6

This question

Can "neither do I" apply to both disagreement with positive statement and agreement with negative statement?

asks whether the advice given by ChatGPT (we can use 'Neither do I' to express disagreement with a positive statement) is incorrect, as the questioner suspects.

I am not sure whether to deliver resounding down- and close-votes, and have confined myself to a comment. Is there a policy on AI grammar advice?

[Edit] The poster of this question has adduced ChatGPT's support (in a comment) for a mistaken (in my opinion) notion about definite articles.

Is it correct to use a definite article before writing in this sentence 'I am good at the writing'?

8
  • 5
    I’m not sure why we would treat bad advice from AI any differently than bad advice from other sources. We have an entire post about how you shouldn’t trust your grammar checker.
    – ColleenV
    Feb 6, 2023 at 12:16
  • 4
    @ColleenV: One huge difference is if you "ask" Grammarly to pronounce on something, it will always respond in the same way to the same input. ChatGPT won't, which means it's potentially a "moving target" that can never be pinned down. And there are an infinite number of ways ChatGPT can introduce bizarre errors that might or might not get noticed, of a type that simply couldn't occur with an actual "programmed, rule-based" app like Grammarly. The holy grail of AI researchers is to get a neural net to explain why it made some decision (currently that's impossible). Feb 7, 2023 at 3:15
  • 4
    @FumbleFingers But our task here (should we choose to accept it) is not to explain why the AI made then answer it did. It is simply to evaluate a single, quoted fixed response and pronounce it accurate or not. That the same question might get a different answer on a different occasion is, as I see it, totally irrelevant. Would you feel differently if all mention of the AI were edited out of the question? That could be done easily enough. IMO it would not change the question in any significant way. Feb 7, 2023 at 3:23
  • @DavidSiegel: Absolutely that's what I'm saying! If the question can be edited so as to have no mention of ChatGPT then just do it, and we're done! But that begs the question Why introduce ChatGPT in the first place? If ChatGPT is the only source of something someone wants to ask about, they should go elsewhere. If ChatGPT is one of multiple sources giving rise to the question, focus on those others, not on ChatGPT. Imho ChatGPT isn't just something to be "minimized" - it has no place here! Feb 7, 2023 at 3:35
  • 4
    @FumbleFingers The source of a learner's question shouldn't matter. If we substitute "my teacher" for chatGPT does it make any difference to the answer? I agree that the question could be edited a bit, but chatGPT is a popular toy, and I don't think the mere mention of it changes whether a question is on topic. If we can entertain questions about slogans on t-shirts, we can let people who are willing answer chatGPT derived questions.
    – ColleenV
    Feb 7, 2023 at 13:33
  • The question has been reopened by a mod other than myself, with the support of two non-mod users. The question log will show details to those with enough rep. One answer has already been posted. Feb 7, 2023 at 16:19
  • 2
    I agree with @ColleenV. I could also ask my old 8th grade English teacher a grammar question and get two different answers from her on two different days. (It definitely happened!) I don't see why that aspect of ChatGPT is significantly different. Feb 10, 2023 at 8:51
  • A new related discussion on the main Meta: Is it allowed on SE to include AI (ChatGPT) generated output in the question post? (disclaimer: I answered that)
    – Glorfindel Mod
    Feb 20, 2023 at 18:38

4 Answers 4

9

I am the one who asked the question mentioned in the question. Regarding whether my question should stay closed (because it employed a chatbot as a resource) or be reopened, here is my opinion: I posed the chatbot numerous questions and for most of them I can determine if they are reasonable or not, but for a few, I am unable to judge. For those I cannot evaluate, it is regarding English grammar, and the source is of little importance here. For example, perhaps a one-year-old says something an adult believes would be ungrammatical but the educated adult cannot explain why. What matters here is the adult lacks the English knowledge, not that it is the toddler who caused the issue.

3
  • 6
    Well argued and said. There was nothing with your question as long as stated your source, which you did, the question is a valid one.
    – Mari-Lou A
    Feb 7, 2023 at 1:45
  • 4
    The ban on many Stack Exchange sites was thought necessary because many users were posting answers that were provided by ChatGPT and copied verbatim. That is plagiarism but even worse many of these users were not verifying whether these codes/solutions etc. worked or not. They just care about the rep.
    – Mari-Lou A
    Feb 7, 2023 at 1:50
  • 1
    @Mari-LouA Some are claiming that StackExchange will be supplanted by ChatGPT, but I thought it is humans who make the ultimate decision and ChatGPT-like bots are simply a type of resource no matter how intelligent they appear. Feb 7, 2023 at 1:58
6

I don't see a valid reason to close this question. In fact, I am strongly tempted to reopen it right now.

As I understand it, this question was posted by a human user. That user put a grammar question to an AI, did not trust the answer, and asked here whether the AI was correct or not. Several regular posters added comments taking the view that the AI was wrong. None made this an answer.

We have had a number of questions that said, more or less, "my grammar check software said X was valid, or not to do Y". People have responded to these. Often to say that the software was incorrect, or at least insufficiently nuanced. Such posts have not been closed as off-topic.

If the linked post had cited the advice of a grammar checker, would it have been closed? I think not. If it had cited the advice of a human not very knowledgeable in English, would it have been closed? I am sure not. If it had cited a book of poor reputation, would it have been closed. Surely not. I don't see the current case as in any significant way.

Unless someone can, fairly shortly, post arguments here why such posts should be off-topic, I intend to reopen the question.

Regular poster FumbleFingers suggests that:

We generally reject "limited scope" questions .... By which I mean the sought answer usually applies to exactly and only the specific "example" presented in the question.

But as I see it the actual question here is how the phrase "neither do I" may be validly and naturally used in response to a question. That is not a "limited scope question". It could come up in many contexts. It is at least as broad as many questions that do get answers here. The question is not "how will this chatbot respond?". Indeed the bot/AI is really irrelevant here. The question is:

Should this phrase be used in valid English? Is the quoted view correct?

That is the question that has in fact been closed, and which I propose to reopen.

4

The question should stay open.

The underlying question is in the title. The asker also happened to include their source, as they should have.

The source happens to be a terrible one, but that's neither here nor there. Users regularly ask for confirmation about dubious things they've heard from unreliable sources like Grammarly, incompetent teachers, beginner-level friends, and learning material not written by native speakers. These are all valid questions which we answer, and we usually tack on our assessment of the source.

Our opinion of the reliability of the source of a statement in a question has no bearing on whether a question remains open.

1
  • 1
    I feel like we’ve had a past discussion here on meta about the source not mattering in terms of whether a question is on or off topic, but I can’t find it. The closest I found was this answer of mine which still explains my position on it even though it’s 5 years old ell.meta.stackexchange.com/a/3116/9161
    – ColleenV
    Feb 8, 2023 at 1:01
3

I voted to close the question because ELL isn't here to analyse the syntax of chatbots. Once four other users had agreed with me, it did get closed, but before that this is what I said to the OP in a comment...

...you shouldn't be treating ChatGPT as a "source" of English for learning purposes. You shouldn't even trust it to write syntactically valid English sentences in the first place, and you certainly shouldn't trust it when it starts explaining grammatical rules to you. It might be right (and it'll be getting more reliable day by day), but even ChatGPT (by far the "cleverest" bot around today) can make some ridiculously untrue assertions. More crucially, it can slip in critical, but non-obvious errors which might not always be noticed even by people who know the relevant "truth".

I didn't really engage with the specific usage being queried by the Neither do I question. I saw that the OP here in this meta question had posted a brief comment rejecting whatever ChatGPT supposedly claimed, so I just "upticked" that comment before casting my closevote as above.


We generally reject "limited scope" questions (name for a programming variable, audio transcription, lyrics interpretation,...). By which I mean the sought answer usually applies to exactly and only the specific "example" presented in the question.

If it's obvious that some question is only being posed because of something ChatGPT said, I think the only sane response is to summarily close it. That's not just if someone has found a situation where ChatGPT appears to be making false assertions - we should resolutely cull all questions asking for help understanding why ChatGPT says something.

Questions asking us to "explain" what ChatGPT says are potentially much more insidious (and invidious! :) than the similar ones we've always had prompted by "writing assistant" recommendations (MS-Word, Grammarly, etc.). At least with Grammarly, there's something "static" that you can pin down and analyze (if debugging other people's software is your bag! :)

With ChatGPT, even if you present it the exact same question / example / whatever, it'll give you potentially wildly different responses. It's bad enough we're the "dumping ground" for ELU's unwanted crap - ELL should not be thrown into the "black hole" of ChatGPT!


I may be a voice in the wilderness here, but I really do think if we allow ChatGPT so much as a toehold on ELL, it will eventually destroy the site. It's ironic that so far as I can see, this is likely to happen because one mod wants it re-opened. I could maybe accept if five high-rep users voted to re-open, but to let something as important as this go through on the say-so of a single mod seems a bit high-handed.

20
  • 1
    Note that there are currently at least two other users besides me who have formally voted to reopen. I haven't checked to see who they are. Note also tht the question did not, as I see it, violate any of the specifications for what makes a questioning off-topic. I have asked the other active mods to consider this issue. Feb 7, 2023 at 2:52
  • if it was any normal question, I wouldn't be bothered. But I at least think ChatGPT-related (howsoever related) questions are such a departure from what we're "evolved" for that the bar for "nodding it through" should be raised much higher. And in this context, I think Doesn't violate any existing policies is 100% irrelevant. Feb 7, 2023 at 3:08
  • 1
    I disagree. Using the output of ChatGPT as an answer is barred. But that in no way means that any question that is in some way related to ChatGPT should be barred, or even viewed with significantly increased scrutiny. Feb 7, 2023 at 3:18
  • 2
    I think you don't understand my objection. Personally I think answers generated by ChatGPT shouldn't necessarily pose much of a problem on ELL (though they may be disastrous on other SO sites). Every day we get garbage answers on ELL, but some of them superficially look convincing. I guess that's partly because some answers come from non-native speakers who sometimes have peculiar misconceptions and "knowledge gaps". So ELL is used to that kind of thing. Me, I like nothing better than to lay waste to a bad answer carelessly upvoted. But I'm not clearing up after ChatGPT! Feb 7, 2023 at 3:27
  • 2
    I think though that there is a difference between a question that is a waste of time to answer because it's mostly derived from AI generated gibberish and a question that is off-topic. If the AI stumbles upon something that is interesting to someone who wants to address it, I see no harm. Downvoting is the appropriate tool to use here, not close voting.
    – ColleenV
    Feb 7, 2023 at 13:44
  • 3
    @ColleenV: I suspect the vast majority of querents on ELL are only really asking here because they don't have easy access to native Anglophones to help them learn English. But in their minds interacting with ChatGPT is already a reasonable way to learn the language. If we allow any questions that originated in any way from chatbots, pretty soon about the only questions we'll get will be people disagreeing with or unable to understand something they got from a chatbot. Most of which might look like reasonable questions to native Anglophones, but they wouldn't be "real". Feb 7, 2023 at 14:45
  • 2
    Nonsense. All questions written by a human are "real". Not every question is within our scope, or a good quality question, or worth answering, but they all have the potential to be. ChatGPT is not new... it's been around for a while; it's just popular right now. The angst over it on SE is unwarranted. It is disruptive, like any new technology, but banning it is the wrong approach and will cause more trouble than it prevents. We should be learning how to deal with it and figuring out how it can be most useful.
    – ColleenV
    Feb 7, 2023 at 15:33
  • 1
    I doubt any of us are gonna change our minds, and I think the whole debate is probably academic anyway. Regardless of what I think, people will do what they want - if necessary concealing the relevance of chatbots to their queries. Unless TPTB unilaterally ban all chatbot-related posts (which I suspect they still might do, even though that's effectively an unenforceable policy), they will appear more and more in ELL (which I think is by far the most "vulnerable" SO subsite, for the reasons given). We'll just have to look at the results in 5-10 years time. Feb 7, 2023 at 16:49
  • 1
    We've already addressed questions derived from AI generated content... Much of the ESL content online is generated; the only difference here is that the author of the question knew they were getting AI gibberish and was transparent about it. I wish there were a way to persuade you that this isn't an existential crisis for ELL. The current lack of engagement from the community is going to kill us way before AI ;)
    – ColleenV
    Feb 7, 2023 at 17:04
  • 1
    @ColleenV: Indeed. I don't have easy access to the numbers, but I have the impression ELL visitor numbers have tanked in recent years, and ELU increasingly gets used by people with very limited knowledge of English. I suspect as time goes by both sites get increasingly dominated by long-term high-rep users who are perhaps sometimes unhelpfully dismissive of "same ole same ole" queries (I don't excuse myself on that front! :) Still, at least you seem to have "come back into the fold" lately. Feb 7, 2023 at 17:45
  • 1
    @FumbleFingers I'm not as engaged as I used to be. I just lurk mostly now. I almost re-disappeared after my last interaction on the main SE Meta, where apparently something in the water makes decent people turn hostile and toxic over the stupidest stuff. I decided to just avoid that part of the network and hang out where the most of the community is too polite and nice to let on that they think I'm a jerk lol.
    – ColleenV
    Feb 7, 2023 at 17:52
  • 1
    Questions along the lines of "X told me Y. Is Y true?" are on-topic regardless of who or what X is so long as "Is Y true?" is an on-topic question on its own. All the questions about Grammarly failures, for instance, are on-topic.
    – gotube Mod
    Feb 7, 2023 at 21:39
  • 1
    ChatGPT itself is new; but NLP/NLG technology has been around a while. How do you think Siri et al work? The reason it’s exploding is largely because they opened it up to everyone. It’s a tool and it’s not any scarier or more disruptive than other impactful technologies have been. The real power lies in training your own models instead of playing with the general one. It could be an amazing resource for learners. I think back to the conversations I had with Eliza and wonder how well chatGPT could imitate a fluent speaker to converse with in any language.
    – ColleenV
    Feb 8, 2023 at 0:55
  • 1
    @ColleenV: I earned my living writing high performance software - including things like crossword and other word puzzle generators - so I'm pretty clued up on such things. I'm sure that "opening it up to everyone" has sped up development, but the sheer amount of processing power (and the billions put up by Microsoft et al) that went into training ChatGPT is mind-boggling. Google will be following suit within months, and we're well into the decade of neural networks now. Feb 8, 2023 at 1:40
  • 1
    The cryptominers drove a lot of the GPU technology that’s being used to innovate in the AI realm now that the energy to mine crypto is more expensive that the currency you can mine with it. My husband is getting into AI assisted art, so I know more about that side than the language side. People are doing really cool stuff with it like Stelfie the Time Traveler. And a lot of others are generating crap faster than ever. It will sort itself out eventually. It’s definitely reached critical mass though.
    – ColleenV
    Feb 8, 2023 at 2:49

You must log in to answer this question.

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