Quote from: yahoo62278 on Today at 01:02:39 AM

I don't feel like people are looking at the long game here as far as AI and this forum is concerned. Yes you can generate posts, yes you could build some sort of detector, yes it probably has some good things. The issue is what happens when this forum is just AI generated posts? All the real users dropping knowledge and helping bitcoin grow disappear. All the companies advertising see it as a spam haven with no real users. Forum starts slowly dying and all these people that were too lazy to make an actual contribution and only thought about making a few bucks will have to look for a new income source and 1 that likely pays a lot less.

Those who are doing all that don't give a f**k about this forum and the value it holds. A person generating his posts through an AI only to save a few minutes while earning money from the paid signature he's wearing barely cares that he's contributing towards ruining a place where Bitcoin was born or its creator(s) were once present, and that how valuable and memorable this place is. They just don't care.

Quote from: nutildah on Today at 02:07:23 AM

This is one of the biggest misnomers to be applied to any type of new technology. The idea that a machine attempting to emulate a human being can further "humanize" its output is absurd. The machine is doing all the work -- there is no human beings involved in "humanization." Thus a more proper term would be "shitifier" as it makes output more shitty albeit in a very robotic way.

You're right, but as a wild guess, I believe so many people use these tools even in this forum to bypass detectors and get marked as AI spammers.

Since a lot of experiments are happening these days, why not try it and see how it does? Roll Eyes

I'll generate a post using ChatGPT, run it through AI detectors, then humanize the text using the humanizer from Quillbot, and then run it through the detectors again, and then I'll add the outcome of both tests below. I'll be using your post (the one I'm responding to) to generate a response for this test.

Response generated through ChatGPT:

Quote

Thank you for sharing your viewpoint. While it is understandable to question the use of the term "humanize" in relation to machine-generated content, it is commonly used to describe efforts made to adjust AI output so it resembles natural human language more closely. This process does not involve actual human emotion or intent, but rather algorithmic adjustments to tone, structure, and phrasing.

Regarding forum moderation, your observation is valid. Identifying AI-generated content can be challenging, especially when users do not disclose their use of such tools. Therefore, it is logical that moderators prioritize content quality and relevance over the method by which the post was created. Posts deemed as spam, regardless of origin, are more likely to be subject to removal.

GPTZero: 100% AI-generated
Sapling.ai: 100% Fake
Copyleaks: 100% AI Content Found
Quillbot: 100% AI-generated

Humanized version using Quillbot's AI Humanizer:

Quote

We appreciate you offering your point of view. Although it is reasonable to doubt the use of the term "humanize" in reference to machine-generated content, it is often used to characterize attempts made to modify AI output so it more closely resembles natural human language. This process uses algorithmic changes to tone, structure, and phrasing rather than actual human emotion or intent.

About forum moderation, you make a fair point. Finding AI-generated content can be difficult, particularly in cases when users fail to reveal their use of such instruments. Consequently, it makes sense that moderators give content quality and relevance top priority over the way the post was produced. Posts judged as spam, from anywhere, are more likely to be deleted.

GPTZero: 8% AI-generated
Sapling.ai: 0% Fake
Copyleaks: No AI Content Found
Quillbot: 100% AI-generated

Surprisingly, the detector from the same source that was used to humanize the text has detected that it's AI-generated, and all other tools refused to accept that. However, this is just an experiment done quickly with no edits or anything like that, but imagine if someone spends a few minutes on each post, running it through two or more such tools, refining the response a bit more in each attempt, they could come up with totally undetectable outcomes.

It's true, these tools are only good at fooling detectors, and a human, reading the post, should be able to determine that it's written and refined by AI, but then, you can't take actions based on your assumptions or even knowledge, and you need proof to do that.