Timeline for GPT on the platform: Data, actions, and outcomes
Current License: CC BY-SA 4.0
16 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jun 9, 2023 at 18:23 | comment | added | NoDataDumpNoContribution | What a thoughtful and nuanced answer. I especially like the term hypernerd. I think it would be worthwhile to study more how questions and answers fared, i.e. take the score into account. Maybe voting has gone down too and that demotivates the nerds or the hypernerds. | |
| Jun 9, 2023 at 0:07 | history | rollback | kaya3 |
Rollback to Revision 5
|
|
| Jun 8, 2023 at 19:21 | history | edited | This_is_NOT_a_forum | CC BY-SA 4.0 |
Active reading [<https://en.wiktionary.org/wiki/cherry_picking#Verb>]. Introduced the abbr. "LFT".
|
| Jun 8, 2023 at 17:40 | comment | added | starball | @kaya3 hehe I wrote that other answer :P (so yes, I'm very much in agreement that question availability is going down rapidly and that the "POV" of the data Phillipe presented about that is heavily warped). I wasn't making an argument about quality- more about increasing complexity / difficulty. | |
| Jun 8, 2023 at 17:03 | history | edited | kaya3 | CC BY-SA 4.0 |
added 717 characters in body
|
| Jun 8, 2023 at 15:57 | history | edited | kaya3 | CC BY-SA 4.0 |
added 818 characters in body
|
| Jun 8, 2023 at 15:49 | comment | added | kaya3 | The issue of "questions per answerer" is discussed in more depth in this other answer. | |
| Jun 8, 2023 at 15:21 | comment | added | kaya3 | Your argument is also missing a step ─ you're saying that ChatGPT causes question quality to go up, and this should cause better retention of hypernerds. But has question quality actually gone up? As I said, in my judgement it has gone down; I'm willing to consider data which shows the opposite, but not hypothetical arguments which reach a conclusion contrary to my observations. | |
| Jun 8, 2023 at 15:12 | comment | added | kaya3 | @starball The numerical observation isn't related to question quality; the ratio of questions per hypernerd will automatically go up if hypernerds are more likely to leave when there aren't enough questions, so the "number of questions per hypernerd" being high doesn't mean there are available questions ─ Phillipe's use of the word "available" here is a statistical fallacy. We can't conclude from this statistic that there are "enough" questions for answerers to stay around for; the decline in the quantity of questions can't be ruled out as a factor despite Phillipe trying to do so. | |
| Jun 8, 2023 at 7:31 | comment | added | starball | I'm not so sure about the "hypernerd" explanation. If anything, I'd think that they'd be the ones to more proportionately find reason to stay / write more answers: I'd expect that if question askers are flocking to ChatGPT, the only reason they'd come back is if they get something that ChatGPT can't help them with- probably something harder- something that a "regular nerd" would be proportionately less likely to be able to answer. | |
| Jun 8, 2023 at 2:12 | comment | added | kaya3 | @blackgreenonstrike Interesting idea, but it's also possible that both are true. The kinds of question that might no longer be appearing on SO due to askers trying ChatGPT first and being satisfied with the answer they get, include both the questions that are written well enough to be definitely answerable, and the megadupe questions that ChatGPT has a lot of training data for; and both of those would appeal to different kinds of very-active users. I think I'd need some evidence before I accepted the premise that there are fewer megadupe questions being asked on SO now, though. | |
| Jun 8, 2023 at 1:45 | comment | added | blackgreen | we should expect that the AI plagiarists' behaviour changed to try to avoid getting caught — and this is the key point. It's entirely reasonable to expect the "drafts saved" metric to decrease in accuracy over time, i.e. to have a dropping recall, as people adapt to the ban. | |
| Jun 8, 2023 at 1:41 | comment | added | blackgreen | Has the quality of the questions gone down? — I think here could actually be the other way around. All the easiest mega-dupe questions are now being asked to ChatGPT, as it will answer anything right away vs. crafting a question here that has any hope to stay open. So fast-gun and low-effort answerers have less questions they can answer quickly and serially. Not necessarily implying that question quality has gone up: just that there are less questions that many people can answer fast. | |
| Jun 8, 2023 at 0:40 | history | edited | kaya3 | CC BY-SA 4.0 |
edited body
|
| Jun 8, 2023 at 0:33 | history | edited | kaya3 | CC BY-SA 4.0 |
edited body
|
| Jun 7, 2023 at 23:55 | history | answered | kaya3 | CC BY-SA 4.0 |