16

I was a bit surprised by CEO Prashanth Chandrasekar saying:

... if somebody goes into the AI tools and asks a question to either ChatGPT or into, uh, Gemini and the AI is unable to answer the question, the mechan... the relationship, the partnership we've created allows the user to post that question onto Stack Overflow directly from there ... (timestamp)
Prashanth Chandrasekar, Embracing Change in the Era of GenAI, MIT Tech Review Future Compute 2024. (Conference date: May 21, 2024)

which seems like it would violate, or come close to violating, Stack Overflow's AI Policy:

Generative artificial intelligence (a.k.a. GPT, LLM, generative AI, genAI) tools may not be used to generate content for Stack Overflow.

Since I expect whatever happens at Stack Overflow, will likely affect Stack Exchange as a whole, ...

Question: Is it (going to be) possible to post questions directly from ChatGPT or Gemini?

I don't believe this feature exists yet (right?). This makes me worry there's going to be a staff post beginning with the word "announcing" and containing the word "community" at least 3 times. Maybe this is connected to the revival of the Staging Ground.

Improve this question
11
  • 6
    I mean, hypothetically, and I've no special knowledge here, auto-posting the prompt into a question might potentially be acceptable. Its going to make for a very bad question, in my experience, cause the sort of questions I ask on SE confuses LLMs badly.
    – Journeyman Geek
    Commented Jul 21 at 13:47
  • 8
    Policy for SE is like a pair of socks to the average person. That person replace socks every day, and not paying much attention to the socks they wear that day. (So it's no surprise at all that the CEO is promoting actions against the company policy)
    – Shadow Wizard
    Commented Jul 21 at 13:48
  • Or we get into a cycle of people getting suspended for breaking site policies via site mechanisms built in by design. The genAI polices are community led for most part
    – Journeyman Geek
    Commented Jul 21 at 13:51
  • 4
    This has been announced for quite a while. The VS Code extension was supposed to get something similar, but it doesn't appear to have happened
    – Zoe - Save the data dump
    Commented Jul 21 at 13:53
  • 15
    Maybe it's just my naivety have not realized this before, but there are two SOs. One that we experience on community daily where questions are opened, closed, edited, reopened and answered. And there's another SO in the company executives heads where there are flying cars, liquid metal cyborgs, Skynet, space folds, parallel universes, perpetual motion machine,...
    – Augusto Vasques
    Commented Jul 21 at 14:47
  • 2
    Also: "because it contradicts Stack Overflow's AI Policy" - the implication is that they [SE] either don't care and plan to do it without doing anything about the rules (i.e. bad UX for anyone using the tools), it's something they [SE] are saying to (potential) shareholders with no plan to actually go through with it, or they [SE] plan to once again forcibly unban genAI content. In either case, if they [SE] do go through with it, it's not going to be received well
    – Zoe - Save the data dump
    Commented Jul 21 at 15:18
  • 5
    Like I've already warned SE (mod-only link) in the context of the data dump restrictions, pushing genAI to main and/or once again forcibly unbanning genAI content will trigger mass-resignations. I know several SO mods, myself included, who plan to resign if they ever go through with something like this. (Obligatory note that mods currently aren't aware of any plans to push genAI to main; this is based on conversations when stuff like this comes up, as well as remnants of the strike from last year)
    – Zoe - Save the data dump
    Commented Jul 21 at 15:21
  • 17
    Your statements in this question appear predicated on a failure to read what was said. What was said was that a user asks the AI. The AI fails to answer or answer correctly. The user can then post the question on Stack Overflow. While this overall description gives us very little in detail, your assumption that it violates the AI-Generated Content Policy is not justified based on those statements. Those statements make no indication that the AI is, in any way, generating content for the question, just that it failed to provide a good answer.
    – Makyen
    Commented Jul 21 at 17:16
  • 12
    That doesn't mean that we're all going to be happy with what actually happens. It's just means that your assumptions are not justified based on the statements made by the CEO, or at least the ones you've quoted. Further, those assumptions are very unhelpful, as they feed into other people's already high level of distrust of the company. I'm not saying everything is fine, nor that this development doesn't need to be monitored, nor that we shouldn't have quality concerns if people are just posting the same thing they are asking an AI (we should have such concerns).
    – Makyen
    Commented Jul 21 at 17:22
  • I suppose you're right, the quote doesn't logically imply that genAI content will be posted; I edited.
    – Rebecca J. Stones
    Commented Jul 21 at 22:53
  • 1
    @AugustoVasques I can confirm that I saw the second SO once. The issue is that it is also upside down and populated by strange, flower like face monsters...
    – ꓢPArcheon
    Commented Jul 23 at 14:10

1 Answer 1

Reset to default
28

Yes, it seems like a terrible idea. No, it's totally not a violation of our GenAI policy.

As I read it, the CEO is imagining that a person will type a question into ChatGPT ("how do I write Javascript code to make every word on my web page blink?"), find the answer from ChatGPT unsatisfactory, and then be able to press a button to cause that question to be posted on Stack Overflow. That is not posting GenAI-posted content. That is posting user-written content: it is posting the question, written by a human, not the response from GenAI. In no way is that using GenAI tools to generate content for Stack Overflow.

It seems to me like a dangerous and ill-informed proposal. I think it's a terrible idea. But not because it violates the GenAI policy -- it's a terrible idea because such questions typically won't meet our quality standards, will be poorly received, will trigger frustration from answerers and curators, and will probably lead to frustration from askers when they inevitably receive negative feedback and pushback.

Push back against this proposal because it's a bad idea.... not because it violates the GenAI policy.

As for whether it's going to be possible, no one from the community here can tell you whether it's going to happen. We can only speculate at this point. We'll just have to wait and see.

Improve this answer

You must log in to answer this question.

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