Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Inside The Decline of Stack Exchange

thediff.co

Post reply on HN
Print view View original

81–90 of 179 posts

Re: Inside The Decline of Stack Exchange

#81

Post by croes »

laichzeit0 wrote: post #31

Some of my most recent ChatGPT interactions: 34.223.203.0/28 what is the end ip of this range? Write python code to strip out characters that are not alphanumeric or spaces How do make FastAPI not return nulls in the response model when values don't exist? Rewrite this javascript to python Braindead stuff that I would normally lookup on StackOverflow but ChatGPT gives me the actual answer instead of me having to sift…

Wouldn't you normally chose the answer marked as answer marked on Stack Overflow or the one with the most points?

ChatGPT doesn't offer that kind of user feedback, so you have no measure if the answer is valid if you don't know the topic well enough

Re: Inside The Decline of Stack Exchange

#82

Post by kodra »

OliverJones wrote: post #23

I would really like to see what Joel Spolsky and Jeff Atwood, the original creators of Stack Overflow, have to say about the current state of affairs. Their business model was recruiting by identifying people with specific skills, and it was effective. I hired and was hired that way. (They sold it years ago to a South African private equity venture and made the pile of money they deserved.)

Someone should feed an LLM all the Joel on Software articles and ask it to comment on Stack Overflow.

But a serious question: is Joel still in charge of SO? Has he retired?

Re: Inside The Decline of Stack Exchange

#83

Post by michaelbuckbee »

Not mentioned in the article is the negativity found with asking questions on Stack Overflow, where your question is marked as duplicate if it's vaguely like something else, if it's not worded exactly correctly, etc.

A remarkable thing about asking ChatGPT questions is that you don't have to be defensive, you can just ask your question no matter how basic, dumb, or simple it is without justification.

Seems a bad state of things that the inhuman AI is extending more grace than the human community.

Re: Inside The Decline of Stack Exchange

#84

Post by cornholio »

I wouldn't exaggerate the death of Stack Exchange, it seems like they have a nice moat and are well positioned to change the future licensing terms of its answers and forbid using them for training AI.

They have already banned machine generated content. When the question of training legality will be settled in courts - predictably, not to the advantage of the AI companies - Stack Exchange will be left with the human network generating the content, while AI companies will need to license any new answers it they don't want their models to remain stale and out of date.

Re: Inside The Decline of Stack Exchange

#86

Post by webnrrd2k »

pbourke wrote: post #42
duxup wrote: post #11

Earlier quoted context omitted.

Entitled "special" people are a problem everywhere. Dev's who are labeled as high performers get greenfield projects that can be pumped out quickly / look good ... someone else does the maintenance, they get all the flack for an obvious bug that wasn't fixed even if they didn't make it. The random executive who is thought to be highly successful who jumps from department to department spreading his fingerprint long e…

As time goes on, I more and more have the feeling that maintenance is 95% of the value in any software system. A well-designed system at launch is not sufficient to provide sustained value, but competent ongoing maintenance is. Phrased another way, you can sustain value by maintaining a badly designed system but not by not maintaining a well-designed system. Ideally systems are well designed and well maintained but w…

Yesterday I saw a post on Lamba The Ultimate that, while a bit long-winded, does a great job addressing this issue. The article picks up in the section headed The Labor of Care, and, for me, the money quote is: ”In order to care for technological infrastructure, we need maintenance engineers, not just systems designers—and that means paying for people, not just for products.”

It's well worth the read for the second half, and I made a post to HN because I'd love to hear more discussion on the issue. (https://news.ycombinator.com/item?id=37116593).

Re: Inside The Decline of Stack Exchange

#87

Post by indeed30 »

TheRealPomax wrote: post #40

Looking at that graph though, what it tells me is that stackoverflow was entirely unaffected by ChatGPT, and the giant spike is just humanity collectively jumping on ChatGPT in addition to folks using Stackoverflow in exactly the same way as before? Same for that "answers over time" graph: that's literally what you'd expect for a knowledge base. There has to be an initial trend upwards as the site starts to accumulat…

> At that point, the curve levels off, and as the number of typical questions die off because they've already been asked and answered many times, the number of new questions has to trend downward

Yes, absolutely, and the article addresses precisely that point. But the questions graph shows a drop of about 40% at exactly the point in time when ChatGPT launches, which is much too extreme to explain away like that.

Re: Inside The Decline of Stack Exchange

#88

Post by valine »

Gollapalli wrote: post #74
valine wrote: post #33

Earlier quoted context omitted.

You’re assuming the bulk of ChatGPT’s coding knowledge comes from stack overflow. I suspect training directly on code bases and documentation is more important.

No, ChatGPT isn’t AI in that sense. It can fire back documentation at you, but a lot of what it does is just look at the word it’s spit out, and pick the next word or snippet that looks right. Same thing with code. It just spits out what it’s trained to interpolate the next snippet to be. Anything involving genuine reasoning needs to be done by a human. It’s a great documentation parser. A step up in search engines.…

If this were true ChatGPT would be useless for any topic that doesn't have a vast Q&A dataset. The purpose of instruct fine tuning is to illicit the model to output in a question answer format. The bulk of the model's knowledge comes from the unstructured dataset.

I'm also not sure how "reasoning" at inference time is relevant here. You seem to be conflating unsupervised offline training with inference. The ironic thing is that the behavior you describe actually works, you can try it right now. Add code documentation to your context window and ChatGPT will happily write code using that documentation as reference.

Re: Inside The Decline of Stack Exchange

#89

Post by no_time »

to me it looks like its too early to draw conclusions from the programming question volume graph. But if it is indeed true that LLMs are leaving SO in the dust, thats crazy.

I’m in my 20’s and this is the first time I feel like my work routine and ways of interacting with tech is getting antiquated. God damn.

Re: Inside The Decline of Stack Exchange

#90

Post by TheRealPomax »

indeed30 wrote: post #87
TheRealPomax wrote: post #40

Looking at that graph though, what it tells me is that stackoverflow was entirely unaffected by ChatGPT, and the giant spike is just humanity collectively jumping on ChatGPT in addition to folks using Stackoverflow in exactly the same way as before? Same for that "answers over time" graph: that's literally what you'd expect for a knowledge base. There has to be an initial trend upwards as the site starts to accumulat…

> At that point, the curve levels off, and as the number of typical questions die off because they've already been asked and answered many times, the number of new questions has to trend downward Yes, absolutely, and the article addresses precisely that point. But the questions graph shows a drop of about 40% at exactly the point in time when ChatGPT launches, which is much too extreme to explain away like that.

Except the graph cuts off before 2023, so it tells us nothing about the actual effect it had, it cherry picked the initial drop off (or worse, they just used https://insights.stackoverflow.com/trends, didn't even look at the time axis, and went "this proves our point").

It's Q3 of 2023, if you're going to argue a decline, why are you not looking at the most important three quarters of data?

Post reply on HN