34

I'm reading the answers at Our Partnership with OpenAI, and a lot of people seem very unhappy about the partnership.

I feel violated, cheated upon, betrayed, and exploited.

And so on. But... why? This is a honest question. Am I missing something important about the announcement?

From skimming over the announcement, SO is going to just sell access to user content for training ML models. Is this what people are unhappy about? But I'm fairly sure that the content is already scraped and/or taken from data dumps for that purpose, so this looks like it doesn't change anything.

Some worry that SO could sell the content without CC-BY-SA restrictions attached to it (the ToS is vague, it may or may not allow this). This is more questionable; is this the main issue? While questionable, I'm fairly sure SO content was already used for ML purposes, disregarding the license.

Or is it just the general stance SE takes, by showing any support to GenAI?

14
  • 8
    @HolyBlackCat before, the AI companies needed to scrape them like any other search engine. Now, they have a whole API dedicated for that, which means getting faster, and real time access, to everything, and all with the full consent of the company, and without the consent of, well, those who write the content. That is the problem, which I was sure was pretty much obvious. So yes, this can be compared to someone who gets things done to them without consent, aka violated. Commented May 12 at 12:29
  • 5
    If the community sees the current AI fad as something dysfunctional and counterproductive, they might also not be keen on the company selling access to our freely given content to AI companies. The company's interest in making money off our content does not necessarily align with the community's interest in how our content is used.
    – khelwood
    Commented May 12 at 12:52
  • 17
    And the partnership announcement would probably be much better received if it would actually have some more details about the deal and more information on how it concerns the communities, instead being a mostly link only post to press releases and blog full of illegible marketing speak. Commented May 12 at 13:25
  • 6
    I think the main thing is it’s hard for users to envision how generative AI can possibly provide attribution. This has always been a problem, but SO wasn’t taking money for letting companies at its (and our) data. On the other hand, it’s possible people are jumping to conclusions a little bit, since there aren’t many details available yet. In that way the anger is a little misplaced I think - I’d be mad about the lack of details, which can’t be because they haven’t been ironed out yet. The deal’s already signed. Most big deals like that don’t consist of a single sheet of paper with “TBD.” Commented May 12 at 14:03
  • 5
    SO user base basically helped to train OpenAI's billion-dollar LLMs and they don't get anything in return, not even attribution to their knowledge/content/time: theregister.com/2024/05/07/stack_overflow_openai . That's how you treat slaves. That's the modern day slavery. Commented May 13 at 15:05
  • 6
    @VanitySlug-codidact.com Please don't compare these things to slavery; slavery is absolutely terrible, nearly beyond belief. Second, OpenAI's billion-dollar LLMs have already been trained on just about everything humans have ever produced as long as them or someone else put it on the internet at some point. SO content and everything else, with nothing at all to do with this partnership. If anything, this partnership is an attempt to claw back something in return, even if it just means continuing to host the content in the format it was originally provided. Commented May 13 at 21:06
  • 4
    @ResistanceIsFutile "And the partnership announcement would probably be much better received if it would actually have some more details about the deal..." Probably yes, although it would depend a bit on how bad the details really are. Some details have been posted now: meta.stackexchange.com/a/399897/260073 Commented May 14 at 20:08
  • 3
    Personally I take issue with any partnership with OpenAI, because in the grand scheme of things I don't think this company is a positive contribution to the world. I think it is a bubble that diverts funding that could have been used more efficiently, and that has a negative impact on the environment. I'm aware that ultimately the absence of partnership won't prevent them from reusing content I've posted on SE websites if they want to, but it doesn't mean I'm OK with making things easy for them.
    – J-J-J
    Commented May 15 at 10:04
  • 1
    Bruce Schneier points out that LLM's disturb the connection between writer and audience in today's newsletter. "(...) LLMs synthesize their own answers, treating content [...] as digestible raw material. Writers and other creators risk losing the connection they have to their audience, as well as compensation for their work." Commented May 15 at 11:08
  • @J-J-J it’s possible - and maybe I’m being too generous - that the agreement is actually there not just to help them but also to keep them honest. Commented May 15 at 11:13
  • 2
    @testing-ma-lady My issue is not with them possibly not respecting attribution and the CC license, but with the nature of the company. However, in the link you give you say that they could no longer use ignorance of the CC license as a legal defense, but ignorantia juris non excusat, so I doubt that it would be really useful relatively to that. But specialized lawyers might have a different opinion, I guess, as it's not my domain of expertise. To me, that's a secondary point anyway.
    – J-J-J
    Commented May 15 at 13:18
  • @J-J-J well, they were doing it before, with apparently only ignorance on their side. So the agreement might be at least in part an attempt to make them agree to do better. Regardless of legal culpability. Commented May 15 at 13:24
  • @S.L.Barthisoncodidact.com And Bruce Schneider is not even the first to discuss this. The World is basically divided into two camps: the no AI shall come near me camp and the resistance is futile, hail the AI overlords camp. Finding a sensible and helpful middleground will be extremely difficult. As a content creator, I would simply stop working under the current conditions until it's settled that using my work for AI applications isn't fair use. And well, that will probably take a while. Commented May 16 at 20:48
  • 3
    arstechnica.com/information-technology/2024/05/… is an interesting read. The lack of actual information on what SO inc agreed to does concern me, and I wonder if keeping that confidential is part of the boilerplate OpenAI has now Commented Jun 1 at 3:07

10 Answers 10

54

In a sense—it's because both AI companies and to an extent the company haven't acted in a manner that engenders trust.

I'd note our first brush with AI was people posting AI output on our sites with little concern for correctness or quality. The network sites to some extent banned it, then... the company went from "eh, whatever" to creating a very strange/onerous set of rules that the moderators could not realistically hold up.

A result of this was anything AI was sus.

I see it as something overhyped and potentially grifty. Some people see it as a potential evil that threatens jobs, creative pursuits and such. Other see it as "the future" though sometimes a little cultishly.

It is, as such, very easy to cause people to be outraged.

We don't have clarity over what the agreements mean yet, though we're told they'll ensure attribution. I have a suspicion it might mean prioritisation with Google's current search engine and the rumoured OpenAI's "It's not a search engine, honest!" product in return for attribution. There's an API apparently but, once again, there's no clarity over what's so special about the OverflowAPI.

We're not really getting treated as partners, nor do we really have a seat at the table. This makes damage control rather difficult. What is easy is manufacturing outrage, since folks already mistrustful to an extent.

It's worth considering why people mistrust these organisations

Historically, OpenAI's used datasets that were essentially dumps of a famous pirated ebook sites (which they now admit), and have been sued. They pivoted from being a nonprofit to, well, being a very profit-oriented organisation.

Google's "Do no Evil" is a memory of the past, as they've pivoted from Anakin Skywalker, the cool organisation that built neat things and did non-obtrusive ads, to the Darth Vader of advertising. I've not really looked into what Gemini's training materials are.

SO Inc literally provoked a moderator strike, because amongst other things, someone decided to stop the data dumps while working out how best to monetise them. I think there's probably a whole SE's site worth of situations where the company has acted in not in our best interests, and sometimes theirs.

None of these organisations have earned a position of trust. Some of them even actively squandered said trust.

With Stack Exchange/Stack Overflow Inc, in theory, their best interests might align with ours—since genAI providers need/want/are hungry for fresh data to feed their models, and being able to tap trusted human sources for information would be valuable. SE gains visibility, since the openAI or google SGE output using SE data sources will cite back, and incidentally this would meet the requirements for a CC share alike licence.

Suddenly the community and its output is a sellable resource, when for years the focus of the company was in providing services like Careers and teams which tapped on the name recognition and community, but weren't directly reliant on the success of the community/Q&A side of things.

I'm not sure if we're now the crown jewel of the network, getting headpats and love, or if we're getting looked at sideways with hungry eyes yet, to be carved up like a Sunday roast. The effects of these agreements and the fact that our site data is now sellable is unclear.

There's a certain pattern here. People are unsure, expect the worst (because we've had to deal with the worst) and there's not really any clarity or reassurance that things will get better that we can trust.

I don't trust OpenAI or Google but I see benefits to a contractual framework for appropriate use, and the theoretical good that our works having direct value to the company could have for community resources.

Much of this is speculative. There's a ton of citations and footnotes this could use. But practically we can't really trust any of the actors to act to our benefit outside their own needs.

0
18

If you look at Microsoft page for Responsible AI usage you will find a pretty good image that presents six keys principles in working with AI-related technology.

Enter image description here

Disregarding Fairness and Inclusiveness, I would advocate that so far the initiative not only doesn't meet the other three principles, it even goes way into deliberately avoiding to address them.

First, I will advocate that relying on AI generated text does not satisfy the Reliability and Safety principle.

To build trust, it's critical that AI systems operate reliably, safely, and consistently.

Yes, I know this is pretty opinion based and every user will have a different view, but in my humble opinion the results you may get from asking a question to a chat bot are surely not consistent and often not reliable. It should not be surprising that most generative text solution include some sort of "warning" than users should check the results themselves. It seems quite a Catch 22 scenario to be honest: I can use ChatGPT to summarize some document, but then I have to read the document and the summary to check that the bot didn't hallucinate content.

Some days ago I have made a simple experiment with the collaboration of Journeyman Geek, asking Google a very specific question to force the AI answers to reference a post I made on the network. The full discussion is available on chat.

Enter image description here

I will only point out for example that in the above screen, not only did the generated answer introduce an irrelevant line that I didn't say (the question is not about skipping time by resting, it is about sequence breaking the game quest line), it also partially misrepresented my intent in the line about "consequences".

That said, if the issue was just related to the correctness of the results I guess most will probably be fine with it as long as there was a clear separation between generated content and human curated answers. Therefore, while I mentioned these concerns for the sake of completeness, I will instead focus on the two remaining principles.

Microsoft defines two more principles: Transparency and Accountability.

Transparency means that the behavior of the system should be explainable and interpretable. Accountability means that who creates the AI systems must be accountable for how their systems operate.

By continuously avoiding to answer the relevant questions the users ask about how the system will operate, Stack Exchange is surely not Transparent and is avoiding Accountability for it own actions.

I don't care at all if "the CM/employees probably are not allowed to post answers that may have legal relevance": if that is the case, then I am pretty sure that someone else with the right "clearance" may post it. Or even better, the company could finally take that... 6-8 years old suggestion and create a shared "Stack Exchange Official" account to post announcements and other official answers (this coincidentally would also put a stop to claims of employees feeling personally attacked by answers posted at the company).

Just in the last few weeks:

I'll stop here but fell free to edit the post if you want to add more examples of ignored concerns.

Anyway, I think the issue should be clear.

  • Generated content can hallucinate wrong info, and users do not like to be associated to things they never actually said.
  • That is, if the LLM will provide some form of attribution in the first place.
  • LLM and image generation models are quite famous for ignoring licensing law in the shade of the legal hole we currently have. Do you really believe that someone paid your usual "by WLOP, Artgerm, etc." for the pictures that Stable Diffusion must have used in its training?
  • Users do not exactly want to provide more ways to Google&Co to cross-reference their PII
  • In all of this mess, Stack Exchange instead of behaving responsible and stepping in to answer users concerns, once again is playing"three wise monkeys" and hiding its head under the sand like an ostrich, posting just a empty buzzword filled useless piece of announcement that discloses a "Responsible OpenAI partnership" without explaining how it will be responsible, what the users should expect or even what the partnership entails in the first place

There is an old saying that claims that when someone fears that answering questions can be bad for them, then they probably don't feel confident that their actions would be welcomed.

4
  • 4
    "if that is the case, then I am pretty sure that someone else with the right "clearance" may post it" I don't think those people ("C suite") know how to use the sites - they don't seem familiar with the company's products at all or perhaps distrust those products.
    – Lundin
    Commented May 13 at 8:45
  • The company has answered Peter's addition to my question, so that one hasn't been completely ignored.
    – wizzwizz4
    Commented May 15 at 10:18
  • @wizzwizz4 I see one answer posted 17 hours ago on your post. My answer is 2 day old, so it was posted before that. Do I miss anything else? I will update the post. Commented May 15 at 11:37
  • I'd note since you posted this post, and since I poked around, the sheer comical badness of google's AI infused search has become a meme Commented May 31 at 23:07
15

For those of us who have been around since the beginning, this move towards AI integration is just one more move where the needs of the community that built Stack Overflow into what it is are being pushed aside for the desires of the company.

It's not as if the evidence wasn't in front of us the whole time That Stack Overflow not only 'could' do this, but would do this -- it was, we just chose to ignore it -- and that at least fuels part of the anger.

Stack Overflow was ostensibly built to 'get big, fast'. This was a stated aim by Joel. Reading between the lines I've gotten the impression that Jeff was less than comfortable with this once he saw the popularity and power of Stack Overflow. It seemed more "this is too much power for one man" than "Hey let's get rich off this rocket ship" for Jeff. Again, this is my view from the cheap seats. So, as this is part of the reason in the head-canon that I have for Stack Overflow, Jeff left.

Once that happened, there was an inflection point: The person who had power and had grown attached to the thing that Stack Overflow was, was gone. As a parent, I know I could never abandon my baby, but watching Jeff interact with Stack Overflow, it appears it had become his child, and that he was forced to abandon it because it wasn't going in the direction he wanted to go in, and he was rather attached to it. I get that, and that must have been emotionally traumatic for a time.

With Jeff gone and any mis-alignment in stated aims gone with him, Stack Overflow began to look for ways to become that rocket ship. We helpfully plodded along, building the corpus of useful programming knowledge, and getting better at this whole thing while we were at it. We got better at editing questions and answers, re-inforcing good questions, downvoting bad questions, seeing the patterns of misuse and abuse, and overall communicating with faceless folks over the internet to improve this thing we all had a hand in building.

In the meantime, Stack Overflow the company was really in a crisis -- they had taken on VC money and kept taking on VC money to keep the (rather nice) NYC office and Stack Overflow the Machine running, but they didn't have a breakthrough rocket-ship trajectory product. That is not a great feeling if you're VC-backed -- as the entire reason for being VC backed is to grow quickly, exit quickly, and get outsized returns for the fund, VCs, the founders (hopefully) and the employees (it'd be nice, but not essential).

So they took on more money and kept trying different things. It was all rather scattershot, but for us that were willing first customers, we were there for it all, and when each didn't pan out to have the rocket-ship trajectory needed to satisfy the VC hunger for returns, it was killed. This hurt us, because we didn't see those events, we only saw things we liked and thought could be improved being unceremoniously dumped without a good (to us) explanation.

Finally, (and I do think of it as bag-holding, but I recognize that's a charged term), Stack Overflow was sold to Prosus (which honestly I really thought Microsoft would be the buyer, and I'm still surprised they didn't), and Joel et. al., were able to drop the bag in someone else's lap and cash out. That was the exit the VCs needed. Now, Prosus has this thing in their laps that they didn't build, and don't understand the culture of, and they have to make money off of it, or at least boost its usefulness in the eyes of another buyer enough to get them to hold the bag.

And here we are at present day.

This AI partnership is just one more piece of evidence that the contribution of the folks that have spent literal years on this network, building it, curating it, shaping it into something useful for other humans on the internet, is being cast aside for the pursuit of money and personal gain, and our work is being directly monetized not just for Prosus, but for any company that finds a way to optimize for what OpenAI provides, and that's not what any of us signed up for. We didn't sign up for nefarious folks to use our knowledge to further their goals, and one of the distinct problems with generative AI is that it commoditizes the act of distilling large tracts of information down into something that can be retrieved in seconds and put to purposes we hadn't thought possible before now.

We've come a long way from the idea that Stack Overflow is you.

1
  • 3
    Certainly the idea that Stack Overflow is you is gone now. Hope you've been well George, I think a lot of the people who put in the work to build this feel the same way.
    – Travis J
    Commented May 31 at 20:51
15

I'm going to twist the intended audience here, tackle one subset of the reasons, and go on a bit of a (hopefully useful) tangent. A lot of the people I see unhappy seem to be unhappy for reasons that don't make much sense to me. I want to take this as an opportunity to clear up some misconceptions. (it hopefully goes without saying that if you don't have those misconceptions, this section of this post is not written to you).

  • This partnership does not mean your content is now going to start being used by OpenAI. OpenAI (and every other group working on LLMs) probably was already using content on Stack Exchange (whether or not their use of it was/is legal is another matter that I don't have the expertise or background knowledge to go into*) way before this partnership. Whether by web scraping, or use of the data dumps, or the API. Yes- there are data dumps and an API (way before "OverflowAPI" existed). Many people don't seem to know this. Scroll down and look in the site footer. The links are right there.

  • This partnership does not mean your content is now going to start being commercially exploited by SO Inc. By participating in the Stack Exchange network and contributing content to it, you (whether you knew it or not- at least according to SO Inc.-) agreed to the Terms of Service, in which you license SO Inc. to commercially exploit your content.

  • That people other than SO Inc. can use your CC-BY-SA content and profit from it is not new. CC-BY-SA does not restrict what purpose the licensed content can be used for. I get some sense that many people don't even know or think about what license their content here is under.

  • That you (potentially) got a temporary suspension if you vandalized your own content in protest is not what you probably think it is. Off-site coverage of this seems to even think these suspensions are being dealt out by the company, and that they are some kind of huge inconsistency with banning AI-generated content. That vandalism of content is not allowed here is not new. These suspensions are being dealt out by your elected site moderators (who- at least on Stack Overflow- were the very people who initiated the AI-generated content ban) to protect the very resource that you care about. The suspensions are to give you time to cool down and stop harming the site. Adding work to their plates (the people who do also have on their plates to deal with all the other issues you care about, like violations of per-site AI-generated content policy violations) is not helpful. See also How can a user unhappy with the LLM partnerships protest constructively (and ideally- effectively)?.

  • Some people seem to be making assumptions about what impacts this will have on this platform- Ex. assuming that this will mean AI-generated content will start being systematically (as in- by the system- without human prompting or involvement) posted. It's totally valid to be concerned about that possibility (I have been for a year now), but no specific details have been posted, so the assumptions are unfounded.

I'm not saying that it's not valid to be unhappy about those first three things (who am I to do that) (and I hope that after clearing up the fourth point, people no longer find it a point of issue). I'm just surprised that so many people are just starting to think about this stuff now. Though I guess that's a good thing.


Misconceptions aside, continuing on the point about lack of detail in the announcement, at a meta level, people are unhappy that the announcement was rather poor. It had no concrete details, and primarily linked to an external press release (technically an off-site resource). If you're rather new to meta, welcome. Unfortunately, at least in the past 1-2 years, this is not new behaviour for the community management team, which is supposed to be the dedicated link between the company and Stack Exchange userbase. More info has been requested at What, exactly, is Stack Overflow's agreement with OpenAI?. Much further back, see also Is SE [going to be] selling our content for AI model training? And what exactly does "reinvest back into our communities" mean?.

This is just another instance in a large pool of company actions (or omissions of action) that demonstrate how little the company's interests / priorities align with ours, and how little they care to communicate with us in a way that that is culturally sensitive. There are exceptions to that pattern, but feelings of dissatisfaction in that are well-founded.

A lot of people also dispute the company's claim that OpenAI is socially responsible (unethical use of content, environmental damages, and other ethics issues like what kinds of content they allow their models to generate).


* Concerning the legal questions of LLMs using SE / CC-BY-SA content, see also:

8
  • 2
    “you (whether you knew it or not) agreed to the Terms of Service” isn't how legal agreements work.
    – wizzwizz4
    Commented May 12 at 22:32
  • 1
    related to above comment: some chat conversation
    – starball
    Commented May 12 at 22:50
  • 3
    Yeah the site has no doubt been scraped countless of times by various shady 3rd parties. But then arguably SO is not to blame; it's not their fault that the Internet is filled with shady people doing shady things. However, if you explicitly provide an official API then you say that you are fine with the content getting scraped and possibly even without attribution. Then they have shifted from a victim of actions carried out by some 3rd party, to the party actively responsible for the scraping. That's a shift in ethics and possibly also with legal consequences (actively violating licensing).
    – Lundin
    Commented May 13 at 8:57
  • 3
    So it matters for the sake of SO the company's credibility. Which was already zero before this and can't get any lower - anyone still trying to communicate with them and trying to talk some sense into them is the very definition of a naive idealist.
    – Lundin
    Commented May 13 at 9:00
  • 2
    @Lundin the API is consistent with the existing subscriber content licenses. I don't understand what you're trying to say. Feel free to ping me in the tavern. I want to avoid lengthy discussion in comments
    – starball
    Commented May 13 at 9:06
  • 3
    The dual license to SO in the TOS is kinda restricted to "as reasonably necessary to," with a bunch of exemples around technical issues/law enforcement. I don't think commercially selling the data (even deleted posts) match the "as reasonably necessary" part of this. So I wouldn't be so sure it does really waive the concern some people have.
    – Tensibai
    Commented May 13 at 14:56
  • 7
    I think this is a phenomenal summary of (and clarification to) the misconceptions surrounding this whole debacle. I've been frustrated to see, what feels like to me, so much wasted breath on misdirected anger. I just don't think it's valid to get mad at SE yet for "not caring about attribution" in some form, before they've actually done much (?) to show that they don't. The announcement communication was atrocious, but they have said the words; folks are more than welcome to not believe them, but that's decidedly different than SE actually acting against them.
    – zcoop98
    Commented May 13 at 19:07
  • 1
    Others using content for personal gain has kind of been the entire goal of Stack Overflow...and it always came with attribution. Can you imagine just taking whole source code from GitHub or ReferenceSource, removing all of the licensing and attribution, and just using it because you were trained on it? Then once used, regardless of how much was used, claiming it at as your own? Audacious to say the least, although perhaps not entirely uncommon given the prevalence for intellectual theft right now. On top of that, it is now not only being offered as a service, but toted to authors as a benefit.
    – Travis J
    Commented May 31 at 20:58
12

If you compare the recent announcement about the partnership with Google with this new announcement about the partnership with OpenAI, the scores are similarly negative but the community outrage/objections were a bit less for Google than for OpenAI. People seem to dislike OpenAI especially much and simply don't want StackOverflow to have a partnership with them.

Then, both these announcements got a lot of negative feedback because they were formulated in a very non-informative way. The company didn't explain anything, leaving us in the dark, open to speculation ("From skimming over the announcement, SO is going to just sell access to user content for training ML models." is only a speculation) and generally a waste of time. People may be especially unhappy about the communication with the company and that might partly overshadow the like/dislike for the partnerships.

Most people speculate that it might be about selling content to these companies so that these can further develop their AI products. If I read the room correctly, many people simply don't want to contribute to AI products of big companies. Maybe they feel like they are not getting enough (nothing to be exact) out from this bargain? Or they have ethical objections or fear of non-human competition.

I think the main difference between what you say and what others say is

..so this looks like it doesn't change anything.

People seem to disagree. For them it makes a difference. If StackOverflow formalizes the partnership with these big companies, it basically went over to the AI side (it was on the AI side from the beginning actually). Maybe people feel that if even this company capitulates, nobody will stand up for them.

And I believe that the partnership really makes a difference. The company wouldn't agree to them if that weren't the case. There must be money in it for StackOverflow and that in turn means that there is something else in it for Google and OpenAI (maybe legal safety, see Can SE just resell our data, relicense it and remove the attribution requirement?). These partnerships are not meaningless. I guess that in general it comes down to additional support for Google or OpenAI and people simply don't want that.

11

Personally, with regards to this announcement, Google before it, and their other AI-related product announcements, it's just too... fake. Empty promises, marketing speak/dancing hotdogs on how this is somehow the future, and socially responsible, and a boon for the community at large... and yet nothing that comes from it has ever helped the community. It's all gone into Stack Overflow for Teams (SOfT). Into selling access to content that we freely contributed to companies that won't similarly contribute to our goals.

First we got the content assistant that was supposed to help people write questions... but it was just a direct prompt into an LLM and instead made questions worse (this is now a feature of SOfT Enterprise). Then we got the editing assistant that was tested secretly on the public SO platform directly before we realized it and called it out on meta (this is now a feature of SOfT Enterprise). Then we got the OverflowAI Search tool that was invite only that I never got an invite to (this is now a feature of SOfT Enterprise). Are we just a testing ground and data source for their enterprise products?

Supposedly some of this income from the AI initiative is going to be reinvested into the community, but all I'm seeing is reinvestment into SOfT. Where's the SG? Where's our improved ask wizard? When will the inbox bugs be fixed? Chat? Review queues? Maybe CM's that actually work with the public platform community (like we used to have)?

3

Why are people so unhappy about the OpenAI partnership? Am I missing something?

Here are some reasons that I've seen:

  • Lack of attribution.
  • Feeling of license violation.
  • Don't want one's work to help a corp one doesn't like.
  • Realizing that AI can replace their work.

(The foregoing only relates what I've seen, not my opinion. Imho OpenAI made AI breakthroughs with long-term positive societal impacts.)

9
  • 1
    If someone following the original X/Twitter/Mastodon thread, then point 3 about "a corp one doesn't like" could be expanded far further. Commented May 13 at 5:36
  • 2
    @MetaAndrewT. I haven't. What's going on there? Commented May 13 at 5:40
  • 1
    These points are so obvious I don’t know what constructive usefulness they add to a deeper discussion, or validation of them. Commented May 13 at 11:59
  • 8
    It answers OP's question Commented May 13 at 15:02
  • They save net time? [citation needed] Commented May 15 at 2:18
  • @DavidRoberts They use net time to save human time (FD, 2024) Commented May 15 at 5:22
  • @FranckDernoncourt Your citation is the archive.is copy of this very MSE page? Is this some kind of recursion joke? (Maybe I'll need some more coffee to get it....) Commented May 15 at 7:13
  • 1
    @S.L.Barthisoncodidact.com it's a "because I said so" reference (also "[Inter]net time" ho ho) Commented May 15 at 7:42
  • Aren't training generational AIs considered transformative use and hence fair use? And you don't need consent of the copyright owner if it's fair use?
    – Poscat
    Commented May 15 at 13:11
3

Overall, the people unhappy are probably the largest contributors. If you were to highlight a group of users who had only a handful of posts, versus a group of users who had thousands of posts and years of time dedicated, you would certainly see a difference of opinion.

I am personally dismayed by the path that the company has been going in the past few years.

A collective effort was spent here by the community to build a body of knowledge that has been a pinnacle of information for the broader community of developers. During that time, assurances were made that Stack Overflow would remain a partner in building this body of knowledge. At first, it was. Now, it isn't. It really is that simple.

Stack Overflow sold out the public platform for a private enterprise version. This had a multitude of side effects, but rest assured, not a single post from enterprise was used in the training of any LLM's. Why? Because it is behind a privatized paywall. Something that was the antithesis of the platform at its inception.

Each step into this bastardized version of scraping known currently as GPT is just one more step away from the community. Anything produced here will just be parted out into a weighted token and used regardless of any agreement in place.

I'm sure all of this just seems abstract. In a place where we all just regurgitated news or gave sports opinions, it would be. We aren't there. We are in the forefront of software development where the newest manual's edge cases can be found. Often, there is only a few articles on the internet as a whole on bleeding edge technology like this.

In those cases, the cases which Stack Overflow content predominantly falls, tokenizing doesn't mean crap. Weights don't mean crap. There is only a handful of references and they will be mixed either verbatim or close to it. This is the main qualm I have. Verbatim plagiarism, plain and simple. It compounds though. If someone else in the space that I am referencing has posted similar content, then that content may be remixed. However, without any sort of reference to what constitutes significance in the tokens, the content can be partially shuffled (#hallucinating). The problem there is that if it is only from a handful of sources, then a majority of the generated content will still be googleable, and may improperly imply attribution of potentially very harmful solutions.

To reiterate, or if you are old school... tl;dr;

  • Stack Overflow's current efforts in AI to offload content for profit further the gap between community and company
  • Content scraped/epoched from Stack Overflow has a high likelihood of being plagiarized
  • Content scraped/epoched from Stack Overflow has a high likelihood of producing damaging solutions in Software for edge cases
4
  • Their data is private and their own, yes. I am unsure what the license looks like, but I would assume it is starkly different than public Stack Overflow. The feature is behind a paywall. Perhaps my wording could have been better there. A founding principle of Stack Overflow was to depart from the notion of Experts Exchange and having areas hidden by paywall. Perhaps that meant content, or perhaps it meant access.
    – Travis J
    Commented May 31 at 19:12
  • @testing-ma-lady: Yep, that's basically the purpose of the Business/Enterprise-tier Teams products – to serve as an internal Q&A platform for a company, so that company-internal knowledge is documented and retained. (SO has its own instance as well, for basically the same reason.)
    – V2Blast
    Commented May 31 at 20:08
  • 1
    @testing-ma-lady - This is clearly getting into a tangent, and I do not need to discuss a hypothetical manufacturing company at this point. Creating Teams as a feature of Stack Overflow was possible due to the success of the knowledge base and user base that built it. Was that community best served by creating a private section of the site? No. In fact, a majority of support staff for the community was removed at the same time as well. Does that represent a departure of outlook for Stack Overflow the company in moving more towards monetized products and less towards community products? Yes.
    – Travis J
    Commented May 31 at 20:42
  • 1
    What follows that divergence of outlook has been a slew of for profit decisions that have been at odds with the interests of the community (Kevin B's post highlights this as well). The recent actions of the company with regards to AI are simply an iteration of departure from the original outlook, so it is relevant to highlight the overall path that was taken.
    – Travis J
    Commented May 31 at 20:49
2

I think people are unhappy because their information is being used to train models. So, Stack Overflow is just getting free money from all the hard work and research we put into our answers, which just ends up getting plagiarized.

3
  • SO Inc has been making money for a decade kn the back of volunteer hard work. Complaining about it now just because the AI bogeyman got involved is a bit naive and nonsensical.
    – Nij
    Commented Jun 1 at 11:27
  • @Nij so you're saying all other 9 people who answered this question are naive and nonsensical?
    – Andy K
    Commented Jun 1 at 19:32
  • No, just this particular complaint and those like it.
    – Nij
    Commented Jun 1 at 21:22
-2

I’ve thought of a (possibly) good analogy for the partnership- and why people aren’t too happy.

Now imagine that the Stack Exchange network server farm is a local, community garden, for a small town.

The members or our communities go to the garden and plant their plants. The plants are all for sharing, and as long as you don’t take too much, you can take other people’s plants.

The garden is free and the moderators make sure no one is greedy, and remove the weeds and odd cannabis plants!

Everyone is overall content and happy to share -reputation- the more you plant the more you can take for yourself.

People have different specialties, like sites and tags, John is good at growing tomatoes, (and software coding,) but Carol is good at growing lettuce, (and hardware fixing.)

They help each other and share.

But then one day things change.

From the big smoke- a company comes and decides to make a profit out of the communities charity, community garden. The company, called OpenVege, harvests the communities hard work- without even asking. The herbs and vegetables are still there- but the self sufficient community now feels violated and betrayed!

If you didn’t get the hint, this is basically what happened, on a bigger scale, to the SE network.

P.S, this took me a while to write so I hope you like it!

6
  • 2
    The converstion from community plots, to corporate megafarm started a very long time ago :( Commented May 31 at 23:48
  • 6
    I don't get the "don't take too much" part of the analogy... or the "rep means you can take more". there's no limit to how much knowledge you can take from SE content except how much is actually there. the restriction is the CC-BY-SA licence. and the CC-BY-SA licence does not place restrictions on who can use or distribute the content, or how much they can use or distribute, or for what purposes they use it, or whether they profit from it. I'd like to say something positive but I don't see how this analogy is representative or useful.
    – starball
    Commented Jun 1 at 0:23
  • @JourneymanGeek I agree! Haha yes- it is just an analogy! Commented Jun 1 at 0:29
  • @super-starball-ultra it is just an analogy- if you dislike it that is fine. Commented Jun 1 at 0:30
  • 2
    "Now imagine that the Stack Exchange network server farm is a local, community garden, for a small town." That’s a nice story but SE has been a business since ages. This isn’t a tragedy of the commons, this is people being oblivious to what they signed up for. Commented Jun 1 at 8:56
  • @MisterMiyagi yes I know! Haha not exactly a garden! Commented Jun 1 at 9:39

You must log in to answer this question.

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