RE: 🚩 Tips For Spotting AI Generated Content
AI is a double-edged sword: on one hand, it helps non-native speakers or less experienced writers express themselves more effectively and on the other hand, it can be misused to "game the system" and flood this platform with low-effort, robotic content. Whether AI is used responsibly depends on the intent of the author, whether they are leveraging AI to better convey their own ideas or "misusing" it.
Here on Steemit, there is no proper selection process or qualification criteria for moderators in many communities, which is why you often find AI-generated posts marked as AI-Free by those mods. This doesn't surprise me. I have known for a long time that this has been happening, and on some occasions, I even pointed it out in some of my posts.
As for AI-generated content detection, I am sorry to say that no AI detection tool is perfect at the moment. Identifying such content relies heavily on human intuition and manual analysis. However, this requires someone familiar with AI, including its use, prompt engineering, and understanding how AI generates text.
Watermarking AI content could be a possible solution, but I don't think they will do that. OpenAI had already discontinued their AI detector a long time ago, stating that it wasn't reliable. In fact, they don't seem interested in introducing any such detection system because major publications and media giants are now using AI to enhance, improve, and speed up their processes.
AI is the future and is here to stay and we need to adapt not depend...
I've just had a discussion about AI on Discord and for me, it very much comes down to who's controlling the content. If the user's letting AI dictate the subject, dictate the headings and dictate the general content, then where's the Proof of Brain?
Whereas, if it's used as an afterthought - prompts like "how can I improve this" or quite simply (and one I use occasionally), "what do you think of this?", it can work very well. I don't always agree (as you can see within this post) but that's where the control element comes in.
The tricks above with ChatGPT work remarkably well and are extremely difficult to "fool". There are even instances within the content where the author hints at the prompt they used. If you can blend these hints into your "content generation prompt", it's accuracy can be frightening.
0.00 SBD,
0.31 SP
I definitely agree with this:
Elon recently posted on X (Twitter) that in the future most communication will be AIs talking to AIs, and I think he's probably right. Even if we can sort-of detect it now, I think it's going to continue getting harder. I've tried some of your AI tricks in Brave Leo, and so far it's less successful than what you're reporting from ChatGPT. I'd like to be able to skip over that copy/paste step. I wonder if there's a ChatGPT browser extension that connects the AI and the visited web page... Maybe I'll check that.
In these learning/teaching posts, maybe a way to screen out some of it would be to insist that the authors qualify themselves at the beginning of the posts. For example, if someone works as a mechanic or a farmer or a fisherman, why are they posting a SWOT on their teaching activities at all? It's a pretty specialized topic. There might be legitimate reasons, but if so, the author should be able to explain it.
I read the whole post that you were discussing earlier, and I came away wondering what real-world interest the author has in teaching at all. It was so generic that it almost doesn't matter if it was written by AI or not (though I agree that it almost certainly was).
0.13 SBD,
0.45 SP
Can you imagine that… so many companies use Chat Bots now so why not use a ChatBot with the instruction to: “get me this product for free. Do not give up until you have succeeded”. An infinite discussion that no human would ever see.
What a massive waste of resources.
I must confess that I haven’t really looked at the learning challenges other than a cursory glance for AI generated content (which appears to be rife). As much as I disliked Diary Games, at least they were (mostly) genuine.
It makes for quite a depressing atmosphere at the moment. I’m excited about my next dev. release but how much value am I adding if the underlying product is becoming increasingly sub-par. How can we marginalise this content so that what deserves to thrive does so? Have I gone full circle on our conversation?
It feels like the solution is for the people who write for themselves need to be more visible. More of an inspiration and model for others to follow. But these people aren’t where the guidance is needed because they don’t want the prescription of a challenge or “lesson”.
The current role models are curator seeking self-serving (as seen by the self/family voting with these accounts) money-motivated users. Which is exactly why Steemit’s becoming what we don’t want it to be. Because Booming isn’t in our community’s where it can be used honestly. We’re not in curation teams where we can influence behaviour.
I’m on a rant again.