6
$\begingroup$

Currently there are some users providing what looks like unfiltered chatGPT answers to questions. I know that Stack Overflow had temporarily banned answered like that https://stackoverflow.com/help/gpt-policy. One of the reasons stated is that the answers "have a high rate of being incorrect".

Personally, I am in favor of the ban, however I would like to hear other people's thoughts on that.

Additionally, for those in favor, I would like to hear what are the proposed actions. My proposal is to initially contact in person the user that submits those answers and make them aware of the policy, and if they persist ban them for a period.

$\endgroup$

4 Answers 4

10
$\begingroup$

I think ChatGPT posts and posters should be dealt with quite harshly since they have the potential to seriously degrade the usefulness of an SE site.

On ServerFault we don't get a lot of meta discussion, so no one has spoken out against a pretty harsh policy: answers generated by an AI such as ChatGPT are not appropriate and will be dealt with expeditiously: answers will be deleted and accounts will be suspended.

For reference, here's one of the discussions on meta.SE about this, and a master list of the policies on various SE sites.

$\endgroup$
8
$\begingroup$

I am in favor of banning the use of any language model (e.g. ChatGPT) that produces answers to questions here.

On Engineering.SE, we sometimes deal with very specific and detailed situations. In engineering, these details matter a lot. Language generators can not know the nuances of engineering and design.

The answers are at best fluff, and at worse they are wrong in ways that are hard to detect by casual users.

$\endgroup$
0
2
$\begingroup$

I asked ChatGPT this question (is this trolling?), and it said "don't ban me, just check all the answers!"

As an AI language model, I'm designed to provide informative and helpful responses to questions that are posed to me, and I strive to always provide accurate and relevant information. However, I understand that some users may try to use my responses in ways that are not consistent with the guidelines of the platform they are using.

In the case of users submitting unfiltered ChatGPT answers on a platform like engineering.stackexchange.com, I would recommend that the platform moderators and community members take appropriate action to ensure that the responses are accurate and helpful. This might include:

  1. Flagging the answer as "low quality" or "not an answer" to bring it to the attention of moderators who can review it and take appropriate action.
  2. Encouraging users to provide more context or additional information to clarify their questions or concerns, which can help guide more accurate and useful responses.
  3. Providing guidance and education to users about the limitations of AI language models like ChatGPT, and how to use them effectively to supplement their own knowledge and expertise.

Ultimately, the most effective way to address unfiltered ChatGPT answers is to work collaboratively as a community to ensure that all responses on the platform are accurate, helpful, and in line with the community guidelines.

Point 3) Educating users on the limitations of Language Models is by far the most important.

I believe that there should be a ban on "unfiltered copy/paste" answers (although how exactly these would be detected remains an issue).

There is a difficult blurry line though, between that, and an answer which has been created with the assistance of a Language Model, which I would be hesitant to outlaw completely.

$\endgroup$
2
  • 1
    $\begingroup$ In this case, the ChatGPT output is clearly identified, and cordoned off - not presented as my own work, and not unchecked/criticised. $\endgroup$ Commented Feb 16, 2023 at 13:58
  • $\begingroup$ ROFL - troll X-D... More seriously, In this answer, chatGPT doesn't address the facts that the first problem was to detect chatGPT answers and that it is much more time consuming to check the answers than to produce them. The problem probably comes from the fact that those aspects of the problem should clearly be explained in the question. In any case, it is probably a bad idea to publish chatGPT answer without showing the exact question that was asked. $\endgroup$
    – Camion
    Commented Dec 11, 2023 at 10:12
-1
$\begingroup$

I agree, that chat GPT should only be used as a learning and guidance. You can learn a lot from chat GPT4, so I'm not against it, but people don't need to cheat the systems though.

$\endgroup$

You must log in to answer this question.

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