Skip to main content
Meta Stack Exchange

You are not logged in. Your edit will be placed in a queue until it is peer reviewed.

We welcome edits that make the post easier to understand and more valuable for readers. Because community members review edits, please try to make the post substantially better than how you found it, for example, by fixing grammar or adding additional resources and hyperlinks.

Required fields*

58
  • 152
    Attribution/citation/sourcing is not negotiable, but it is not included in the current tests. That seems like it's going to skew the results quite a bit. Commented Feb 4, 2025 at 17:54
  • 44
    If you can push on major AI providers to improve the attribution capabilities of their output, that would be a wonderful thing for the common good. I’m skeptical, but if it is possible, it will probably require a significant change to the type of results they’re generating that would potentially skew the results of your experiment. In my experience, when you’re using a system that tries to provide attribution, it hallucinates answers and then slaps on a tangential attribution that doesn’t actually say what’s claimed, unless it’s operating in a very narrow scope like an internal knowledge base. Commented Feb 4, 2025 at 18:01
  • 73
    Does seem a bit disingenuous to claim so much intent to have properly cited answers... when stack hasn't even proven it's possible, while moving forward with actually publishing ai generated content without it. Given how long it takes for stack to iterate on things now days it could be years before this ever gains any form of citation feature, long after far too much of it has been posted to reasonably clean up the mess. Commented Feb 4, 2025 at 18:21
  • 58
    We need fewer, higher-quality Q&As - AI could be useful as a personal assistant, to provide personalised responses, but it doesn't belong in public posts. But I guess that idea isn't as profitable as just scaling up volume at the expense of quality. Commented Feb 4, 2025 at 18:39
  • 77
    I don't quite understand the name of this experiment... This isn't an answer assistant, it isn't assisting anyone at answering a question... instead it's an attempt at filling in for the lack of answerers. Commented Feb 4, 2025 at 18:43
  • 77
    "Users with at least 50+ rep (default value) on the specific Stack Exchange site will be able to see and evaluate the private answers." - Currently, the association bonus is 100 rep, which means that I could join Arts & Crafts, Raspberry Pi or User Experience (UX), earn the bonus and immediately start reviewing answers without any prior participation on those sites. Wouldn't it be better to increase this threshold? Probably the same rep required to access the review queues (which at least requires some participation on my part before I can review anything). Commented Feb 4, 2025 at 19:11
  • 152
    "Committed to the Stack Exchange network being a place for human-curated knowledge and information." No. It is a place for human created and curated knowledge. You will have hard time finding experts willing to curate AI nonsense. Commented Feb 4, 2025 at 19:39
  • 27
    "The current experiment is integrated with an existing data partner...We are not able to disclose the specifics during this phase of the experiment." - Sounds legit. Commented Feb 4, 2025 at 22:13
  • 51
    I think this undermines the GenAI ban on some SE sites. As long as for example the GenAI ban lasts on SO, this answer generator would have no chance there. But when it is tested now, it may be introduced nevertheless. Wouldn't this again result in a moderator strike? If we do not pay attention, we might end up with more AI answers than human answers. A competition of AI generated answers and human generated answers will only result in alienating human answerers even more. This might not end well. AI and human content must be extremely well separated if AI content is to used at all. Commented Feb 4, 2025 at 22:30
  • 67
    What's the point of Stack Exchange if AI-generated answers are permissible? Querents could just type the question into Google and get the same (theoretically) result. Humans have an intuition that AI may never have, making the humans the only valuable contributors to Stack Exchange. ... Unless your goal is to be rid of the pesky humans.... Commented Feb 4, 2025 at 23:33
  • 36
    While a lot of people are unhappy about this, I do appreciate you taking the time to openly let us know what is happening, even if we get little say in it. Commented Feb 4, 2025 at 23:49
  • 52
    Just call a spade a spade. The last time you did an experiment, it magically just became a new feature with some random unexplained marginal growth metric Commented Feb 5, 2025 at 9:39
  • 51
    @ꓢPArcheon I just went through the metas on the 3 sites that "volunteered" and found no prior discussion on any of the sites, just announcements: "we are going to do this experiment now". I don't think there was any public discussion at all. Getting the majority of all moderators to agree on a small site with just 1 or 2 active moderators won't be that hard. Commented Feb 5, 2025 at 13:43
  • 26
    Generally, the people qualified to properly proofread an LLM generated answer are already the ones answering questions. This seems like it will just encourage people who have a moderate understanding of the topic to mark the AI as correct if the output "looks about right" (which it usually does, creating text that seems relevant is what LLMs do best) TL;DR: I am willing to help humans solve problems, I am not willing to proofread LLM guesswork. Commented Feb 6, 2025 at 23:16
  • 23
    To me, the most frustrating thing about seeing AI answers on Stack sites is the thought "If the asker wanted an AI-generated answer, they would have asked an AI." Adding this "Answer Assistant" will change that thought to "Why bother with asking a question on <StackSite> at all if an AI is just going to answer it anyway? Just ask an AI, skip the middleman." I realize you want these answers "curated by the community", but I have negative interest in doing that... I like helping people, not AI 🫠 Commented Feb 7, 2025 at 14:15