Our partnership with Google and commitment to socially responsible AI

So what exactly do we get for curating a multi-billion dollar corporation's data set for them, other than they will make us something cool that we can pay them for?

As of March 2024 Alphabet (Google) has a market cap of $1.662 Trillion. They could afford to pay people to curate their model data if they refocused their sprawling HR bureaucracy.

I hate to be that guy but... most of the other pieces of news about social software companies selling data to AI companies talked about eye-wateringly high amounts of money getting paid.

I'm hoping there's a financial or otherwise incentive here - and some of that goes back to the community in terms of support.

And while I see someone asked about money - I'll ask, does at least some of this include ensuring that tech debt for the foundational software behind Stack exchange gets paid off, and that resources are available for the community that generates the high quality Q&A, considering the recent losses of key staff?

Practically - what's in it for us? What's here that the folks who create, curate and occasionally cheerlead have to be excited about this announcement?

I noticed that the robots.txt for all network sites has been updated. Google-Extended is no longer blocked. Looking at Internet Archive Wayback Machine captures, the last time Google-Extended was blocked on Stack Overflow was Tue, 02 Jan 2024 19:07:37 GMT. By Tue, 02 Jan 2024 21:41:45 GMT, it was no longer blocked in robots.txt.

Previously, I raised concerns about the process by which bots and crawlers are blocked and unblocked, especially since it's generally done silently - without discussion or announcement.

So, this brings up several questions:

• Does this agreement with Google include terms preventing others from crawling SE data for AI training? Is this an exclusive deal or is the company also open to other agreements with other companies to give access to SE data for AI training?
• If no, why are OpenAI and Amazon blocked from crawling and training on SE data? If yes, what is the duration of this agreement and why were all of these bots blocked much earlier? See the earlier question on the process for making decisions on blocking and unblocking crawlers and other bots.
• Are there updates on the "guardrails" regarding the Data Explorer and the data dumps? This was announced previously and there were allusions to restrictions on who could use this data and how they could use it, especially in AI model development.
• What is SE's commitment to ensuring our contributions are available to the broadest possible extent under the Creative Commons BY-SA license?

I do want to be clear: SE does have every right to control who can access their platform and how they can access it, including placing limits on crawlers, API usage, and more. However, the content being restricted is composed of the effort of countless people to create and curate. I firmly believe that contributors should have input into and visibility into how access to our contributions is controlled and how they are monetized by the company.

The story of this partnership started about 16 years ago - with you, the community that made Stack Overflow what it is: the internet’s predominant resource for high-quality curated technical Q&A.

This is approximately as tone-deaf as rolling up to a meeting of Georgists and opening with "the story of Monopoly's commercial success started about 120 years ago, with Elizabeth Magie and her passionate critique of rent-seeking."

AI is antithetical to our understanding of quality, curation and Q&A. Nobody involved at the beginning could have predicted what AI would look like in 2024. Users didn't consent to this, either. You can say that the content license does not require such consent, but this misses the point entirely.

This resource is ubiquitous and invaluable today, but it took countless hours of hard work and dedication from many community members to get it to this point.

Then why is the company treating it so flippantly? Why is the company advertising our own content (as repackaged by this venture) back to us, while doing nothing to resolve or even acknowledge countless problems pointed out by the userbase that impede human interaction with the system - many of them nearing a decade old if not older?

What reason can you offer as to why a sensible programming expert should choose Stack Overflow as a platform for sharing knowledge in 2024, when personal blogging is easier than ever and numerous alternatives to the Stack Exchange Q&A model exist where the staff actually respect and listen to their users?

Does "strategic partnership" just mean cooperation, or is there actual money flowing from Google to Stack Overflow?

It was a bit hard extracting this information from the marketing fluff.

As such, questions on Stack Overflow (whether by a community member or assisted and curated by AI) are posted only after human review.

The blog post mentions in parentheses here the possibility of AI-assisted or curated questions. Does this indicate a change in policy regarding the prohibition of AI-generated content that is currently in effect on Stack Overflow?

https://stackoverflow.co/company/press/archive/google-cloud-strategic-gen-ai-partnership:

In addition, Stack Overflow plans to leverage Google Cloud’s state-of-the-art AI capabilities to improve their community engagement experiences and content curation processes.

What are the planned features for SO, and what's their release timeline?

The phrases "partnership with Google" and "socially responsible" are mutually exclusive.

Google (or Alphabet) is a corporation engaged in mass surveillance of the public, including what should be private communications, not only for psychologically manipulating people into buying things, but also in the service of government intelligence agencies. It also engages in political censorship or semi-censorship of online content in web page, video and other search results - by different rules in different world states, including the US of course.

Several other answers point out the details of how an SE-Inc-Google deal is indeed problematic and questionable, but one could have just safely assumed such details were to be found somewhere, given the general principle.

Our mission is to set new standards with vetted, trusted, and accurate data that will be the foundation on which technology solutions are built and delivered to our users.

Where will this "vetted, trusted, and accurate data" come from?

While that description does apply to some highly curated parts of SO, most parts are more accurately described as a garbage heap. I somehow doubt that the company is suddenly changing track from the "more users are more better, quality is an afterthought" mantra of the last years to "let's ensure we only have high-quality contributions", and filtering out all the accumulated garbage seems improbable as well (keep in mind lots of it has a positive score).

From here:

In addition, developers using Gemini for Google Cloud will be able to access Stack Overflow directly from the Google Cloud console, bringing them greater access to information so they can ask questions and get helpful answers from the Stack Overflow community in the same environment where they already access Google Cloud developer services and manage cloud applications and infrastructure.

Does that mean that you plan to allow posting questions through a 3rd party interface? Potentially bypassing the guidance and quality filters we have in the ask wizard on the website?

What is the "added value" of this, if any? I mean to the communities that actually authored the content, not to the receiver of the Google dollar?

The content of this site represents countless years of experience, condensed into carefully worded answers to specific questions. There is already a mechanism to search out appropriate answers to questions, and it works.

The considerable risk of this idea is that our content, correctly perceived to be of very high value, becomes cattle-feed for a badly thought out tech experiment, and might actually end up doing more harm than good due to the shortcomings of the experiment.

This is NOT why I chose to spend time writing answers that share my technical knowledge here. I suspect that I am not alone in that sentiment.

No doubt when I joined the site there I ticked a box agreeing to some carefully worded legalise, and even if I didn't there would be little I could do. But this feels to me very much like an abuse of trust and good faith, and I suspect that I am also not alone in that.

Does this deal include the content of the public Q&A sites under a non-CC license?

Too funny.

Duplicity:

Reminder: Answers generated by artificial intelligence tools are not allowed on Stack Overflow. Learn more

Stack Overflow plans to leverage Google Cloud’s state-of-the-art AI capabilities to improve their community engagement experiences and content curation processes.

Reminds me of SO policy of no politics on SO, then SO ownership and management decided to fly their political banner colors in the SO logo.

The model of "A.I.", as I see it, is a marketing racket.

People provide all of their own data to an algorithm humans create, then humans tailor the outcome to suit the consumer of their own data regurgitated back to them. Don't like the answer, no problem, the human will adjust the algorithm to output the answers that suit your fancy.

The Register seems to think this means SE will be charging for API access.

They have a very sensationalist headline reading "Stack Overflow to charge LLM developers for access to its coding content" Though there's not really much from the actual press release, and more from previous statements backing it up.

Assuming they're incorrect - since the article seems to be a mix of past statements and this, can we confirm/clarify that there's no change to access to APIs / Data dumps and if so, what any changes related to these might be?

And as much 'love' as we all have for the ol register, maybe get someone in touch with them to clarify.

It isn't clear what happened between now and 11 months ago. Specifically, what changed since the accepted answer (official company post) to this question Is SE [going to be] selling our content for AI model training? And what exactly does "reinvest back into our communities" mean? excerpt:

The money that we raise from charging these huge companies that have billions of dollars on their balance sheet will be used for projects that directly benefit the community. ... We may need to tighten up access controls to prevent abuse, but there will be a method for community members to access the api and its data for their own use.

But the community is - you are - being denied your rightful attribution as it stands right now. Prashanth is saying that you should at least get to benefit from the financial impact. This is about protecting your interest in the content that you have created.

I thought that Stack Overflow content was sold to Google then, in April 2023. Note the caveat about there being

a method for community members to access the API and its data for their use

which wouldn't need to be mentioned if licensing and/or ownership rights had not changed in some manner.

Perhaps the financial and other terms of the transaction weren't completed until now, in early March 2024?

Perhaps access to Stack Overflow content was sold to Google in April 2023, but not exclusive access? This conjecture is supported by the plural form ("huge companies") used when stating that--as protection of our interest in the content we created--we will be benefiting as a community from:

money that we [Stack Exchange management] raise from charging these huge companies that have billions of dollars...

Perhaps the current "exciting new announcement" by OP is that Google now has exclusive access to the Stack Overflow "repository" of content. That seems necessary for the following, per OP:

to provide Stack Overflow content directly within Google Cloud.

If an additional granting of (exclusive or other) rights occurred, then a second financial transaction was likely completed between Google and Stack Exchange management. It remains unclear how the infusions of funds, whether in April 2023, March 2024, or both are realized as benefits to us as the community.

An AI-related partnership with Google AND an independent commitment to socially responsible AI isn't possible. This isn't casting aspersions upon Google, but rather a consequence of the following:

1. Alphabet has AI principles which are reflected in the business strategy and standards of its wholly-owned subsidiary, Google.
2. A partnership between Google and Stack Exchange will result in Stack Overflow having de facto the same policies and level of commitment to socially responsible AI that Google does.

Alphabet makes AI policy recommendations for governments worldwide. Given that vast scope, I don't think that Stack Exchange is likely to be broken out separately.

For what it's worth: I've had enough. I'm deleting my profiles.

Edit (May 9) after the months that have passed, and in light of the announcement that SO has now also partnered with OpenAI, I don't fully stand behind this answer anymore. Scroll down for more.

Original post

I'm not happy about this announcement, particularly the fluffy buzzy optimism.

Yet I won't join the general criticism, and indeed say that the actual strategic decision Stack Overflow made here is sensible. Perhaps it is in fact the best decision available (albeit rather in a least-bad-option sense).

Why? Well, let's go through the alternatives. In hypothetical other universes:

1. SO takes a stance to keep the platform itself AI-free but the data fully available for anybody.
   This is what I would personally find the most amiable option, but a sober look on the actual ramifications makes it not so great. Namely, it would mean that every AI company would just feast on the data without any consideration for things like attribution for post authors at all. Users looking for help wouldn't care about that and for them just using a 3rd-party AI would be the increasingly more convenient option. As a result, SO would end up being more or less a ghost town plundered for its past wealth but with little relevant new activity.
2. SO stays out of AI and locks down access strictly to ensure the data stays in human-only hands.
   This is undesirable in many ways, both in principle due to things like privacy but also practically because this sort of gatekeeping would make it very unattractive for new users. It might in the short term succeed in making the platform more of a "clean, high-quality human content" place, but over time it would again just make it less and less relevant. Simultaneously it would make AIs worse for programming purposes (assuming the access-control mechanisms actually work in that way), because they couldn't access the SO data; opinions may vary on how good or bad that is, but it would stop mattering when an alternative platform pulls ahead of SO in terms of user- and post count and eventually also catches up on quality (perhaps not quality of the average post, but in the sense that far too many questions simply don't get asked on Stack Overflow anymore).
3. SO opens the gates, encourages companies to train their AIs on the posts by SO users without consideration for original author attribution, and then attempts to get back the wins by integrating some of these AIs back into the platform.
   I suppose this might briefly make it more popular with newbies, but it would alienate experienced users who are nothing, but cows being milked for content in such a system and increasingly don't get to interact with other users at all, but everybody just with AI.

In the light of all that, revisit the actual decision in a (perhaps naïvely) amenable reading:

4. SO stays mostly how it currently is in terms of user interactions, but makes it more cumbersome to mine data in real time. Looking forward, it cooperates with companies who will develop "socially responsible" AI that helps keeping the platform attractive to new users (some of whom will hopefully eventually become experienced and valuable members of the community), whilst also respecting the users who answer questions both by making AI-generated help give attribution to the authors whose work was crucial there, and by keeping them in the loop with votes, future questions, etc.

Again: I'm not very optimistic that this will work out. Will the "responsible" AI be attractive enough to help-seekers (compared to independently trained ones)? Will it actually respect answer authors? Will the quality of information circulated on the site be, if not improved, then at least kept at current levels? - who knows. But it is better to try than not to, and of all the possible strategies, I would indeed assess 4. as the most promising.

Of course, one thing that really would be desirable is that SO discloses the economic side: how much does Google pay for being the "responsible partner"? Are there other contestants, and how do they compare in terms of both money and prospective social responsibility?

May 9 edit

By this point, a status update on what concretely SO is doing to ensure the supposedly non-negotiable "socially responsible" is overdue. Instead, what they have done is another corporate-speech oh we are so good press announcement, and with OpenAI, of all companies. The same OpenAI that started the whole debacle by unleashing ChatGPT in a state that absolutely does not honour the relevant issues like attribution.

It does now look an awful lot like all this "responsible" stuff is just empty talking, written to distract from the cashing-in. I still have hope that SO will prove the critics wrong, but I'm no longer willing to place any trust in it.

I will not contribute to the platform until more concrete information is available. I see several users are self-vandalising their contributions now. This is perhaps a bit of a knee-jerk reaction, but by as it stands now I consider joining in. Too much is at stake to just let these companies get away with pretty words but no evident actions in the right direction.

In the "Defining socially responsible AI" blog post, the first part of the vision is about attribution being non-negotiable:

All products based on models that consume public Stack Overflow data are required to provide attribution back to the highest relevance posts that influenced the summary given by the model. With the lack of trust being felt in AI-generated content, it is critical to give credit to the author/subject matter expert and the larger community who created and curated the content being shared by an LLM. This also ensures LLMs use the most relevant and up-to-date information and content, ultimately presenting the Rosetta Stone needed by a model to build trust in sources and resulting decisions. Sharing credit ties to attribution and with attribution—trust is at the heart.

There are several problems with this.

It is not yet settled that training a model requires attribution. In the United States, there are court cases that may resolve this. However, until there is a more recent precedent from the courts, cases such as Authors Guild, Inc. v Google, Inc. and Authors Guild, Inc. v. HathiTrust have found that text and data mining is fair use. There are cases where models have regurgitated protected text or images, but there are also pending court cases to address the question of responsibility.

SE has no right to take action to protect our intellectual property. The official stance of the company is that they cannot help us if our intellectual property is misused. Several YouTube channels are taking and using content from the SE sites with insufficient attribution, but the company isn't and can't take action. Should the courts decide that model training requires attribution, it would be our responsibility as the copyright holder to take legal action against the model developers.

Why is the company taking these actions? Why block AI model developers from a rich, robust source of training data and selectively try to "help" the intellectual property owners without consulting us first?

What assurances does the community have that this will not open the door for plagiarism via uncited works?

I assume some of us perhaps understand that our work benefits others, even corporate entities... but that was always supposed to come with citation.

Has Stack Overflow the company considered that there may be a class action exposure to violating millions of user's content licencing?

Google is a proprietary company which obsoleted their own question and answer platform.

Stack Exchange is a question and answer website. It has its own server. It stores the emotional labour and research effort of uncountable user. For the public good.

Why you think it is justified to give control of a company over this huge volume of labour? Why do you think it our contribution should be used by the chatbot of only one company? Feels disgusting.

Update: Better keep this site content and moderation totally manual/human.

I have an idea.

I suggest allowing AI to post answers and then train AI on all that AI generated data.

WIRED reports (my emphasis):

The Google deal will also test how users of the version of Gemini for Google Cloud integration can create new data for Stack Overflow. People who don’t get a satisfactory response from the chatbot will be able to submit their query to Stack Overflow, where once approved by moderators it will be available for the website’s community of users to answer. As the companies prepare for the demo in April, they also are talking about letting users submit improved answers back to Stack Overflow.

Are we stepping away from the volunteer moderator model and paying folks to check queries sent over from Gemini?

SE originally banned ChatGPT content, ostensibly because of concerns about accuracy. All my interactions with generative AI have convinced me that that concern was well founded. I never get code from ChatGPT or other LLMs that works out-of-the-box.

Since Google is blocked in China, does that mean SE will also be blocked in China now?

(It is vital to know if anything could affect users' social credit scores kept by a foreign power.)

Stack Overflow will work with Google Cloud to bring new AI-powered features to its widely adopted developer knowledge platform.

And what does Google give back to Stack Overflow users who provide efforts to solve technical issues or share theirs?

I still feel that the best reward for the huge amount of voluntary work incorporated in this Q&A website would be to direct everyone having a question precisely to this website and not just to the naked answers, thus feeding back to its community. Anything that predates the answers bypassing actual participation is detrimental to the concept of collaborative Q&A.

It's like printing Wikipedia on a paper book and selling it; you can do it but it sucks.

janw, Journeyman Geek, CollenV, and others have asked what the community is getting out of this. We can’t discuss financial and contract details, as you are all aware. What I can share with you is that resources will be put into the public platform.

We have some exciting updates to share in the upcoming weeks. Some initiatives that we know the community cares deeply about are being prioritized. Those announcements will be coming soon.

I know it’s frustrating that I can’t be more specific than that right now, but I do mean it when I say those updates will be coming in weeks, not months. In 6-8 weeks. ;)

I had seen here. It just says that we need to trust AI, how can we more effectively support developer needs? How do we “trust is at the heart”?

Please don’t announce something that doesn’t contain important points. I don’t think it answered me how we trust AI! If we use AI, it just ruins this SE only. It was hard to control the AI.

The reason I use SE is want to learn English and Mathematics from the Stack Exchange. And I can get the correct and trustful answers from Stack Exchange. Why do I want to use SE if AI was used by SE? I think that I will use other sites or communities instead of these sites.

Why do I need to use “Gen AI stack exchange” if it is the same as the function with the ChatGpt? (I mean for the future.)

I hope the Stack Exchange Team or community should think wisely first, and I hope that SE always bans ChatGpt everytime.

Partnership with Alphabet/Google means I will probably not contribute here again. I just came from contributing a small answer.

Alphabet is a war criminal company and blissfully provides partners with the huge US spy agencies. I'm surprised that this partnership fits with Stack Overflow team's ethics. I read somewhere that Stack Overflow won't comment on the arrangements, particularly the financial side. This is anything, but transparent and community friendly.

I'm not a core user, but there are lots of people out there like me.

The initiative to make Stack Overflow and the Stack Exchange sites more accessible will just dumb the sites down and make them far less useful.

In an ever-evolving world, we’re always looking for new ways to get this repository in front of the world’s technologists: those who can benefit from it and contribute back to it. And with this new joint initiative with Google, we’re laying the groundwork for this reality.

Given that there's no talking SE out of this, I hope this is just a start of this effort. Siloing Stack Overflow (and/or all Stack Exchange data) in one AI platform is a detriment to society at large. This data has been generated by Stack Exchange communities. It's understandable that Stack Exchange is a company, and as a company needs to be profitable. That said, this data should be available to the benefit of all mankind, not just one AI system or company.

I implore Stack Overflow to ensure that the data to remain democratized and available for all seekers, corporate or otherwise. Stack Exchange must not contribute to a "exclusivity" situation, like with Streaming providers today, but instead with all of the data on the internet. Imagine one AI being completely oblivious to entire swaths of the internet, just because of an exclusive agreement with one company or another. That would be exceptionally awful.