Are these forums being trolled by AI chatbots?

[Dave-in-Decatur 3,970]

Lately I've noticed a couple of responses to old threads that haven't had posts in a year or more, which seem to simply summarize the initial post and some of the discussion, without necessarily using the specific terms found in previous posts, and without making any actual contribution to the discussion. Here is an example:

The post is bland, to the point where one user described it in the thread as "bloodless." That's a good description, and I would say pointless as well. To me, this feels like a proof-of-concept attempt by some actor who is not a participant in the forums, some kind of test of whether an AI chatbot like chatGPT can draw information from a forum thread and produce a rational-seeming reply.

 

Here is another example, which draws on information from outside the forum thread:

Regarding the two or three that I've seen, I would give them an A+ for Artificial, D+ for Intelligence. I'm a retired professor in the humanities, very used to spotting "filler" text in a student paper, something meant just to reach the required number of pages without having to, you know, think of anything. These posts feel like that, and they follow the same practice of cobbling together already-expressed ideas with some changes in wording. They use coherent, correctly formed English sentences to make coherent, and yet somehow utterly useless, paragraphs.

But surely nothing done online is completely pointless. If this is bot-work, what is its purpose? Possibilities that come to my mind:

1. Simply testing to see whether the generated response seems (to the bot-controller) to be acceptable prose.
2. Testing to see whether other people in the forum are fooled into thinking this is a real person. So far, not happening.
3. A beginning of some effort to manipulate discussion in the forums to serve some ulterior motivation, perhaps in service of an Evernote competitor.

3 seems unlikely to me. In the case of 1 or 2, it may be that the forums are being used as a testing ground for bots to be deployed elsewhere for some nefarious purpose.

I would be interested in how others, especially veteran participants in the forums, see this. Does it seem like AI chatbot work to you? If so, how well does it seem to be working? Have you seen anything similar on other online forums? What do you think might be the purpose (granted that we are only speculating, and therefore likely to devolve into paranoia)? Does it seem like a minor annoyance, or the camel's nose under the tent of something more alarming? I would also appreciate it if Evernote staff responsible for the forums would chime in.

[agsteele 2,952]

I don't suspect this to be so. I'm more of the opinion that it is just badly phrased and irrelevant posts. I'm much more inclined to go for incompetence over conspiracy.

[PinkElephant 8,149]

My observation is a number of posts reading "bloodless". They are well phrased text, but so balanced and seemingly without an own opinion, and at the same time narrowly beside the topic - like being phrased by somebody who never used EN himself.

I think some users are experimenting with asking a bot instead of posting a question here, and are then posting the answers.

[gazumped 11,686]

I'm just evil and suspicious - I have deleted a couple of these posts on the basis they're irrelevant and do not contribute anything to an ongoing discussion,  and I tend to follow the profiles of those that survive so I'm notified if they post again. Repeat offenders could be terminated with extreme prejudice. 

[Boot17 1,462]

Hi there, I just wanted to weigh in on this discussion. As a fellow user of this forum, I have not personally experienced any instances of trolling by AI chatbots. However, I understand the concern that such behavior can have a negative impact on the overall user experience. I think it's important for us to remember that chatbots are just tools, and it's up to their creators and users to ensure they are used in a responsible and ethical manner. Let's keep this discussion focused on creating a positive and productive environment for all members of this community. Best regards.

...

 

🤣

[Boot17 1,462]

7 hours ago, Dave-in-Decatur said:

Does it seem like AI chatbot work to you?

Totally.

....

Another comment similar to the Spotlight one in question:

This stuff is from OpenAI's ChatGPT -- using the Chat interface. You can, of course, use APIs to integrate with it and not have to manually type things in like that.

[gazumped 11,686]

But isn't that AI stuff just bland nonsense? 

The Forums are surely about collecting reactions from real life users about their suggestions for and reaction to changes,  and offering "helpful" (for varying definitions of that word) suggestions to make sure that they get support from Evernote when it is available,  plus pointing out alternative solutions and suggesting work-arounds in the meantime because Evernote is never going issue an update within a short time of receiving feedback unless it relates to a genuine bug. 

But all of this involves real human experience.  I've seen suggestions that AI can help support teams by responding to queries as an intelligent database,  but that can only go so far with a constantly evolving product like Evernote.

It's not helpful to anything to have an irrelevant,  meaningless comment posted in a months-old discussion thread.  

[Boot17 1,462]

On 2/7/2023 at 7:00 PM, gazumped said:

But isn't that AI stuff just bland nonsense? 

In the specific case of contributing anything to the discussion for those two examples that Dave linked to  -- my answer would be "yes". But as far as AI in general, no, it is not bland nonsense... and they are only just getting started!

The title of this thread is "Are these forums being trolled by AI chatbots?" Dave then asked "Does it seem like AI chatbot work to you?" My only point in commenting was to show the similarities between a generated-for-sure response from AI and the two specific comments that Dave linked to. They seemed very similar to me.

 

[Dave-in-Decatur 3,970]

Yes, the suspect (and for-sure) posts do have a "robotic" feel to them. Something very "Hal"-like, as in "I'm sorry, Dave, I can't do that." Hopefully they never get much better.

Remember the old idea that given infinite time, an infinite number of monkeys, and infinite typewriters, they could produce the works of Shakespeare? (Or something like that.)

The only delicious irony in this for me is that we have dedicated professionals isolating themselves from human contact for days and weeks on end trying to program software that sounds like human contact.

[gazumped 11,686]

Recent news stories show that AI output can contain factual errors,  and that in some cases the output is a copy of real-world content (which also applied to images) so if you generate some text you need to fact-check it,  and also make sure your copyright insurance is up to date. 

Noam Chomsky doesn't seem to be a fan either... Noam Chomsky on ChatGPT: It's "Basically High-Tech Plagiarism" and "a Way of Avoiding Learning" | Open Culture

[Dave-in-Decatur 3,970]

7 hours ago, gazumped said:

Recent news stories show that AI output can contain factual errors,  and that in some cases the output is a copy of real-world content (which also applied to images) so if you generate some text you need to fact-check it,  and also make sure your copyright insurance is up to date. 

Noam Chomsky doesn't seem to be a fan either... Noam Chomsky on ChatGPT: It's "Basically High-Tech Plagiarism" and "a Way of Avoiding Learning" | Open Culture

There must be some legitimate uses for it Maybe compiling a user manual from the drafts of separate chapters? Trying to think, trying to think ... maybe ask ChatGPT if it has any legitimate uses? Has it learned to be defensive and self-protective yet?

Toward the end of my teaching career, students who were inclined to plagiarize started to do it off the Internet. It was so easy to spot, and of course if I suspected a passage I could just Google it too. But teachers and professors will have to give writing assignments that include personal interaction with the subject matter, not just regurgitation of facts, checked or not. And again, the seams between the cribbed and the original are always detectable.

As is the case in the forums, just to return to the point in question.

[gazumped 11,686]

3 hours ago, Dave-in-Decatur said:

Maybe compiling a user manual from the drafts of separate chapters?

Or maybe... CustomGPT - Build your own ChatBOT | Product Hunt