-2

The recent changes to data dump requires an agreement (emphasis is mine):

I understand that this file is being provided to me for my own use and for projects that do not include training a large language model (LLM), and that should I distribute this file for the purpose of LLM training, Stack Overflow reserves the right to decline to allow me access to future downloads of this data dump.

However, the post Announcing a change to the data-dump process also reads:

It’s important to say that when you breach the agreement that you make when downloading the dumps file, we do have the option to decline to provide you with future versions of the data dumps. But we really don’t want to have to do that. [Philippe]

Does Stack Exchange plan to exercise that option and decline providing future versions of the data dumps to people who train LLMs on it or share it to people who may train LLMs on it?

Improve this question
0

2 Answers 2

Reset to default
19

Yes. 💯.

The only reason the Company is making changes to how they distribute the Creative Commons Data Dump is so that they can gate it for LLM use. This has been a constant goal in all of the data-dump posts on Meta for the last 15 months.

Translated into business terms, in case an exec reads this: "Due to macroeconomic headwinds, Stack Overflow is being frugal with investments, and only allocating resources with a clear financial benefit."

Improve this answer
4
  • 5
    As I said on chat, it is pretty clear that "in the last two years of moderators begging for better tools all they had on their mind was to find a legal way to restrict access to their assets - that being Subscriber Content - so that they could sell it to OpenAI and even make them pay an extra for "Exclusive access". Because let's be honest. The real point to the dump changes is not to "protect us" from the companies that want to steal our data for their LLM without ethical attribution. It is just to paywall access so that those same actors make them money."
    – SPArcheon - on strike
    Commented Aug 1 at 9:11
  • I would even go further. All these events paint a picture that points at Prosus only reason for buying Stack Exchange being needing a "factory that produces LLM training material". I will give to them that they were pretty "smart" with this choice. Far before the big "codepilot and tech/IT AI support tools" they already saw the potential of getting exclusive access of the content of what was one of the most popular site for professional IT Q&A. Sadly, that content came with a Creative Common string attached, so since then all has been going in their mind is finding a way to kill it.
    – SPArcheon - on strike
    Commented Aug 1 at 9:18
  • 8
    @SPArcheon-onstrike You're giving Prosus too much credit in their motivations for buying the company. The whole OpenAI thing caught the company by surprise, and they've been stumbling to find some solid footing ever since. Prosus's public filings show that they currently value the company at a fraction of the acquisition value.
    – AMtwo
    Commented Aug 1 at 14:04
  • 5
    I think at the time Prosus bought SE - it was part of an ed-tech portfolio + the potential of teams. There's a few 'broader' issues with that segment of prosus - with write downs on quite a few investments. Predicting OpenAI and similar products might have meant a much more agile reaction than we've seen, and might have avoided some of the painful belt tightening we saw.
    – Journeyman Geek
    Commented Aug 2 at 13:34
12

Yes, they plan to do that. And maybe even more. In my opinion, this is probably precursor to turning off data dumps altogether (emphasis: my opinion).

The agreement reads:

I understand that this file is being provided to me for my own use and for projects that do not include training a large language model (LLM), and that should I distribute this file for the purpose of LLM training, Stack Overflow reserves the right to decline to allow me access to future downloads of this data dump.

Saying 'But we really don’t want to have to do that' is nonsense. If they do not plan to use that "right", why bother making the changes. No company would invest money, resources, and time (especially companies run by profit-oriented CEOs) on something just to reserve the right and leave it unexercised. They made the changes knowing it'd cause a backlash. What? Just to feel happy about their "rights"?

They're asking us to give them a stick and consenting to getting beat up by it, but then saying they wouldn't really want to hit us. Why should I believe this? It's not like they haven't beaten me up already, and over and over (plenty of examples on Meta).

Improve this answer

You must log in to answer this question.

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