3

enter image description here

There's a ban on AI content, because it is generally known and accepted that AI produces slop. The whole reason of this site is connecting with people who have knowledge and experience. So why then, are the SO devs advertising their own AI?

Has anyone tested to see if they get punished for answering a question with the SO AI?

I understand that the ban is a community decision and the AI is a dev decision. That's the point. The community has already made it clear that we don't want this sort of thing in the community, but the devs have plopped it down here anyway. This will probably just increase the amount of AI slop generated posts and answers in the community, despite our clear indication that we don't want it. So why are the devs doing this?

25
  • 9
    See Labs experiment launch: stackoverflow.ai Commented Jul 11, 2025 at 0:58
  • 3
    What happens when you click on that link? Does it look like you're on Stack Overflow? It doesn't look like it to me. Commented Jul 11, 2025 at 1:10
  • 2
    The one you circled in your question called AI Assist. Commented Jul 11, 2025 at 1:38
  • 1
    @PresidentJamesK.Polk It goes to stackoverflow.ai, but the point isn't "Where does the link go", the point is "Why is the link here?" By putting it here, they're going to increase the number of people using LLMs, which will probably cause some increase in the number of people posting AI generated content in this community. Commented Jul 11, 2025 at 1:42
  • 11
    The point is, it's an experiment that is not happening on Stack Overflow but rather on another site. You're exaggerating the risk that someone will put AI stuff on Stack Overflow because of this innocuous link, stuff that we can just delete anyway. If it becomes an unexpected problem, we can deal with that. Commented Jul 11, 2025 at 1:46
  • 5
    eh, it's a very real risk, but one we have no control over. the company decided that because it's just a link that goes somewhere else, the community's feedback is even less important than normal. Commented Jul 11, 2025 at 3:37
  • 7
    Put another way, make sure you express your concerns on the MSE post, where at minimum it'll be findable by those who want to look back at this mess later. Commented Jul 11, 2025 at 3:53
  • 2
    A feature that our corporate overlords are offering despite it being useless. The community still will downvote and delete all AI generated content. I asked how to disable it and it hallucinated a response Commented Jul 11, 2025 at 4:12
  • 6
    I don't think that it's "the devs" making these decisions. Commented Jul 11, 2025 at 5:59
  • 3
    Just because the link goes to a slightly different domain I wouldn't say it's somewhere else. Surely the new feature is part of the brand and the portfolio of the company. There are also links back to more or less relevant content and the ask page. One could see it as part integration of AI. Obviously some will draw the line earlier and others later. Commented Jul 11, 2025 at 7:18
  • 2
    Also see meta.stackexchange.com/q/411281/334566 Commented Jul 11, 2025 at 8:32
  • 6
    "I see people are already rushing to the defense of the devs to point out that the ban is a community decision and the AI is a dev decision. I understand that. That's the point." — then how are we meant to discuss this? If you're hoping the staff will explain something to you, then who is "we" in the title? Commented Jul 11, 2025 at 11:08
  • 9
    "We" the community don't want AI. The company is absolutely not included in that. Commented Jul 11, 2025 at 14:04
  • 3
    @user4581301 But the big companies (Google, Microsoft, Anthropic, OpenAI, Meta) are working on that and are worth Trillions. Who needs more? Anyway I'm not getting paid for that. Commented Jul 11, 2025 at 22:03
  • 2
    Does it need to be more complicated than "Stack Inc. needs income to stay afloat?". The site itself isn't doing too hot in its current state. Right now would be the exact moment to get the training model done in any case because the repo is right now mostly human in origin; wait too long and it is absolutely ruined by AI generated content. Commented Jul 14, 2025 at 9:03

2 Answers 2

Reset to default
13

I think you've conflated two very different things.

The community rejected AI content being merged inline with human content. And it did this for very good reasons. It especially rejected AI content maskerading as human content by way of users and bots posting it as if humans were authoring the answers.

The feature you're pointing to is not injecting AI generated content into a space for user written content and it is not expecting the moderators / curators of the user written content space to also moderate and curate the AI generated content.

Neither is it blurring the boundary between the two spaces such that readers are unaware of the division between user generated content and AI generated.

I don't believe it's wise to deny that LLMs have a part to play in helping people learn. This would be as nonsensical as to suggest Google Search has no part in people's learning (AI driven since 1998). Of course there are many well documented dangers from simply trusting LLMs but, in the right hands, LLMs are a really powerful tool for accelerating learning.

LLMs are undoubtedly changing the way we code and they look set to change a lot of things in life over the next few years, but the technology is still in its infancy and people are still learning what it's good at and frequently making mistakes.

Ultimately the world is changing and LLMs will be a larger and larger part of a developer's life. SO can either try to experiment with new and innovative ways to use them or it can stick its head in the sand and pretend the world isn't changing.

No we don't want AI content on SO Qs and As. But that's a different topic than suggesting SO can't have AI on the site at all.

2
  • 2
    "And it did this for very good reasons." The good reasons were not only the masquerading (after all there was also an experiment with clearly marked AI content), but also the substandard quality of the generated content. People were concerned about quality and probably still are. Offering the same service somewhat distinct in the site might be a hint, that quality of AI generation has improved and maybe AI should actually be part of Q&A in some sense or that the service with act as additional even closer distraction. Something doesn't add up exactly. But it probably doesn't have to. Commented Jul 11, 2025 at 21:55
  • 3
    @NoDataDumpNoContribution No I don't agree. The quality is a problem if you mix it with authoratative questions and answers. But LLMs are a tool just like google search is a tool. We don't believe every link we read on google. We must learn how to use LLMs for what they are good at (and they are good at some things right now). I have to cross reference every 10 year old SO answer already. I have to cross reference LLM output too. That doesn't mean I don't use either. 75% of the time LLMs get me where I need faster and the other 25% wastes too little time to be a risk. Commented Jul 11, 2025 at 22:12
9

So why are the devs doing this?

The "devs" explained why they are doing it in a meta post Labs experiment launch: stackoverflow.ai.

In short they want to give people a new way for quick help with the additional possibility to also see content from here.

You probably don't see any advantage in that? Maybe even a risk? They may see it differently.

They probably concluded that if you can't beat AI, you can as well jump on the train and try to put up some ads for the human curated library part. This doesn't mean that their support for the ban of AI content in the Q&A part changed although admittedly one could say that this links both worlds closer together. That is probably their goal. I could imagine that they have something like "best of both worlds" in mind.

From another point of view one could also see it as defeat. In this model, LLM usage comes first and SO usage only second (if only there was a way to do it the other way around) and they surely will have to pay for the usage of the LLM somehow. Or people might prefer the original LLM locations anyway and the effort is wasted.

We'll see how it goes. Maybe it even helps sparking new interest in human Q&A. Although some of us think that the mission is accomplished already anyway and all (or almost all) programming questions already have been asked.

Interesting times.

4
  • 2
    Although some of us think that the mission is accomplished already anyway and all (or almost all) programming questions already have been asked.. Forgive me being blunt but that is complete nonsense. Very many new questions are still being asked, just not on stackoverflow. Tech hasn't stopped evolving over the past decade it's accellerated. And changing tech is continually driving new questions. But a lot of the new questions are heading over to competitors, especially github discussions. Commented Jul 11, 2025 at 20:23
  • @PhilipCouling Good to know where the questions are. So I can specifically check GitHub discussions. Anyway I was just summarizing answers to questions about the diminished rate of new questions. Commented Jul 11, 2025 at 21:59
  • Oh I'm well aware it's been bouncing around the (meta) echo chamber. I could probobally name it's most vehiment proponents. But it remains something of a conspiracy theory that neither fits the available evidence well nor has supporting evidence of it's own. Commented Jul 11, 2025 at 22:26
  • @PhilipCouling My impression is that meta got more extreme with fewer participants. Probably some kind of selection effect. Anyway, you know where all this leads to? Human Q&A as the last resort. We are supposed to answer only the questions that chatbots cannot answer and then only so that they can in the future. Who will volunteer for such a bargain? I admit that I have big problems with it. I'm not donating my time and knowledge, so Stackoverflow AI gets better and my name is never even mentioned once. They will never solve the attribution problem. Human Q&A might just be dead at this point. Commented Jul 12, 2025 at 8:21

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.