Hacker News Reader

Comment by CodeCompost | Hacker News Reader

People may be dismissive of Serverfault and the other sites but pretty much all LLMs are trained on them.

Yes, it's like a lot of other content that no one will bother to maintain anymore because it is easier to ask an LLM. Not sure where they are planning to get their new training data from ...

I mean, most frontier labs train on user input by default, seems like a pretty good source of information.

Average user input quality is not optimal by definition. SO's upvote system and reputation system is what really refined quality. Much like Reddit's system is helping a lot for conversation. Much like academia's publish and citation system is an even better system.

I'm not saying LLM won't be able to train further, but tough and sound reviews are paramount to improving the corpus of quality content. And with the replacement of SO by LLMs, one can wonder if one quality system went down the drain. Anyway there's still code reviews on GitHub and in private repos, but we're still far from getting the human judgement out of the loop.

I do expect there will be more closed knowledge forums. Unfortunately one person exposing it to LLM provider will be enough to break the whole idea. But I believe there will be many that will keep secrets for some in group.

"Average user input quality is not optimal by definition."

Absolutely.

"SO's upvote system and reputation system is what really refined quality."

Absolutely not!

SO's problem was exactly that they gave average users moderating power. That lead to half of the questions being closed because they were over the head of the average user/moderator regardless of being interesting and useful questions or not.

This issue was mostly on SO itself. For most other stack exchange servers, it worked really well.

I'm sure there were plenty of people like me who couldn't help because they didn't have the right amount of "karma".

You can train on questions, but I will rarely answer to a chat with a SO-quality answer.

People do still talk about programming online, just not on Stackoverflow.

I think since AI finally provided a non-adversarial alternative to Stackoverflow it lost the critical mass to be a one stop shop for all tools/languages/projects. So people just reverted to communication forums specific to those tools/languages/projects.

There are plenty of project mailing lists, GitHub issues, language forums, blog posts etc. that are still active. E.g. just the Rust users forum gets more than 2-3 posts per day.

Stackexchange network has died for me way before LLMs were even a dream.

Passive aggressive commenting and deleting questions because it looked remotely similar to another one was enough for me.

I was there for learning and sharing, not for the proverbial fighting cage.

I can spend 5 minutes more and use a search engine, and then take my own notes and keep them to myself. Thanks.

The aggressive moderation on StackOverflow was so annoying. I still remember multiple times where I found a question about my exact issue, but it had been closed as a duplicate, and the "duplicate" was basically irrelevant.

And plenty of times I had an answer to a question but couldn't because I didn't have enough 'karma'. Sorry I'm not going to waste my time gaming the system to answer a question and be helpful.

The AI "understood" the concepts enough. It can generate readable code in new syntax like Typst, D2lang etc; refer 80 page documentation scattered across urls (looking at you AWS); clone tools from Github to find undocumented features.

It was great while it was good. The demise wasn’t quality as such, but rather the insane moderation - closing everything as duplicates or out of scope etc.

CoSOMO (Content Stolen; Obsoleted, Move On)

Ouch ... brutal, disheartening, but true.

You guys tbh are just infuriating. StackOverflow and ServerFault in particular are CC BY-SA. I contributed to a lot of websites which are currently switching from public access to login required. I did it under the assumption that anyone will be able to read it. Thank god Stack Exchange and HN so far are not showing petty resentment.

You are absolutely right. These contributions were given willingly under a very permissive license. That is on the contributors.

It might not have been that apropos to roll out in this context, but the point goes for all LLM vacuuming.

PS: Although those contributors might have been willing to let anyone, corporation or person, manually use their submissions, they might have felt differently about remix machines.

> Although those contributors might have been willing to let anyone, corporation or person, manually use their submissions, they might have felt differently about remix machines.

How does a corporation "manually use" a submission?

How people are not "remix machines"? I mean it is a common joke that bad programmers used to copy-paste code from Stack Overflow.

Surely you don't mean that GPT-6 itself is at fault here.

It “““manually””” uses a submission by working with data and information like the Old World did, before our wonder machines made copying and plagiarizing (in the non-consentual case) into a factory. One might consent to how things work in a world with certain preconditions but not when those preconditions change.

“Remix”: you can make any analogies you like. Huh, people are non-deterministic because you don’t know their brainstate. You can go nuts with that kind of misanthropic sophistry for all I care.

Of course I don’t blame a computer program for anything.

And all the sites trained on humans, something which we never found offense with, for some interesting reason.

Reply on news.ycombinator.com