Show HN: StackOverflow.gg – AI-generated answers to every coding question
stackoverflow.gg
stackoverflow.gg
Maybe it's chaotic like you and me, but it cant be non deterministic, it'd mean it doesnt even really answer questions and just spout random tidbits :s
"Top StackOverflow expert on XX", so the incentive is still there.
One solution to spam could be to keep the voting system, but not publicly reward users with points (like on HN).
Edit: come to think of it a Markov generator would probably be better on the question end so the hallucinated answers have more work to do.
This process may vary based on the TLD. For The Channel Islands, you can start here: https://disputes.gg/
IANAL
You really can't use the exact name of the people you are cloning. Even something like "stackrobot" would risky. Safest thing is to not use any words from the original product.
Source: Ages ago I had a personal domain based on a longstanding nick that happened to partially coincide with the name of a major motion picture studio. That it was a paralegal who called me said they weren't anxious to take it seriously, and all they were really after was just getting me to say it was nothing to do with them so they had me on record in case I later did make it something to do with them. No big deal, but it wasn't like if I had a treatment burning a hole in my pocket I could've parlayed the conversation into a chance to submit it. And even at that, I doubt this will be anything like so simple and nonadversarial.
Of course. Here's the plausible chain of events:
1. Someone at the co. reads about this on HN 2. They tell their boss who might tell their boss, etc. 3. It finally gets to someone who is responsible for telling the lawyers. But they also say "hey this is kind of cool. We'll send a C&D, but maybe reach out to see if there's anything interesting there."
(Not least because escalating something to counsel, and then having side conversations about it with whoever you just brought to counsel's attention, is a bad idea!)
And that's if they see something here they might want to investigate further, which given SO's stance on ML-generated answers up to this point, I tend somewhat to doubt.
I have, in fact, seen this exact scenario!
I don't think it's likely, but losing a domain is a small price to pay if that were potentially a goal.
Deliberately infringing someone's trademark with a domain name seems to me like a dicey way to look for attention, though. But hey, "it's a bold strategy, Cotton, let's see if it pays off for 'em."
But you could start a company called Stack Overflow that sells cookies in Austria. They may still try to sue you but then it comes down to an actual argued case and debates and such. This is just complete copy.
There is some imagined distinction between a hallucinated answer from a LLM and one from a person posting "try {random code}" that are both equally wrong - but only the LLM is deleted (because deleting the wrong answers are an attempt by a human to answer(?) or would hurt feelings / engagement (?)).
If the answer is from 8 years ago, did the API change between then and now? Or was the answer for the same version that you are using?
Part of the challenge of StackOverflow is finding the question that matches your situation. But that is a pointless act with ChatGPT -- just ask it your exact question. Duplicates be damned.
Bard mentioned time-traveling UFOs for no relevant reason.
I see various software as browser addons/extensions... is that a security loophole?
Seriously, I think it means questions which don’t yet have community answers.
But that doesn't get you the sweet sweet ad money I guess. Quick Stack Overflow, make another blog post no one gives a shit about, and continue banning 90k rep users over rollback wars.
I'm kind of curious, because I've heard the line of reasoning about keeping it "clean" multiple times. Like its only purpose is to be a pristine knowledge base where every question is asked only once, and a perfect answer is selected and enshrined for all time. That's all wonderful and magical, but how does also serving it's purpose as a QA platform for new people on the struggle bus prevent that first goal?
I don’t ask questions, because people take a very carefully explained, explicit question and post irrelevant stupid comments on it (“why are you doing this?”) instead of answering the question.
I don’t answer questions, because people upvote wrong answers and vindictively downvote answers to the actual question (which is why everyone posts comments instead of answering, because you can’t down vote them for it).
People care more about karma than actually helping.
The quality issues are real, but really?
I care not at all if 90k+ karma folk leave. Just leave. If you don’t like what it is, go hang out somewhere else. Why are you there if you’re not trying to help people?
What I care about is the people patiently answering questions in a way that is correct, not antagonistic, not ducking off because it took more than 5 minutes and they can get better karma somewhere else.
…almost like, a dedicated support community who care about a topic. That’s a nice idea. You could probably find one of those for your topic of interest and get help there from nice people. That’s what I do now. The folk there are generally quite lovely, because they actually care.
The people on stack overflow, by and large, are not nice, do not care and won’t help you if you have any really hard problems to solve.
An AI would be just as good, if it had decently accurate answers.
Well, this reminds me of "if my grandmother had wheels, she'd be a bicycle."
Failing to clean up bad questions is encouraging all the bad behavior you're talking about, since answering yet another question from someone who couldn't be bothered to write his own regex is an easy 25 points that nobody will ever look at again.
Or would those people be happier if they just had their questions answered?
The answerers with their karma; happy! The askers with their answers: happy.
The only “losers” are people who hoard their karma… because (I guess?) now other people are getting karma? …and I just do not care about those people.
The community has spoken, to the extent whether the majority of usage becomes the defacto standard, and basically filtered out all but the most basic coding issues that are just one peg above homework assignments.
> The people on stack overflow, by and large, are not nice
From your long rant, I'm not sure if you're coming off any better. To me it sounds like SO should be happy someone like you isn't contributing.
Is quite a harsh personal attack, and probably contributes to the author's feelings regarding hostility and gatekeeping.
But you're right, I should've avoided going personal.
They never said that. They said that stupid comments get posted.
I literally not do care remotely what you, or other folk like you on SO say. :)
You’re why I no longer go there.
Please see the comments under the OP. While the question is technically sound (I've needed this exact thing for a REPL I was writing), the comments are perfectly useless, self-entitled and discouraging drivel. Notable example:
> This is an incredibly odd thing to ask
(emphasis theirs). How is this even slightly odd. Are REPLs odd? Of course not.
They ask a question with some weird requirements, people ask why those requirements are in place as it makes it hard to solve, the asker explains why. Everything is good.
I see people complain about how their "well researched and well articulated question got closed" all the time, but when prompted no one every actually provides a link to their question. Which makes me believe it actually just was a low effort noisy question after all..
If anything, HN is much more snarky, just look at the comments on anything here.
It's clearly not allowed. People shouldn't be surprised when their answer gets closed. The goal of stack overflow is to build a quality mapping from questions to answers. There is no way to map a subjective question to an answer.
A heavily policed community is not a welcoming one.
Visit new questions: https://stackoverflow.com/questions?tab=Newest
Scroll until you find an upvoted question. Count how many downvotes you see on the way. You can often get to page 2 with dozens of downvotes before the first upvote.
Hard mode: Scroll until the net upvotes are more than the net downvotes.
But I tried: scrolled two pages. Most questions were at 0 votes. But I checked every down voted question, and they were all useless. No code, just a vague statement of trying to do something.
That's exactly it. Someone out there is asking for help. They don't know how to ask or where to start. Then they get shut down by the community for being useless and vague instead of guided on the right questions to ask.
This is the first downvoted question I see: https://stackoverflow.com/questions/75396340/java-17-switch-...
It's someone who does not understand switch and wants to know why their code doesn't match with their understanding of switch. There's no given reason for the downvote.
ChatGPT at least takes the effort to make sense of the question and verifies that this is what the asker intended.
Today that question would be downvoted as a bad question, and so would many of the answers.
It's kicking out the ladder after you've climbed it and is hostile to anyone learning to program today.
All those questions you deride as "useless" is someone who is struggling to do something, who takes the time to ask for help and gets downvoted and hostility in return.
Was your comment intended to be the perfect example of this? replying to a joke about insults in answers... by going on an entire rant about how it's easy not to get your question closed (which was not mentioned at all).
Note, I do like the comments on HN, and spend more time here than on SO. I personally like how they're both direct and don't accept too much fuss. But I also see newcomers here getting burnt by "that orange website", not being used to how it is.
However, even when ignoring all that, if you really ask a "well-researched and well-articulated question" on a complex matter, you will not get an answer in most cases. For those after reputation points, it is far easier to answer simple questions. You are far better off raising it as a GitHub issue if it is an open-source framework.
Ask, and ye shall receive.
Here's my SO profile screen[0]. You can tell that it's me, because of my dragon logo, which is actually a refinement of my signature from the days I wanted to be an artist[1]. I actually make an effort to be "non-anonymous," these days. I feel that it helps me to be a better netizen.
The first thing that you may notice about my profile, is that the Answer[2]/Question[3] ratio is about 0.6. That's unusual for many of the more senior members (I've been there for about 15 years. I used to have another account, and nuked it in disgust, many years ago, then decided I was being a big baby, and came crawling back. I've been a member since, but the name was changed, somewhere along the line, as I decided that "non-anonymity" was my lodestar).
That's because I don't go there to answer questions, or prove how smart I am. I go to solve problems, and remove roadblocks, in my day-to-day work.
Here's my latest question[4]. It's a really basic question. One I could have found the answer to, myself, but I would have had to spend a bunch of time in the ugly, chaotic documentation that Apple has for their bundle hashes. Probably would have taken me an hour or so.
I should tell you a bit about myself.
I have been writing software, since my days of 6800 machine code, in 1983. I've been writing Apple software since 1986. I have been shipping software, since 1986/7 (depending on how you define "shipping").
This ain't my first rodeo.
I've also been training and mentoring folks since I was 19. One thing that I learned, very early on, is that the people I'm training, usually (not "sometimes," "usually") have more expertise than I, in matters outside the scope of the class.
I learned to respect my students. I'm not better than they are. I just have a bit of knowledge that can help them, and it's an Honor to be allowed to provide it.
When someone asks me "what I do," I generally respond with "I solve problems."
That's what I do. Every. Single. Day. Every day, I am presented with a "Argh! We're all gonna die!" problem; often multiple times a day[5].
And I solve them. That doesn't make me unique, in the slightest. Lots of people here, can easily say the same, and they probably solve much more difficult issues than I. That's one of the reasons I like coming here. I'm not the smartest person in the room. I have found great comfort in surrounding myself with people that humble me. It can be difficult, at times, but I feel that I need to learn something new, every day, and I don't get that, by staying in my comfort zone.
I also speak Swift without an accent. I've been writing Swift, seven days a week, 52.14 weeks per year, since the day it was announced, in 2014. I've also been shipping apps in Swift, since not too long, afterwards.
Everything I do is geared towards "ship."
That often means that my solutions and workflows are not always optimal, or "academically pure." Duct tape and baling wire. Make it work. I take a great deal of care in the Quality of my work, but it is not always "exemplary." It needs to work, work well, and be maintainable. I also like to get to the point, where I leave it behind, and move on to the next project[6].
SO helps me to quickly find "correct" solutions. I don't like duct tape and baling wire. I'd like to do it right, the first time. That's pretty much exactly the spirit behind my latest question [4].
Note that the question got almost immediate downvotes and close votes. That happened within seconds of it being posted.
But it was answered. It was answered in a rather condescending way, but I got what I needed. I then took that answer, and configured it into something that I could ship[7], and, eventually, into something that I could share[8].
Now, that was a productive interaction for me. It gave me the answer I needed, I was able to greencheck someone for answering it, even if they did so, looking down on me, and I was able to repackage their response into something that can be quite useful, in the future.
When I ask well-researched, well-documented, windy questions, they are usually ignored, like so[9], and I end up answering my own question. Sometimes, these self-answered questions get fairly popular[10], or even unanswered ones[11].
It's fairly plain that the SO folks think of people that ask questions as "supplicants," and to be looked down upon. Look at all the folks with massive scores, and single-digit question counts (the digit is frequently "0").
Here's an exercise: Go to Meta, and suggest that people applying for positions of authority have at least a 10% question-to-answer ratio.
Why would you want a general that doesn't know what it's like to be pinned down in a shellhole?
Enjoy your downvotes and attacks.
Happy now?
[0] https://stackoverflow.com/users/879365/chris-marshall
[1] https://news.ycombinator.com/item?id=34654770
[2] https://stackoverflow.com/users/879365/chris-marshall?tab=an...
[3] https://stackoverflow.com/users/879365/chris-marshall?tab=qu...
[4] https://stackoverflow.com/questions/75361518/accessing-an-io...
[5] https://littlegreenviper.com/miscellany/thats-not-what-ships...
[6] https://littlegreenviper.com/miscellany/leaving-a-legacy/
[7] https://stackoverflow.com/a/75364118/879365
[8] https://github.com/RiftValleySoftware/RVS_Generic_Swift_Tool...
[9] https://stackoverflow.com/questions/62697317/strange-link-er...
[10] https://stackoverflow.com/questions/66736704/ibdesignable-ui...
[11] https://stackoverflow.com/questions/26736909/how-do-i-add-th...
I suspect that we could have actually found a lot in common.
In my experience, I have found friends in the most unlikely places.
(And, while I wouldn't ever install this, good to see that FF is still on the charts)
"Why are ChatGPT responses written a few words at a time?"
It replied, a few words at a time:
- - -
The responses are broken up into smaller chunks for several reasons:
1. Better readability: Breaking up the responses into smaller pieces makes it easier for users to read and understand the information being provided.
2. Interactivity: ChatGPT's response format allows for a back-and-forth interaction with the user, making it feel more like a conversation and allowing the user to ask follow-up questions or clarify information.
3. Error correction: By responding a few words at a time, the model can quickly correct any errors or misunderstandings in real-time, leading to a more accurate overall conversation.
4. Efficient communication: Responding in smaller chunks allows the model to quickly provide information and reduces the latency in the communication process.
Overall, the goal is to create an efficient, user-friendly, and interactive conversational experience.
For those who don't know, it is like Markov Chains: probability of next word (or a group of words encoded as a token) is calculated based on previous words and computationally intensive. It just uses not just probability between 1-2 previous tokens, it uses 2048 token window (roughly 1500 words) to predict next token, then puts next token into window and goes on.