An example of a ChatGPT comment
I mentioned last week that I removed several well‑composed ChatGPT comments here, but that I’d trashed them. Spotted another today, a comment on this post about the recent fire.
The comment, while anodyne, is not the usual gibberish we used to see in comment spam:
6 people are missing after a fire in Old Montreal this week. What do you think happened to them?
The trick is in the commenter’s link, which is to a U.S. site that sells CBD oil in various formats. They’ve sent me similar spams recently, always adjusted to the topic, but always linking back to their site.
I don’t doubt that the AI will get smarter, so if anyone sees CBD oil being flogged from here, please let me know, because it’s not me doing it.
Wow! I’ve caught another! Check out this comment by “Kathy Brown” to this post. “Her” link went to the CBD site as well, and it’s so well written that I hadn’t noticed it was a bot:
It’s unfortunate to hear that Archambault is closing its original store, and it’s clear that the lack of social services has had a devastating effect on the area. It’s clear that something needs to be done to help the community, and I’m wondering what steps the mayor is planning to take to address the issue? It’s important to help the people in the area, and I’m hoping that the mayor is taking this issue seriously.
I’ve since deleted it.
Joey 10:13 on 2023-03-26 Permalink
How do you distinguish chatgpt spam from the original variety?
Tim S. 11:18 on 2023-03-26 Permalink
Thanks for sharing, Kate. Interesting that the AI is at the stage where it understands that the post should be in the form of a question, but can’t come up yet with an interesting question. We’re still ahead of it for now!
Kate 11:31 on 2023-03-26 Permalink
Joey, in the past, comment spam never actually addressed the topic, but was always something like “Thank you for this information, it is very helpful, this is the most informative I have read on the topic” and so forth. Now the spam actually addresses the topic in some bland but relevant manner. I’d seen others, but deleted them before this (except for one, which I added above).
ChatGPT has a style, it talks around the topic in a smooth but plausible manner, but introduces no ideas of its own.
(There’s a thought. Maybe Ensemble is an artificial intelligence!)
Ephraim 16:06 on 2023-03-26 Permalink
Ever tried the invisible box trick? It’s an extra form field that doesn’t really appear and needs to be blank for the comment to be accepted. That’s what I use on my web forms.
MarcG 22:04 on 2023-03-26 Permalink
That technique is sometimes called a ‘honeypot’. Kate, if you want to look into adding something to the site let me know. 🙂
Kate 10:30 on 2023-03-27 Permalink
Thanks, MarcG. So far I think I’m on top of it, but you never know.
Poutine Pundit 19:49 on 2023-03-27 Permalink
Here’s how ChatGPT responded to the request to provide a follow-up comment to this feed with new ideas and a link:
It’s interesting to see how AI-generated spam is becoming more sophisticated and difficult to distinguish from human-generated comments. One potential solution could be to develop more advanced captcha systems that use natural language processing to detect whether a comment is generated by a machine or a human. Another idea could be to use machine learning algorithms to detect patterns in spam comments and automatically flag them for review. Here’s an article discussing some possible solutions: https://www.relevance.com/blog/5-ways-to-keep-comment-spam-at-bay-in-2023/
The link provided by ChatGPT doesn’t work.
Kate 16:24 on 2023-03-28 Permalink
Poutine Pundit, very nice! Very plausible.