AI-generated/spam? So we know what to look out for.
Thanks.
 
Can you please explain how this post ("Japanese Summilux") is AI-generated/spam? So we know what to look out for.
It's the very last post in that thread at the time of writing:

1767711172947.png
Brand new account with only one post and all of the hallmarks of "AI"-generated text in the post; not only is it regurgitating old info without offering anything new, but it uses certain key words and grammatical structures that become easy to spot once you've seen them a few times.

I love this place because it's a relative sanctuary from the low-effort attention-seeking posting behaviour and the flood of GenAI slop elsewhere on the internet. If this stuff starts clogging up here as well, it'll be a sad day.
 
Sure, these are some of the red flags for that post:

1. it was a new account that has never posted before.
2. it's a reply to a five year old thread, with a 2 1/2 year gap to the previous post.
3. the wording/phrasing itself screams "I'm not a person"
4. the odd reference to RFF itself: According to forum posts, members on Rangefinderforum and Leica Society International...
5. references to Reddit posts

Note that 1 and 2 are not always indicative of spammers, but it is common for them to attempt to establish legitimacy by adding posts to old threads, flying under the radar...
 
Sure, these are some of the red flags for that post:

1. it was a new account that has never posted before.
2. it's a reply to a five year old thread, with a 2 1/2 year gap to the previous post.
3. the wording/phrasing itself screams "I'm not a person"
4. the odd reference to RFF itself: According to forum posts, members on Rangefinderforum and Leica Society International...
5. references to Reddit posts

Note that 1 and 2 are not always indicative of spammers, but it is common for them to attempt to establish legitimacy by adding posts to old threads, flying under the radar...
I guess it's your #3 that is the challenge for me -- I don't think I can easily spot something that indicates it's AI-generated. Heck, so many people (not forum members here!) are lousy writers so I appreciate something that is well written. Examples would help here.
 
GenAI stuff is rarely "well written". It ticks a lot of grammatical boxes but it has no soul; there's no real "voice" to it. No sense of character.

Quite often, it uses repetitive syntax, lists, and paragraphs of weirdly similar lengths. That's the easiest thing to spot. You can see how the generated post uses "According to..." over and over again, for instance.

It also just outright makes up nonsense because it's not "intelligent" at all - it's just fancy auto-correct. The phrase "The Reason behind the male japanese names" is almost certainly ChatGPT having insufficient material to create context or "understanding" for the conversation at hand. Sometimes it won't be something as obvious; part of the reason ChatGPT et al. have become so popular is that they appear to "know" a lot to people who don't know anything about the subject, but the text its generating includes obvious mistakes (or just outright garbage) to anyone with slightly more than a surface-level knowledge of said subject.

Like I said, it gets a lot easier to spot once you've seen it a few times. The various GenAI systems are getting better, but they're still a long way from human. We don't need Voigt-Kampff tests yet - things they're spitting out are still landing firmly in the Uncanny Valley.

God, I'll be glad when this bubble bursts.
 
It's not difficult to spot the tell-tale signs once you've seen multiple examples of AI results. Bear in mind it all depends on what prompts are used to generate the text.

A spammer is not going to put in a whole lot of effort here fine-tuning prompts, methinks. This is more than likely a bot that was able to register, once in, the bot picked an old thread and a few key words from it and used those keywords for the AI prompt to generate the reply.

Generally speaking, human posters do not reply to these sorts of threads in this stilted, academic manner with subheads and bullet points.
 
Can you please explain how this post ("Japanese Summilux") is AI-generated/spam? So we know what to look out for.
Thanks.
Things like the following are a hint:

"Accuracy: According to forum posts, members on Rangefinderforum and Leica Society International claim that the Canon lens really ..."

Or

Unique Character: According to fans on Reddit and the Leica Forum, the Canon lens..."
 
Forum posts are typically conversational, as opposed to presenting an academic response.

Just for kicks, I sent the text thru ChatGPT with the following prompt:

To illustrate this to forum users, could you rephrase this paragraph so it sounds like a real person wrote it? Dead giveaway this is AI generated are the referencesto 'users on reddit, youtube, rangefinderforum, Leica forum', especially since this was posted on the rangefinderforum!

Here’s a more natural, forum-authentic rewrite that keeps the substance but removes the very obvious AI tells. I’ve written it in the kind of voice you’d expect from a knowledgeable rangefinder shooter posting from personal familiarity, not citing “users on Reddit/YouTube/etc.”:

The Canon 50mm f/1.4 LTM is often called the “Japanese Summilux,” but that label has always been a bit awkward. The Canon actually predates Leica’s Summilux and stands on its own as a serious lens from the late 1950s, with a different design philosophy and rendering. It’s a fast f/1.4 in a compact screw-mount package, which made it an impressive and relatively affordable alternative at the time, but that doesn’t mean it was trying to be a Summilux.

Optically, it has its own character—smooth bokeh, a bit of classic glow wide open, and very good sharpness for its era—distinct from Leica’s take. Some people find the comparison helpful as shorthand, others see it as flattering, but many feel it undersells what the Canon actually is. Given that the Canon came first and has a unique look, calling it a “Japanese Summilux” is historically misleading and doesn’t really do the lens justice.

If you’d like, I can also:

  • Make it sound more casual / more opinionated
  • Dial it back to a shorter reply-length post
yes, make it sound more casual, LESS opinionated if possible.

Sure — here’s a toned-down, more conversational version that sounds like a regular forum member trying to be clear rather than make a point:

The Canon 50mm f/1.4 LTM often gets referred to as the “Japanese Summilux,” but that comparison doesn’t really tell the whole story. The Canon lens actually came first and was already a very fast f/1.4 lens in a compact screw-mount body, which made it stand out on its own at the time.

In use, it has a different look from a Summilux. Wide open it has a bit of that classic glow, with smooth bokeh and good sharpness once stopped down. Some people find the nickname helpful as a rough point of reference, while others feel it’s unnecessary since the Canon has its own character and history. Either way, it’s probably best understood as its own lens rather than defined by comparison to a Leica design.

it's still not sounding/reading like a personal viewpoint. any other tweaks?

Got it — the key is to make it read like lived familiarity, not a summary. That means: fewer complete explanations, a little imprecision, and phrasing that sounds like someone thinking out loud rather than presenting a position. Here’s another pass with that in mind:

I’ve always found the “Japanese Summilux” label for the Canon 50/1.4 a little odd. The Canon was already around before the Summilux showed up, and when you actually use it, it doesn’t really feel like it’s trying to be one anyway.

To me it has its own look—nice bokeh, a bit of glow wide open, and plenty sharp once you stop it down. I get why people use the comparison as shorthand, but it also kind of glosses over what makes the Canon interesting on its own. I’ve never thought of it as a Leica substitute so much as just a really good fast Canon lens from that era.

If you want, I can:

  • Make it sound even more off-the-cuff
  • Add a small personal hook (“I’ve had one for years…”, “On my M2…”)
  • Or make it read like a quick reply mid-thread, not a standalone post
 
Wow, thanks for the context and tips on spotting these. It's all really disturbing, as we keep getting closer to a Blade Runneresque world.

What is the point of these AI bot posts anyway? Are they redirecting you to an external link?
 
Members: if you come across posts that appear to be AI-generated, please report them.

Mods have a difficult enough time with spam posts that are not related to photography, but it appears the spammers are now using AI for relevance, in order to infiltrate.

We want to rid the forum of these accounts.

I'm leaving this post in place (see link below) as an example.

(The Report button link appears below every post.)

Thank you!



How do we know you're not an AI, hrmmm? An AI would do just that - demand we stop the AIs. It's a Goedelian Knot, I tell ya.

I kid, I kid. I'm glad we're taking active measures against this crap. I spent the last few years doing interdiction against AI-enabled 'bots and they are truly awful. Well over half the social media content these days is likely AI and it's getting worse.
 
I did a blog post recently on a chatbot response I got on Photrio. Here is an excerpt:

My question:
-------------------
I recently acquired a Yashica YE. Everything works well except the frame counter which does not advance.
I would appreciate any insights on getting into the counter mechanism and fixing the problem.


Here was the first response I got:
--------------------------------------------
Could be a number of things. You will have to open the camera and examine the counting mechanism to see what is wrong.

Here is a case where a little plastic piece broke, causing film counter failure...


The lack of specificity in the reply along with veering off into a lengthy explanation of an unrelated problem signaled that the response was AI-generated. Someone had clearly just submitted my question to a chatbot and posted the result without further elaboration.
 
Back
Top Bottom