I feel like weโve been getting overwhelmed lately with AI-generated posts. Iโm not talking about using translation tools, but people just lazily opening up ChatGPT and asking it to โmake an OSM forum post proposing tag XYZ.โ We are seeing posts that are so obviously ChatGPT generated, down to the em-dashes, bold headers, corporate-speak, and overly feel-good calls to action. Some recent examples are here and here and here. Even the OSM foundation is working on a policy change to remove egregious AI content.
I propose that the reaction be added as a way to flag posts that are obviously AI-generated. This would serve the following purposes:
Flag to other community members that the post may include content that is misleading, misrepresented, or flat out wrong, because AI was used to created it.
Provide feedback to post authors that their post reeks of AI and that they are probably getting a weaker response than they would otherwise and/or annoying people that have to read excessively verbose content.
Provides a less confrontational way for community members to flag a post as AI. In the past I have directly called out AI posts, and as you can see, this provokes a very negative response.
I want to recognize that many in our community do not speak English or do not speak it natively. I embrace and recognize the use of AI for the purpose of machine translation. However, what we are seeing is more โplease generate a post in English that says such and suchโ rather than โplease translate the following post that I wrote to another languageโ. I think we should have the option to react negatively to the former user of AI while still encouraging/allowing the latter.
I am a fan of AI, and the things that it can do. I wrote this post with my own hand, with zero AI use. I hope that this proposal can help us reduce the amount of AI-generated garbage that we have to wade through on a regular basis.
In my opinion, there is no need for an additional emoji. I still stand by my answer and opinion in the popcorn thread: Either we allow all available emojis or, conversely, none at all.
If any (AI-generated) content violates the current (or future) community guidelines, we have a flag-icon to report the post to the moderator team. A robot emoji reaction cannot trigger a notification to the moderators and would therefore be ignored.
And if your concern above is valid, then the reactions to AI-generated content is already close to zero, so thereโs no need for an extra hint about it, right?
Does Discourseโs Reactions plugin let us offer all the possible emojis? If so, it isnโt documented anywhere. Thereโs a different plugin, Retort, that offers all emojis by default and allows you to leave multiple emojis on a single post, similar to Slack or Discord. But it might not be as mature as the Reactions plugin, and migrating to it might be messy.
The converse is Discourseโs default behavior, only a single emoji: . If we were to return to that, all of this communityโs expressions of non-heartfelt emotion would vanish faster than ChatGPT can say โCertainly!โ
True, though youโll love hunting for and in that massive list.
For completeness, the Discourse Chat plugin does offer every emoji by default with the same emoji picker that you see when composing a post, plus some shortcuts to the most frequently used emoji. (Donโt ask why this chat room uses so much.) I guess the Chat plugin wouldnโt really be a solution to the LLM-generated posts that @ZeLonewolf is concerned about, but itโs curious that the Reactions plugin doesnโt just reuse this UI. Maybe worth proposing on the Discourse forum.
It depends on whatโs posted, I guess, but some sorts of AI content would definitely go against this rule (see discussion of that here). Even where that line isnโt crossed, there has been a fairly clear pushback against AI content in this forum. My personal perspective is that it can be bad if it dilutes the original human message (and the most egregious uses of AI in these forums certainly do that).
I would like to ask the DWG to lock this post. This is obviously a joke proposal, and ZeLonewolf has a history of emoji proposals for trolling purposes.
Not that I disagree with this proposal, but I just want to bring up a concern I have:
Truth be told, I kinda just write like that sometimes? I love em dashes, I think using them captures the way I speak irl which is what I try to go for in my writing. I like using headers on occasion cuz I try to outline what Iโm getting at before I dive in, since I know Iโll make my point badly if I donโt think about it ahead of time. I work a corporate job so (unfortunately) corpo speak is just part of my regular vocabulary now. And I was taught in ESL early on to include what youโre asking for as part of your writing so that itโs clear what your intention is โ I think that could be read as a call to action.
What Iโm getting at is that I donโt think I would appreciate it or feel welcomed to contribute if I had people accusing me of using AI for writing in a style that comes naturally to me. This is a problem that the art community online is facing where people accuse them of using AI for works they created by hand. That sorta behaviour makes me less willing to share my artwork online, and I could imagine something similar for text.
I donโt think thatโs your intention with this proposal โ there are times where something is obviously AI. Iโm just hoping this doesnโt open the door for people to go after others in the grey area where itโs harder to tell.
Rest assured, your spaced en dashes distinguish your writing from the unspaced em dashes that ChatGPT prefers. Of course, both are perfectly valid choices. I used to be on Team Em Dash back in the day (also Team Two Spaces Between Sentences). But if a post has enough of these very subtle cues, people have become pretty good at telling when ChatGPT was used with the default settings, sort of like telling the difference between a human voice and a text-to-speech engine. And this automation alone is no problem from a moderation standpoint, as long as the post has substance and a bit of human character shines through.
(Personally, I take pride in writing differently than what these tools mass-produce. But I do wonder if eventually my writing style will become a liability as society becomes acclimated to LLM-speak and comes to expect it instead of something more bespoke.)
First off, if you are truly a non-native English speaker, congratulations โ your prose is flawless as far as I can tell and certainly doesnโt come off as AI.
Itโs the obvious over-the-top AI posts like this one that I am particularly trying to target. Now, I canโt predict how people will use it, but I highly doubt a legitimate human response would be perceived as AI. I realize that there is a whole controversial world of bad โAI detectionโ software running rampant in academia these days, but weโre talking about humans making human judgement.
I like that icon but frankly spoken I do not see any advantage in adding it to the reaction panel. If someone posts AI generated text it is fair enough to let them know that this is not appreciated in the forum, but this can better be done with a few polite words than by use of a bot icon. A commentless bot icon as reaction to a post is surely not
Having said that I can not see any reasonable Pros for such an reaction emoji, but several Cons:
The bot emoji certainly has the potential to offend other participants, specially those who did not use AI for a post but earned this reaction anyhow, like @Malle_Yeno just pointed out.
Additionally the bot emoji has a potential for misuse by being placed intentionally under posts which are most probably not AI generated just to wilfully annoy someone.
The introduction of โyourโ popcorn reaction (which I understand as absolutely harmless) caused a discussion with more than 200 posts, still going on and a second one with nearly 100 posts, still going on. This is such a waste of time that we should not trigger another very similar issue.
As AI is developing fast I have no doubt that in near future it will be much harder to tell if a post was created by AI or not. The more difficult it becomes to distinguish between AI and human generated, the bigger becomes the space for misplacing a bot emoji reaction and the less sense would such a reaction make.
Following the rule of acting in โgood faithโ we should assume that someone creating a post with AI did that in good faith and not to annoy others. In such cases it is surely better to let them know in a polite way instead of flagging the post with a bot icon (as already expressed above).
Iโd say this are reasons enough to downvote your proposal.