« first day (4599 days earlier)      last day (499 days later) » 

9:13 AM
Just checking how is the AI-generated content situation on our site? There's this thing on SE meta.stackexchange.com/questions/389617/…
@user202729 We've seen a few odd answers, but those were clearly rubbish by any objective standard: the fact they were AI-related wasn't the major issue in that sense
9:30 AM
@JosephWright well the AI-related problem was (and it) that it's as Knuth put it "all expressed in marvelous English prose" and so the rubbish sounded very convincing.
3
9:53 AM
@UlrikeFischer :)
@UlrikeFischer I meant more that we've not had a major issue, and that I'm therefore not as worried as I suspect some other mods are about this
@UlrikeFischer In other news, I see there are 75 people down for TUG'23 - pretty good going!
10:23 AM
@DavidCarlisle, I upvoted your wrapfig answer. At least your left margins align... :P
@mickep all you need is a box, a tcolorbox is a dozen nested boxes and a million lines of tikz, but definitely enough for a green tick:-)
2
@DavidCarlisle It seems so. Interesting.
Seems to be the wrapfig day.
 
8 hours later…
6:16 PM
@JosephWright I was reading about the moderator strike on meta.stackoverflow.com/questions/424984/…, I was wondering what the stance of TeX.SE mods is in this?
6:36 PM
TLC3 has arrived!!!
3
6:52 PM
@Marijn First I know of it
@Marijn I'm not sure many people would notice a strike on TeX-sx
@JosephWright ;-))
ah ok, it seemed somewhat SO oriented but several mods of other sites have also joined I understand
@Marijn The main site (SO) is very different from the others
I know, I am also active there and I read both Meta.SO and Meta.SE regularly
@JosephWright Do you (TeX.SX mods) actually apply any of the automated AI detection tools whose ban caused the strike?
6:55 PM
but in general the actions/attitude of SO the company of course also affect our site, not just Stack Overflow
@TeXnician No, anything automated can only be done by the Powers
@JosephWright TeXnician means off-site automatic ChatGPT detectors, which are used on SO by moderators as supplementary information when deciding to delete posts or suspend users
@Marijn Well yes, but I think there's been a complete misunderstanding of the policy. My feeling is the Powers are worried that 'AI detection' is coming out as discriminatory against non-native English speakers, and that is a real business issue (think being sued)
@Marijn Exactly. I meant the tools as mentioned in the open letter.
@Marijn I suspect that was already a violation of the mod agreement, TBH
7:02 PM
@JosephWright in the discussion it was mentioned that mods already use automatic tools and heuristics for things like spam detection or other low quality content, so if a ChatGPT detector is against the mod agreement then any of those things would be against the agreement as well
@Marijn Might be that I have quite a narrow reading of the agreement: as we don't need to worry about large numbers of problem posts, it's always been easy to err on the side of caution
also the argument was that those moderators never decide based on the output of such a tool alone, but also take into account for example the number of answers posted in a short time, a comparison with the writing in previous answers by such users, structural/rhetorical clues, and close inspection of the answers which often reveals that the answers are basically nonsense
anyway, I'm glad we don't have such problems on our site, but I am a bit worried/displeased about the approach of SE staff in this current issue and various others over the past couple of years
@Marijn All the other stuff is what the Powers are saying mods should be using ...
@Marijn Oh, there's a reason we've called them the Powers from day 1
@Marijn I suspect at least one thing, that perhaps the Powers haven't expressed, is that mods are supposed to be there for where automation fails: if stuff can be picked up automatically, that really should be down to the Powers alone
sure, but the mods don't automate the process, they just use automation to streamline their manual decisions
@Marijn AI can either become a threat (others like ChatGPT are harvesting SO/SE content and taking over, visits go down, we already see that) or a benefit if SO implements its own AI better and close to the user-crafted content.
7:15 PM
@Marijn I can see that too - in the same way we can see e.g. the IPs (clearly automated)
@StefanKottwitz I think AI can definitely be an advantage (I'm an AI researcher myself, so I'm a but biased...) and SE is also trying to use it in this way, for example the experiment they are currently running with suggesting a good title after you write a post, which I think is a good idea.
However, if people post large amounts of unchecked non-working/besides the point answers to any SE site I think this should be moderated very strictly, and if the company restricts such moderation in any way then that is a bad thing.
@Marijn I see that point, we or some sites may see a massive amount of generated answers, and that doesn't fit to our approach of carefully checking for duplicates, proper tags, being on-topic, quality, being up-to-date, best practice, etc., so I understand the other mods
@Marijn Like I said, my understanding is that the issue is that automated tools are getting it wrong too often; that doesn't mean of course that imposing something is a good plan, and some data would be good
8:18 PM
@Marijn What you have to remember with SO is that ehy have 100s of flags per day and need to do them super-fast, whereas we are used to ... a more leisurely pace (currently there are 6 flags, I might look again at them later)
@Marijn You have prompted me to pop into the mod lounge to see what's happening :)
I hope that the answer to this will be short:
I am trying to change the way chapters appear in the toc of a book doc class.
I am trying to do `renewcommand` to `\l@chapter`. I make some changes to the various dimensions that already appear in the command but nothing seems to change. I also tried to remove the `\bfseries` that appears there and still the chapter heading is in bold face just like in the default case. What am I missing?
I made some changes to `l@section` and `l@subsection` and it is pretty straightforward. Only chapter (and part, with `\l@part@) are not easy for some reason.
@tush Problem solved. I haven't seen any change since bidi is also in use and it modifies those commands, so I need to use \AddToHook{package/bidi/after}{...
9:15 PM
@JosephWright -- "6 flags", sounds like an amusement park. I almost always prefer to flag "thank you" non-answers rather than down-voting, but I always contemplate whether I should bother the mods. I think I'll stop worrying.
@barbarabeeton Oh, those are always easy :)

« first day (4599 days earlier)      last day (499 days later) »