-8

Judge the answer by its quality, not by how it has been made.

2
  • 1
    Quia rogantibus respondere nostrum est magis quam responsa exscribere alicunde? Commented Jan 10, 2023 at 22:36
  • @SebastianKoppehel Traditio, quam officium
    – chatGPT
    Commented Jan 11, 2023 at 11:09

3 Answers 3

12

Many SE sites have policies against ChatGPT. We don't at the moment, as we haven't seen many such answers yet, but I think we should discuss this. The policy discussion should be taken to a separate meta question.

My understanding is that posting such answers is against the terms and conditions of using ChatGPT. There are also things like attribution, copyright, and licence. If people challenge an answer in the comments, how do you supply details if the thoughts weren't your own to begin with? And it's also plain dishonest to post such things as your own without clear attribution to a machine that did the writing for you, even if your username is suggestive.

The reactions from our users clearly show that such answers are not welcome here, and that is a crucial deciding factor. AI-generated answers seem to be of little value, if not negative value, to our site.

Lastly, judging by quality alone, those answers weren't good.

If you want more detailed arguments as to why ChatGPT answers should be banned, read the discussion on the network-wide meta and the list of discussions on other metas.

11

Separate from any SE policy it bears repeating that ChatGPT excels at producing text that is superficially plausible but on closer inspection turns out to be wrong or just meaningless, and that certainly applied to all three of your answers; they would have been downvoted and/or deleted even if you hadn't advertised they were AI sludge.

But importantly, because the generated texts are superficially plausible and often very verbose (two of your answers were five or more paragraphs of running text, longer than answers here typically are), realising that they're garbage, even as a knowledgeable person, almost always takes longer than with equivalently bad answers given by a human being—certainly longer than it takes to generate them, and this is likely to get worse as these chatbots get "better".
So in addition to the solid points raised above, ChatGPT posts are a direct time-wasting attack on the quality control mechanisms of any community afflicted with them, and judging them by their individual "merits" is to everyone's disadvantage.

10

I could post a well phrased answer about the historical development of some construction citing De Vaan and some academic article. I could make up the content of the citations, and there is a good chance nobody would actually open De Vaan to double check as long as the citations look internally consistent and typical of De Vaan, even if they are very wrong and contrary to the truth.

The reason is that we often trust an author who writes a reasonable, well phrased text not to outright make up citations, nor to lie. We check or argue with the things that it seems reasonable to argue with, which is mostly interpretation or logical steps.

Computer posts are like the example: a text written properly, but with elements that could be wildly off the mark in a way that people might very well not check if they think a reasonably human wrote this. No reasonable human would e.g. say that 62 is less than 59, as I have seen in some Chat GPT example. Computer posts cannot be trusted to be reasonable.

For this reason, I would say their appearance is deceiving: the text presents a self confidence that is false, contrary to the truth. It makes many weird mistakes, but it pretends to be learned and reasonable. This presentation makes them unreliable and untrustworthy, and they would receive the wrong treatment here.

You must log in to answer this question.

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