8
$\begingroup$

We will see more and more questions that involve ChatGPT, in two ways:

  1. Questions about the inner workings of ChatGPT and similar Large Language Models (LLMs)
  2. Questions about "advice" dispensed by ChatGPT - "ChatGPT told me X, can you explain this to me?"

We currently have a tag , which is apparently mostly about case (1) above, but it starts to get used for case (2), e.g., in this question.

I propose we create a tag for questions about advice given by ChatGPT, clarify the distinction in the tag wiki, and start the thankless and likely perennial task of cleaning up incorrect tag usage.

Thoughts?

$\endgroup$
2
  • 10
    $\begingroup$ Perhaps rather create tags "llms" or "large-language-models" and "llms-advice" as above, and synonymize the ChatGPT tags to these - because the next incarnation of LLMs will yield the same challenge, but not be called "ChatGPT" any more. $\endgroup$ Mar 14 at 14:20
  • 1
    $\begingroup$ I concur with this proposal. Without explicit demarcation, genuine questions asking about LLMs would be interspersed with those questions under the advice of ChatGPT; the latter should be dealt differently. $\endgroup$ Mar 14 at 14:23

2 Answers 2

9
$\begingroup$

It would make sense for 'chatgpt' to be a synonym of 'large-language-models', as it is rather too specific. The way that would work is that 'chatgpt' would appear in the list when you look for a tag to add to a question, but what gets added is 'large-language-models'.

I can't say I favour the 'chatgpt-advice' tag—why not have tags for any other source of (mis)information? who on earth would want to follow them? It would seem its sole utility would consist in heading off some questions from being wrongly lumped together with questions about L.L.M.s, & it might not even be terribly effective at that. (Furthermore, its existence might be taken as encouragement to use ChatGPT.)


† And I don't think that ought to be encouraged. Human error, as found in textbooks, blogs, lectures, &c., may be instructive—offering a chance to dispel common misconceptions or suchlike—, but the same can't be said of ChatGPT's confabulations & non sequiturs. I don't think there will be many cases where ChatGPT's "advice" is worthy of critique.

$\endgroup$
5
  • 1
    $\begingroup$ I agree that ChatGPT is "hype" as mentioned in a comment to another answer and won't often be "worthy of critique." The nightmare scenario = it will probably be perceived as a superior authority because of the hype & the label of "artificial intelligence", particularly by those seeking easy answers to tough questions in many fields, not just stats. The whole StackExchange network, academia & beyond ought to brace itself for a fresh flood of misinformation, poorly written code & challenges to expert credentials from this toy. At some point, experts may be forced to label it as just that: a toy. $\endgroup$ Mar 22 at 15:50
  • 1
    $\begingroup$ I think this makes the most sense. I edited the chatgpt tag to include a warning against using it for questions about advice given by ChatGPT (as well as against such advice in general): stats.stackexchange.com/tags/chatgpt/info I also created an LLM tag (stats.stackexchange.com/tags/llm/info). I can't propose chatgpt and gpt as synonyms of it, because I don't have the rep yet. I don't know whether anyone has, and whether chatgpt and gpt should by synonymized... perhaps you as a mod want to do this manually? $\endgroup$ Mar 24 at 6:41
  • $\begingroup$ @StephanKolassa: Thanks! I can do that - was wondering though what the distinction is between the 'llm' & 'language-models' tags. Should one of those perhaps be a synonym of the other, or should we edit the tag wikis to clarify? $\endgroup$ Mar 27 at 6:17
  • 1
    $\begingroup$ It looks to me like we should synonymize "llm" and "language-models". Sorry, I didn't even think to check whether there already was something like that out there... But then, I'm absolutely not an expert on either kind of language model, so perhaps an actual expert could chime in? $\endgroup$ Mar 27 at 6:45
  • $\begingroup$ I don't know if there's value in maintaining the more general (?) 'language-models' tag & the more specific (?) 'llm' tag. I'm no expert either, but imagine there might be little & middle-sized language models too. $\endgroup$ Mar 27 at 6:48
4
$\begingroup$

This could be generalized to other sources beyond chatgpt, for instance

"Y told me X,"

with Y being chatGPT, quora, reddit, google, online course etc.


Chatgpt is just a current instance of a language learning model or 'online oracle'. In the future we may have additional names from chatGPT rivals like Google's bard, and who knows Amazon is gonna come up with a third version as well.

A more generic name might be welcome. Potentially we can keep a tag chatGPT for the moment as an alias for the more general term.

Aside from this, it may not seem like GPT is gonna be used as an important resource for statistics advice. Tagging potential questions that have specifically chatGPT as a source might be a bit exaggerated and is giving this source of information more attention in the tags than needed. Rather than making a separate tag, to lure away users from wrongly using a tag, it might be better to discourage the use of the currently existing chatGPT tag for such problems.


The same is true for the first description of the chatgpt tag. This is more like a language learning topic. The chatGPT is just currently a hype but doesn't seem to require a special tag (except as an alias to help people find the correct language learning tag)

$\endgroup$
3
  • 2
    $\begingroup$ I agree that ChatGPT is hype, but I'd say let's create a tag, at least so that the questions are easier to find and we are able to better track how many are there, easier find duplicates, etc. $\endgroup$
    – Tim Mod
    Mar 14 at 15:34
  • 3
    $\begingroup$ @Tim I agree with creating the tag, but we need to keep in mind that the tag will eventually need to disappear and merge with other tags or just function as an alias for a more general tag. I imagine that it is useful to keep this in mind now already; to prevent a potential divergence from the greater picture. $\endgroup$ Mar 14 at 15:37
  • 3
    $\begingroup$ Yep, it'll disappear, hopefully with ChatGPT itself. $\endgroup$
    – Tim Mod
    Mar 14 at 15:41

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .