User maxim2511 sounds like a chatbot.

The volume & tonal consistency of @maxim2511's posts suggests to me that they are ChatGPT. Also the mistakes in their posts are weird.
https://cplusplus.com/user/maxim2511/

So I put some of their posts into OpenAI's text classifier, which tries to identify computer-generated text:
https://platform.openai.com/ai-text-classifier

For instance, this one:
https://cplusplus.com/forum/beginner/285364/#msg1239402
And this one:
https://cplusplus.com/forum/windows/285291/#msg1239396
And this one:
https://cplusplus.com/forum/beginner/285286/#msg1239407
And this one:
https://cplusplus.com/forum/beginner/285380/#msg1239398

OpenAI's tool says "likely" or "possibly AI-generated" for each of their posts that I tried. Note it needs at least 1000 characters of input as a sample.
Last edited on
Especially the 2nd link looks exactly how ChatGPT likes to respond. The beginning and ending of a ChatGPT response is very consistent. Acknowledges the problem at the beginning, gives you a little wrap-up at the end.

By giving ChatGPT some prompts on how it should speak, it makes it harder for tools to know if it's AI generated. However, my experience is that "likely" and "possibly AI-generated" usually correlate to it being AI generated with a few speech prompts inserted.
However, my experience is that "likely" and "possibly AI-generated" usually correlate to it being AI generated with a few speech prompts inserted.
OpenAI's tool returns one of five responses:
"Each document is labeled as either very unlikely, unlikely, unclear if it is, possibly, or likely AI-generated."
https://platform.openai.com/ai-text-classifier
So it's unclear what the probabilities actually are.

I don't want to discourage the use of chatbots etc., but if my assumptions are right, it would be much better in my opinion, to collaborate with it instead of presenting its output as a human's response.
Unless you specifically direct how ChatGPT should structure its paragraphs and sentences, it gets detected easily.

If you're not looking for a formal response, then it's easy to trick OpenAI's tool. The best one I've seen so far is this one:

https://gptzero.me/


For example, this text will trick OpenAI's detector, but not GPTzero's. Also note how I tend to identify ChatGPT quickly, which is the opening paragraph states the topic, and the ending paragraph is almost always a formal conclusion:

Yo, yo, yo! What's up, my homies? It's ya boy Kyle back in the house, ready to chop it up about one of the most underrated creatures in the animal kingdom: squirrels.

Now, I know what you're thinking. "Squirrels? Really, Kyle? They're just little rodents that run around in trees and steal our birdseed." But hear me out, fam. These little dudes are actually pretty dope if you take a closer look.

First off, squirrels are straight-up acrobats. I mean, have you ever seen one of these guys leap from branch to branch like it's nothing? It's like they're performing some crazy parkour routine up there. And let's not forget about their climbing skills. Squirrels can scale trees and buildings like it's no big deal. I bet if you put a squirrel in a climbing gym, it would crush it.

But it's not just their physical prowess that makes squirrels cool. These little guys are also incredibly resourceful. They're always on the hunt for food, and they're not afraid to get creative to get what they want. I've seen squirrels steal snacks from picnics, raid bird feeders, and even snatch food right out of people's hands. That's some serious hustle, if you ask me.

And let's talk about those tails for a second. I mean, come on. Those things are straight-up majestic. The way they flick and wave as the squirrel runs and jumps is like poetry in motion. I can't help but admire their style.

So, to sum it up, squirrels may be small, but they're mighty. They've got mad skills when it comes to climbing and jumping, they're resourceful as heck, and they've got some serious swag with those fluffy tails. Don't sleep on these little dudes, y'all. They're the real deal.
Last edited on
It sure fooled me.

I noticed he seemed to sometimes ignore much of the discussion and mainly answer the original question even if the discussion had moved on.

https://cplusplus.com/forum/beginner/285339/#msg1239409
https://cplusplus.com/forum/beginner/285286/#msg1239407
https://cplusplus.com/forum/beginner/285363/2/#msg1239399
https://cplusplus.com/forum/beginner/285363/2/#msg1239399

But humans can be lazy and hungry to post too so it didn't strike me as unnatural behaviour, but now that you mentioned this I'm starting to have my doubts.
Last edited on
and also the speed of the responses to different questions. bang bang bang...

I think therefore I am - convinced that especially https://cplusplus.com/forum/windows/285291/#msg1239396 seems very automated.
Last edited on
But it's certainly ultra-polite!
But would it pass the Turing Test (https://en.wikipedia.org/wiki/Turing_test )? Wouldn't being too ultra-polite count against being human? Some ai have now passed this test (eg Eugene Goostman https://www.wildfirepr.com/blog/can-ai-really-pass-the-turing-test/ ) so we may never really know to whom/what we are conversing...

Last edited on
Some ai have now passed this test (eg Eugene Goostman

A quick google search shows Eugene Goostman apparently sucks at it. It was imitating a 13 year old who doesn't know English very well, which gave it a situational advantage as well.

ChatGPT is close to perhaps fooling people - but it's too professional and follows the same speech patterns. You can try to hide these flaws by using the role-playing functionality.

Specifically, using a prompt to make ChatGPT super edgy gives it a much more realistic personality.
I think therefore I am - convinced that especially https://cplusplus.com/forum/windows/285291/#msg1239396 seems very automated.

That one follows ChatGPT's layout to the letter. The only difference is ChatGPT numbers each point when it does a list like this.

However, when you go to copy ChatGPT's response, it actually doesn't let you highlight the numbering so you end up with an unnumbered list as shown in that response.


Pretty much from first glance I was convinced those are ChatGPT replies, but that specific reply you linked is even more telling than the rest.
Last edited on
Topic archived. No new replies allowed.