107

I was going to post this as an answer (among many) under @Philippe's Statement from SO: June 5, 2023 Moderator Action, but that's currently locked, apparently due to concerns about the title, out of all things. Quoting from the linked post:

[Philippe]    Both sides share a deep and unchanged commitment to quality on this platform.

Each day of muzzles on AI moderation (including, and culminating with, moderators' strike) adds to the backlog of GPT nonsense that gets posted as answers across the sites, and allows more of it to get upvoted and/or accepted, even though the answer may be worthless or plain wrong. This degrades the "quality on this platform" and goes against all the reassuring feel-good official SE posts of late.

For context, most of my firsthand experience comes from MSE (math) lately, where GPTs can write deceivingly fluent prose, but prose which either has little bearing on the question, or makes no math sense, or sometimes both. However, the "answer" may sound authoritative enough for the OP to upvote/accept it, and that creates a precedent for subsequent readers to trust it and perhaps upvote it more. This can and sometimes does snowball, which erodes users' trust in the long run. Somebody will eventually figure out if an answer is wrong, and then will trust SE less because it misrepresented it as a good answer. And such word goes around quickly, more so if it turns out to be due to misguided policies rather than just accident.

To the illusion apparently held by SE that 'forget the policy, f*** the moderators, regular users can take care of this via usual downvote/close mechanisms' -- that's simply not true. I am a regular user, and I had as few as 3 downvotes reversed for "serial downvoting", even though the respective answers were blatantly GPTs (and eventually recognized/closed as such). As I wrote here before: "Of course, humans can and do make silly mistakes, too. Difference, however, is that most of us refrain from posting every one of them out of a sense of propriety, and, anyway, we would not be able to generate them at the rate of hundreds per minute even if we wanted to for whatever trolling reason."

As a footnote, all this talk from the CEO about profitable growth seems to lose sight of the basic fact that whatever the metrics are—page views, click-throughs, engagement, sponsorships—those are ultimately driven by the user base of the SE sites, the perceived quality of and trust in the sites' contents, earned over many years, and the thankless work of all moderators towards those commitments. All of that is too easy to lose, though SE appears to be blissfully oblivious to that.

Knowing Philippe is in the position to speak for the company, how do you reconcile your "commitment to quality on this platform" statement with the reality of the latest AI guidelines which IMHO directly and immediately hurt the quality on this platform, per what I wrote above? Or, are there any ulterior motives that you could share with the rest of us, maybe?

18
  • 46
    The CEO would do well to remember that it is ChatGPT that could not exist without Stack Overflow, and not the other way around
    – Foo
    Jun 7, 2023 at 7:23
  • 9
    That is one of the more interesting issues about the company's current stance -- by allowing "AI"-generated content on SO, they mechanically disqualify the site from participating in further "AI" training. Could it possibly be intentional? Jun 7, 2023 at 7:28
  • 33
    @FrédéricHamidi no. They simply don't see more than few days or weeks ahead. Jun 7, 2023 at 7:36
  • 4
    has anybody actually seen any GPT answers since the mod strike started. I haven't yet.
    – Kaia
    Jun 7, 2023 at 8:15
  • 5
    @Kaia I saw someone post a bunch of very blatant, very incorrect ones on Puzzling.SE yesterday. I downvoted and flagged them and all but one have since been deleted.
    – F1Krazy
    Jun 7, 2023 at 8:18
  • 4
    @Kaia Yes, a few.
    – dxiv
    Jun 7, 2023 at 8:33
  • 7
    Yes, large quantities, just like before the strike, @Kaia. Jun 7, 2023 at 10:29
  • 7
    The price also depends on the volume. On small stackexchanges with only say <10 new questions every day, it might be possible to only curate for quality and accept AI content if suitable. For SO, with the much, much higher numbers and where everyone thinks he/she is a programmer or could be one, it's different. But it definitely depends on the topic and size of the community. That's why there isn't a network wide policy on AI generated ban. That's why the price will be a bit different everywhere. Jun 7, 2023 at 10:41
  • 2
    @Trilarion Importantly, volume is influenced by moderation. If those posts are being removed, the people posting them have less incentive to continue doing so. If not, the quantity will increase. Jun 7, 2023 at 13:58
  • 3
    @Kaia The way that question is phrased, what exactly are you trying to insinuate?
    – Dan Mašek
    Jun 7, 2023 at 14:48
  • 1
    @Dan Mašek I'm not trying to insinuate anything. I was trying to see how many GPT answers were getting posted. After posting, i skipped through the first-answer review queue and found a couple, so they're definitely there if you go looking
    – Kaia
    Jun 7, 2023 at 21:51
  • @Kaia OK, I understand, thank you for clearing that up.
    – Dan Mašek
    Jun 7, 2023 at 21:59
  • 3
    Upon opening the bounty tab, my favorite place to find low quality answers, the question at the top of the list had a new gpt answer from a user who had posted 6 in the past 8 hours
    – Kevin B
    Jun 7, 2023 at 22:03
  • Could you please post a link to some of this "backlog of GPT nonsense"? Or just from Math SE, where your "first hand experience comes from", where AI nonsense has been upvoted/accepted and nothing been done about it? Is this "nothing being done" owing to strike or was it this way regardless? Are there links to some examples of this from before strike?
    – user13267
    Jun 21, 2023 at 2:50
  • @user13267 Yours is a good question, to which I do not have a good answer. Reason is that GPT posts tend to not last long on MSE. Even after mods' strike started, users who are unaware of or unconvinced by the strike continue to downvote and close the obvious nonsense. There is a list of 10 such posts here, and there is another one that I linked here, but they all point to posts that got deleted since, thus only visible to users with 10k+ rep on MSE, which you do not appear to have.
    – dxiv
    Jun 21, 2023 at 3:17

1 Answer 1

67
+50

... which erodes users' trust in the long run.

Worst of all, it erodes trust in the readers about themselves, and about maths.

If they believe that the senseless gibberish is the solution to a problem, then they might be hard on themselves for not understanding it. And their self-confidence, self-respect, and self-image (as someone fit for coping with math issues) suffers.

In essence, SE suddenly runs a huge disinformation network that sabotages humanity.

SE has a responsibility towards humanity as a whole, in their undeniable role as a distributor of information.

5
  • 14
    Wow. When you look at it this way, the thousands of people who've been volunteering their time to help SE so far (mods and curators of all types) perhaps "should" turn saboteur and try to bring down the SE network before the flood of AI-generated misinformation gets strong enough to undo all the good that SE has done over the years. Jun 7, 2023 at 12:02
  • 27
    @Randal'Thor I can't for the life of me understand how StackExchange was not made into a non-profit. How Joel Spolsky and Jeff Atwood invited us to build this empire of knowledge and human collaboration, only to toss it to the meatmincer of psychopathic capitalism. I know, in hindsight it's easy to deliver judgements, but the way the history of StackExchange played out, it is a shocking monument to a huge crime against collective human dignity.
    – Levente
    Jun 7, 2023 at 12:29
  • 13
    @Levente Well, they earned themselves and eventually early employees who stuck around a substantial payout. It would have been a tough sell to them to say "actually, we've decided your stock options won't be worth anything because we're going full charity". The psychopathic part now is the part that's actively destroying the value of what's left without any real chance to extract anything useful from it. Making money here and having a quality public network should be on the same side, because the reputation of the public network helps the sales of their paid products. Jun 7, 2023 at 14:02
  • 1
    @BryanKrauseisonstrike WRT the creators earning money for themselves and a few others: I meant, from the get go, they were selling our collective dignity for cash for themselves. How (in)visible this outcome was back when they decided to extend SO with the additional, non-tech SE sites? WRT network reputation contributing to earning potential: I may speculate that SE are not optimistic of any earnings in an AI-dominated world. They may expect losing substantial traffic to (Chat)GPT. So they may attempt to inflate the apparent value (number of new users), to attempt a quick cash-out.
    – Levente
    Jun 7, 2023 at 14:47
  • 20
    Another thought: from various communications since the Monica case, I find it quite noticeable that SE may value the information they try selling, but not the people who produce it. We, the community of producers / curators / maintainers of information are just a stubborn, meddlesome annoyance to them, whose management imposes a cost, and whose failure to manage causes unpleasant fallout in media. Who wouldn't want to replace this untameable powder keg of a community with a tame golden-egg-laying AI goose? (Even if many of those things that that goose produces are rotten inside...)
    – Levente
    Jun 7, 2023 at 18:00

You must log in to answer this question.

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