-18

In a discussion which I will not link, because it has already generated far more heat than light, multiple users appear to be claiming that the strike letter stands for a blanket opposition, either to AI in general, or to AI on the SE platform. Normally, I would not worry too much about random chatter in the comments, but the answer which makes this claim is being heavily upvoted, indicating that the community might actually agree with this characterization. I strongly disagree, and I want to lay out my concerns in greater depth.

Before signing the letter, I read it very carefully. I wanted to be sure that I agreed with everything in the letter, so that I would not be seen as advocating for something I might later regret.

Let's start with the TL;DR at the top:

Stack Overflow, Inc. has decreed a near-total prohibition on moderating AI-generated content in the wake of a flood of such content being posted to and subsequently removed from the Stack Exchange network, tacitly allowing the proliferation of incorrect information ("hallucinations") and unfettered plagiarism on the Stack Exchange network. This poses a major threat to the integrity and trustworthiness of the platform and its content.

This opening succinctly captures the context of the strike, what the company did to deserve it, and why the moderation diktat is causing problems. Specifically, it identifies three main issues:

  • Moderators and communities have been unfairly stripped of their autonomy.
  • AI answers are sometimes wrong, and undermine confidence in SE answers.
  • AI answers are plagiarized, because the person posting them did not write them.

I agree with all three of these criticisms. The letter goes on to reiterate all three of those points, and adds a fourth:

In addition, the details of the policies issued directly to moderators differ substantially from the guidelines outlined publicly, with moderators barred from publicly sharing the details.

The letter then goes on to discuss what the signatories will no longer be doing, why the company should care about those things, and how the company can remedy the situation (in short: retract the bad policy).

Nowhere within the four corners of this letter is there any discussion whatsoever of the social benefits or harms of AI writ large, aside from the existence of confident-but-incorrect ("hallucinatory") outputs. Nor is there any discussion of the company's overall approach to AI, or the CEO's apparent enthusiasm for it. There is a rather vaguely-worded yellow banner at the top of the page which suggests there are unspecified other problems, but that's not part of the letter, so I don't need to try and interpret it. The FAQ doesn't change anything either; it just expands upon the issues described above. In short, the letter says nothing at all about AI except as it pertains to the specific moderation policy that set this whole thing off.

This is a good thing. A letter like this needs to be focused on one topic, one problem to be solved, or else the striking group will come into conflict with itself and eventually splinter. Personally, my feelings about AI are deeply conflicted, and I am uncomfortable with taking a public stance on whether it will do more good or more harm in the long run. Having read this letter thoroughly, I was convinced that my signature could not be interpreted in that way. That is why I signed it.

If people want to oppose the new site on the grounds that (they believe) AI is evil, I have no objection to that. If people want to point out that it is tone deaf and imprudent for the company to return to the subject of AI at this delicate moment, that too is fine by me. There are a thousand other problems with the way in which the company has handled this site launch, and I wrote an answer pointing out one of them (they are asking us to help them launch this site, but we're all on strike and probably won't be willing to help). But if people want to claim that I, as a signatory, am personally opposed to the existence of an AI-related site, merely because I signed a letter which doesn't actually say that, and in spite of the fact that Artificial Intelligence already exists, I feel I have no choice but to speak up, and clarify that that's not what I meant.

Maybe I'm overreacting. Maybe the answer in question got upvoted because folks agreed with everything else in it (this was a rather small point in a rather long answer, after all). Maybe I misinterpreted the discussion (at least one person claims I have). Maybe you'll all look at this and say "well, Kevin, of course we don't think the strike letter meant that!" I sincerely hope for that reaction, anyway, because I have no desire to pick fights with the Judean People's Front. Or to be more direct, I don't want to have to reach out to the folks running the strike letter and ask them to remove my name. But if the community is dead set on reinterpreting my signature's meaning, I may not have a choice.

Since this is a Q&A site, here's a question: Given everything I've explained above, have I interpreted the strike letter correctly, or am I wrong and it's really about AI? Is there some subtext I'm missing? Was there some other discussion, perhaps on Discord or elsewhere, that would change the apparent meaning of the letter?

23
  • 21
    I would argue moderator trust, autonomy and treating the mod team and community as equal partners is the real core issue but the specific issues at hand are practical achievable goals. Jun 22, 2023 at 23:42
  • 2
    @JourneymanGeekOnStrike: On meta, downvotes indicate disagreement. Right now, this Q is sitting at -4, which I'm interpreting as meaning "No, we disagree with you. The strike letter is about AI and you should withdraw your signature." If that is not your interpretation, please consider writing an answer.
    – Kevin
    Jun 22, 2023 at 23:44
  • 9
    FWIW, I downvoted because your bolded question is opinionated and the decision to (or to not) retract your signature is a decision for you, and not really one we can answer. If you edit your question to just ask about the scope of the letter, then that’s different
    – cocomac
    Jun 22, 2023 at 23:46
  • 30
    "Does the strike letter oppose AI?" ─ can be answered very easily by reading the strike letter.
    – kaya3
    Jun 22, 2023 at 23:47
  • 7
    "I strongly disagree, and I want to lay out my concerns in greater depth." ─ OK, so you don't want to link to the thing you're writing a rebuttal to, because it's not a productive discussion, but you want to continue and participate in that unproductive discussion, and expand it into another Q&A. I can't say I see the point in doing that rather than responding to the thing you want to rebut in the place where it is posted.
    – kaya3
    Jun 22, 2023 at 23:50
  • 4
    @kaya3-supportthestrike: Have you ever tried to have an argument in an SE comments section? It sucks. The thing I'm responding to is like one sentence long, plus a bunch of people arguing over that sentence in the comments. I don't see how that would help inform your opinion, but you can find it via my profile if you think I'm hiding something.
    – Kevin
    Jun 22, 2023 at 23:52
  • 23
    It's not that I think you are hiding anything, it's rather that there is a mismatch between your stated desire and your actions. But I took up your invitation to find the offending answer, and it seems the claim you object to is: "So many of the network sites are rejecting AI. There's a whole strike on because of how hard y'all are pushing AI." But that sentence does not mean the strike letter "opposes AI", it's about the way SE is pushing AI on the community which does not want AI-generated content here on Stack Exchange. So your rebuttal is premised on a severe misreading of that.
    – kaya3
    Jun 22, 2023 at 23:58
  • 2
    @kaya3-supportthestrike: My concern is not with the specific wording of that answer in particular. It's with whether I have badly misinterpreted the strike letter, which IMHO is the actual thing that matters here, not the wording of some random MSE answer. If you think I've done a poor job of conveying that, then please suggest an edit.
    – Kevin
    Jun 23, 2023 at 0:02
  • 11
    For what it's worth, if you want to see downvotes as answering the question, I would interpret a downvote here as "no, of course signing the letter doesn't mean you hate AI!". I don't really see how a downvote would mean "yes, signing means you hate AI".
    – terdon
    Jun 23, 2023 at 0:15
  • 2
  • 16
    I just don't get what you're after here. Did you read the strike letter and want to support it? Then feel free to sign it. Or not. All this AI talk completely misses the point of the strike. Furthermore, one can welcome AI advancements without agreeing with the policies that Stack Overflow Inc has set around it.
    – mason
    Jun 23, 2023 at 3:32
  • 6
    Considering that someone felt they had to post the same comment "It's deeply disappointing to see early and vocal supporters of the strike so eager to jump on the bandwagon" at least three times on the "prompt design nomination post" and got multiple upvotes every time I think this question if well deserved, and it feels a little blind to claim otherwise. This is not something the OP hallucinated out of their mind. Some users see ANY form of support to AI related project as a betrayal of the strike.
    – SPArcheon
    Jun 23, 2023 at 10:02
  • 3
    @user13267 obviously I don't support many of the things the company tried to do recently, and in a way I agree that even the new proposed site is leaching resources that could be used for other improvements that the community has been asking for year. Yet, imho, it seems very obvious that there is also a minority of users that downvote everything about AI or attempting to talk about AI use out of resentment to the company instead of evaluating every proposal separately.
    – SPArcheon
    Jun 23, 2023 at 11:17
  • 5
    Your interpretation of downvotes: "No, we disagree with you. The strike letter is about AI and you should withdraw your signature.", is widely flawed. You didn't write a one-line question. You've had way too many statements and questions in your post, that one can oppose to, hence proceed to downvote. FWIW, I did not downvote, I do disagree with many of your interpretations, and don't think the strike letter is about hating AI. Your decision about your signature is somewhat immaterial.
    – M--
    Jun 24, 2023 at 2:39
  • 5
    As someone with no feelings about AI (except maybe hope?) , my downvote here means I disagree that the strike letter means we hate AI. Don't try to put words into my mouth. It's not really about AI at all. It's about how SE Inc. treats us, "us" being human beings and contributors to the (wide) community, and the value of our moderated answers here. Jun 24, 2023 at 13:36

7 Answers 7

54

I read this question very carefully, and as someone who helped draft the "The problem with AI-generated content" in the initial moderation strike post, I wanted to chime in a bit.

I personally don't hate AI by any means, and I signed the strike letter. Mithical, one of the representatives who is conversing with SE about the strike and who is a moderator on Artificial Intelligence (of all things...), likely also doesn't hate AI (though you could probably do some mental gymnastics and figure they are somehow hate-moderating that site or something). I think I can speak for a lot of people who signed the letter by saying... We don't hate AI. The potential for AI to help all of us in a variety of tasks, including curating SE, is quite high. A lot of us also, myself included, were actually quite interested in this example of generative AI on SE. Now... It kinda grew into a little bit of a dumpster fire, but the bones of a good idea were certainly there, they just need some refinement. Well, maybe a lot of refinement...

The strike isn't about AI... Not really, anyway. It's about the lack of communication between SE and its supposedly trusted userbase about issues that directly affect us. terdon did a great summary about the problems, but I wanted to highlight one more: The ban on AI content had widespread community support, and continued to have widespread community support across the damn network. Why, oh why, would you strip away that decision from the community via an edict from on-high? That's why I signed. This was the single most anti-community action since the Monica debacle, and it just continues to get worse as time wears on.

So, look... By signing the letter, you're not saying you hate AI, even if it's caused a lot of us some headaches here on SE, especially on SO. You're just saying that you're unhappy with how things came to this point, how it's continuing to go, and that we need to have a chat about it.

1
  • 22
    As someone whose "real" job has involved a fair amount of work with AI, including multiple papers published on the use of neural networks and other ML techniques, I also think it is absurdly stupid to claim that the strike is about a general hatred of AI. That said, I do willingly admit that I have a deep skepticism of generative AI, and what I've seen happen just on Stack Overflow has nearly driven me to fear and outright oppose its proliferation (not because of gen AI, really, but because of humans and their gullibility). Still, the scope and aims of the strike are quite clear and focused. Jun 23, 2023 at 6:35
47

I was involved in the drafting of the letter and have been part of the strike since the beginning and, frankly, AI is the least of my worries. My main issues are:

  1. The company is breaking their promises to us. Again. They are showing that when they sign an agreement, what they mean is that the other side should be bound by that agreement, but not them.

  2. The company is slandering moderators. Again. They are accusing us of being so dumb as to rely on AI detector software which, as anyone who's used them will know, simply don't work very well. Which is why we never relied on them.

    They are also accusing us of wrongly suspending users. If true, this is very serious, and we need to know about it so we can apologize to the affected users and ensure it doesn't happen again. However, they have yet to show one single example of a wrong suspension. I kid you not, they haven't shown any of us, any mod across the network, a single incorrect suspension over AI.

    This, by the way, is a problem in and of itself. Moderators are human, and of course we make mistakes. Given the number of sites and users and the number of suspensions issued across the entire network, you would expect some to be wrong even by pure chance. If you can't find any wrong suspensions, that means you haven't bothered to look. So the company feels it is easier to accuse than to actually sit down and check for themselves.

  3. The company has imposed quality criteria on the sites. This is a new level of over-reach, one that has never been seen before and one that goes against the founding principle of all SE network sites to be "run by us, the community". If the company tells us what constitutes a "good" question or answer for us, then the sites are no longer run by a community of users but are instead run by a for-profit entity for its own ends.

    On top of that, the specific quality criteria they imposed are just nonsensical and basically force us to allow AI generated crap, even blindingly obvious AI generated crap, with all the issues that you point out in your question, on the sites.

All this to say that while the specifics of the AI policy triggered the strike, the problems run much deeper, so no, I cannot see how it makes sense for anyone to interpret a signature on the strike letter as a vote against AI in general. Artificial intelligence can be an amazing tool if used well and in the right context. I personally am less convinced about LLMs like chatGPT, but even they, despite their current limitations, clearly have huge potential. I don't want them to be answering posts on SE, mind you, I come here to get human experts answering my questions, but that doesn't mean the tools themselves are bad.

24

Thinking that I might be involved here, I'll offer my perspective.

I signed the letter because I am in complete opposition to the company's latest moderation decision to prevent moderators from getting rid of AI-generated content.

At its heart is a familiar blight that I've seen in these interactions - an utter failure to communicate with the community that is claimed to be so treasured and valued. At least, that's what's written about us in blog posts and such. We're beloved as long as we're on the same page.

My firm belief is that Stack Overflow's value as a network is in the quality of its posts. For there to be a network, good answers are a must. By preventing moderators from removing AI-generated answers (which are by no metric "good" in any way), the value of the network is diminished in my eyes.

By and large, the network has rejected the use of AI in their content. So a decision - that doesn't have a lot of particularly strong statistics behind it, by the way - that is completely the opposite of what the community is actually saying is something I fundamentally oppose. I don't really want to accuse anyone of lying, but it's a hard sell for me to believe that the circumstances around this decision are entirely truthful and honest.

I do want to circle back on the "blight" I had mentioned earlier. Honestly it feels like all of us in this mire never really advance beyond some kind of uneasy "truce" after something a bit crazy happens. It's been escalating slowly and slowly ever since the Summer of Love, where the company's decisions were for the most part a bit rushed but no big deal at first, then with increasing escalations just not involving the community that is claimed to be at the heart of all of this.

I've posited that they probably mean a different "community" than the community I'm seeing, but... not much I can do about that.

Anyhow, the letter itself isn't saying "AI is evil, ban AI" or anything along those lines. Signing it doesn't pigeonhole you into that perspective.


Personally I think it's fine for a tool, but the use cases that the company wants to advance to help the community feel like they'd add more to our workload, with needing to check if a title makes sense to begin with (because the OP who uses that tool can't do it themselves) and whatever in the hallucinated heck this tool was when trying to suggest formatting edits (which, again, the person who would use this tool a lot isn't the sort of person who would pay a lot of attention to these kinds of edits, which is actively harmful to the quality of the site).

3
  • Somehow I finished writing my answer right at the same time you finished yours. Please do let me know if I'm reading too much into what you had said over there.
    – duplode
    Jun 23, 2023 at 4:37
  • 4
    Looks like you're in the ballpark @duplode. I realize my speaking style can be a bit on the extreme boundaries, but you've managed to condense it pretty well. Personally I do think that the policies the company is pushing is still related to the chief reason I signed on, which is that they're simply not communicating with the community they claim to value. But yeah, nothing's missed here.
    – Makoto
    Jun 23, 2023 at 4:54
  • 3
    "I realize my speaking style can be a bit on the extreme boundaries..." Hold my beer. Wait, no, give it back, it's too damn hot here! Jun 23, 2023 at 7:46
16

I would say it is a specific issue we are concerned by.

It's less about the "rules on AI" or "AI" than the moderation teams requesting guidance, being told to do it ourselves, then overruled without warning.

The social contract was "We'll take care of our communities, you help us when we need it" and certain other considerations.

However, in the face of systemic issues, we'll never get back to moderating our communities if we wait for perfect or good enough. We've had issues we've brought up that probably predate most of the community team, or maybe even most employees.

We've been aware of worse engagement, finding it hard to replace core community members who left, and discoverability issues with various aspects of the platform. We've asked for better representation in the company - and even spoken out against how the company treats its staff. We can't fix that now.

Showing the company trusts us on this issue, however can be a starting point.

It could be a good many other issues, but this is what we have now, and we need to know the company's willing to put aside its, well hubris, to trust that we know what we're doing, and communicate as equals.

We don't oppose AI. We feel cynical use of AI tools for answers will dilute the quality of information here and cause both short and long term issues for future users. We'd like to be trusted to take care of our communities. That's not a big thing to ask, when we've been doing it for years.

14

To answer this question, I find it inescapable to consider the highly upvoted answer it mentions, and in particular the passage from it you find objectionable.

This is just so tone-deaf to everything the community is saying about AI.

You couldn't be more disconnected from reality if you tried here. So many of the network sites are rejecting AI. There's a whole strike on because of how hard y'all are pushing AI.

While I can't speak for Makoto, the natural way to read this passage is to take it as referring to events that have transpired in SE, so that:

  • "everything the community is saying about AI" refers to the views condensed in the letter;

  • "So many of the network sites are rejecting AI" refers to the policies actually adopted by network sites to avoid being overwhelmed by unverified AI-generated posts; and

  • "how hard y'all are pushing AI" refers to the company making the policies mentioned above ineffectual, which is a direct reason for the strike.

(With respect to the third point, one might object that the mere neutering of the policies doesn't qualify as "pushing" anything. However, the whole set of AI-related initiatives currently being prioritised by the company are a possible frame for making sense of the neutering, even if they weren't the triggers for the strike. Regardless of whether such a frame is ultimately correct, adopting it is well within Makoto's remit as the author of the answer.)

The overarching theme, therefore, keeps being the AI-related policy decisions by the company that affect the operation of the sites and the workings of their communities.

Now, if one has broader concerns about AI, and in particular concerns about overzealous deployment without due regard for externalities and side-effects, it wouldn't be surprising for them to interpret the current events at SE under that lens. These broader topics are not addressed by the letter, as the answers by terdon and Spevacus note, and so various views about them are compatible with signing it. The flip side is that signatories of the letter are free to comment on the situation at SE in a way that is framed by their broader views about AI without fear of somehow tainting by association all the other signatories.

Summing it up: there is no reason to extrapolate from an answer written by a single person to however many other signatories there are, and your signature is not being reinterpreted.

5

There are two "mostly separate but not completely" issues at work here.

First is the strike.
We can lie all day if we really want to but the strike trigger, the final straw, was a policy about the moderation of post generated using simple text generators (with ChatGPT being the most common one and the name that everyone used).
I won't repeat all the reasons here, basically users are concerned that the network would be polluted by grammatically correct text that has no guarantee of being meaningful or to not contain unreal, made up things.

On the other side we have a company obsession with AI.
The generated content policy is just a part of a bigger picture. AI generated post edit suggestions, alleged monetization of the data dump, firing people and redirecting resources to side project like an AI research and demo site... all contribute to make users quite angry.

You can get where this is going:

  • You have users who are just angry about the misuse of a text generator as a knowledge extraction tool (those exist but ChatGPT is not one of them).
  • You have users that are angry about the fact that the company is directing resources to AI at the cost of firing people while ignoring all the small usability requests that are on hold since 6-8 years B.C.
  • And you have people that hate AI technology with passion and consider them to be the new "Internet of [redacted] (warning - twitter link)".

The first group could be actually interested and willing to help on the new site, the second group will probably see it as the next resource leech and hate you for helping and the third one will hate you because AI.

All the three groups probably signed the letter. But they are not the same.

6
  • 5
    "[#3] And you have people that hate AI technology with passion..." I've personally spoken with many of the people who have signed the letter, and all of the Stack Overflow mods who have signed, and none of them reflect this sentiment. So, while this may be theoretically valid speculation, it is inaccurate and does not reflect the reality of what the strike is about or what the strikers' positions are. Jun 23, 2023 at 10:20
  • 2
    @CodyGray-onstrike then maybe you could explain why someone would be "deeply disappointed" at mods nominating themselves for helping on a separate site about AI related content. As I said on my other comment though please don't try to sell me the "it is just about the timing" excuse since that user is apparently mad even at mods that started their answers with "when/if the strike will end"
    – SPArcheon
    Jun 23, 2023 at 10:34
  • 1
    Because the strike is a battle of stamina and attrition. If you're already planning on what you're going to do once the strike is over, half the battle is already lost. There's some indication that the company just plans to wait us strikers out, hoping that we'll just give up on the hard-line stance and come back to the work we've been doing for the past decade-plus for free for them. Jun 23, 2023 at 13:03
  • 1
    @CodyGray-onstrike sorry, your further comments on other questions made me rethink your position. I will stop at two things. First, I would argue that the commenter was quite lucky that they didn't say that line to some other mod. I have also personally spoke with some and I can ensure you I could name a few that would really get upset at someone telling them how to act to not "fail their expectations and disappoint them. Second, I would also argue that said users are probably aware of that leak too (I assume those are the indications?) and I doubt they will fall for the "free cake".
    – SPArcheon
    Jun 23, 2023 at 16:15
  • 1
    Third, I think that if you really fear users thinking what to do after the strike then this is an hate war, not a strike. I personally still carry resentment for how Monica was harmed but I don't hide my idea behind "I am on strike now" - I blatantly say that I can't bring myself to trust the site enough to be engaged in help that involves for example things like signed agreements. This seems to be less about the strike and more about resentment against the company. Not that I can blame someone for this, but let's call a spade a spade.
    – SPArcheon
    Jun 23, 2023 at 16:19
  • 2
    Fourth.... shall I point out that this user was so eager to copy and paste his comment that they even commented on wizzwizz4 reply? As Adám pointed out, I am not even sure they took some time to read that answer before deliberating that they "were deeply disappointed". So, forgive me if I disengage now but this is getting quite ridiculous and I really don't have time to argue more
    – SPArcheon
    Jun 23, 2023 at 16:21
5

I upvoted your question because I agree that this is something we should clarify. However, I don't think the downvotes generally indicate a general adversity towards AI in general. And I'm afraid your formulation is somewhat problematic.

  • Some users will no doubt downvote to indicate that they do not agree with the final question, i.e. to indicate that they do not agree that the strike letter is an indication of a deeper skepticism towards AI.

  • Some users may well be downvoting because, as a matter of fact, bringing up this topic can very well be construed as an attempt to splinter our unity. In fact, casting doubt over the very simple and straightforward analysis of the actual text of the strike letter you provided yourself seems like something the company would do to sow fear, uncertainty, and doubt on the strike and the motivations of the participants.

  • Certainly, many of us are deeply skeptical of the company's sudden turn to jump into the current hype around large language models, and how they ended up offending the community in the process; and some might have downvoted to express this limited form of specific, limited "AI skepticism."

  • Finally, I'm sure there as a small fringe of downvoters who genuinely hate AI in all its forms. This is hard to grasp for the rest of us, but after all, there are also people who insist that the Earth is flat, etc.

6
  • 3
    I think this question was caused by multiple comments like this one (picture if the comment is deleted here). The comment have multiple upvotes. As you see some expect that users who signed the letter will oppose any AI related project. Also, this is quite tangential to my answer to so I mostly agree here.
    – SPArcheon
    Jun 23, 2023 at 10:13
  • 5
    But that's not what blackgreen is saying at all, and I think that interpretation is equally as disingenuous as the one that motivated this question, @SPArcheon. Jun 23, 2023 at 10:18
  • 1
    @CodyGray-onstrike seriously? When the comment starts with a direct passive attack "It's deeply disappointing to see early and vocal supporters of the strike so eager to jump on the bandwagon.", "deeply" being bold? When they claim to be mad of the timing "It's deeply disappointing to see early and vocal supporters of the strike so eager to jump on the bandwagon." on a post that starts with "I'm cocomac - When the strike ends, I'd like " (emphasis mine)? Sorry, I am not buying this.
    – SPArcheon
    Jun 23, 2023 at 10:22
  • The link is to an identical comment under starball's nomination, which doesn't have any conditions pertaining to the strike or its goals. The way I read it, blackgreen is simply asking strike participants why they are eager to provide the company with their contributions if they are serious about striking. The fact that this initiative (too) is AI-related seems unimportant.
    – tripleee
    Jun 23, 2023 at 10:46
  • @tripleee they say things like "jumping on the bandwagon" and "when the CEO said the company's genAI efforts have community support he was right after all" and you say that "The fact that this initiative (too) is AI-related seems unimportant"? Again, we have to agree we disagree.
    – SPArcheon
    Jun 23, 2023 at 11:13
  • 4
    "A direct passive attack"... isn't that a three-way contradiction? Anyway... Yes, seriously. Yes, I read the comment; I know what it says. I also know the author. Yes, the timing and the optics are the concern. A strike succeeding is largely about perception and impact. Jun 23, 2023 at 13:01

You must log in to answer this question.

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