I'd like numbers on what % of questions are scolded for having been asked
Just think of a genuinely good question and show you've done a modicum of research and people will happily reply to it
stackoverflow.com/users/8610114/sensanaty
Seriously just check out questions I've asked on SO. They're nothing groundbreaking, a lot of them are pretty beginner/basic questions, yet I've never really had adverse reactions to any one of my questions. Some go unanswered sure, and I might have one or two marked as dupes (they are more or less) but if you've ever browser the "new" portion of the site it's inundated with hudreds of shitty, low effort questions, so it's understandable.
SO is also an excellent jumping off point for figuring it out yourself. I've had dozens of times where I figured out a solution midway through typing a question, since you have to make good quality questions there in the first place.
You're being whined at for not knowing the very specific incantation which will summon the results. Which, if you knew, you likely wouldn't need to ask anyway.
"You're stupid and your question is stupid. I forbid answers. I will keep it up and nofollow all of the scalped resources it links to in the hopes others can come here and benefit from this lack of answer"
Which is my point. ChatGPT is the most basic modicum of research and it is excellent for that. But it also can do a lot more.
Collaboration on writing a better question is the only way its answers could get to the next level and customized for your specific use case. I'm imagining it spinning up a generated SO clone for you to browse and scolding you just the same (but at least giving you a decent link), or at least asking why an answer you found isn't good enough.
The way collaborating with AI works now is too wordy and error prone. You don't know what you don't know. Chat bots still don't talk back much and that's a problem.
When we get to this level of AI, maybe some people will finally understand that getting your expectations broken is not a form of hostile communication.
technology moves on
in 2030, when stackoverflow have folded due to lack of visitors, and all new data on new libraries/APIs are locked up beyond OpenAI's reach
how useful is the AI going to be?
the whole thing sows the seeds of its own destruction
It is, as before: if the service is free, you are the product.
That being said, I think that to the extent AI is an actually helpful product, as opposed to social media, public forums, or free video hosting, the tradeoff is way more worthwhile. In my most humble opinion. I'm sure your calculus will lead to different outcomes.
yes, it's asked to produce something, but it won't see the result after substantial effort required to actually make it work
the interaction is not going to be nearly enough to produce useful training data to replace e.g. stackoverflow
and if GPTBot can reach your site, you are the product, they would make you their product even if you don't use their free service
but now they are not allowed to complain when others train models on GPT-4 data, they can become someone else's product too - see Orca, TinyStories and Phi-1.5
Then I download the docs and feed it into the LLM I have running locally.
if the fair use lawsuit succeeds and copyright protection over open source is essentially terminated: what companies are going to open source anything?
> how useful is the AI going to be?
In 2030 the AI can go and and search for the answer itself, by installing the environment in a container and trying to find a solution. It's something that is in some way possible even now, but probably not practical yet.
But to do that, I'd have to make sure there are no duplicates and frame the question with every appropriate detail.
Or, I could just have a quick conv with the AI. Sure, it'll get some things wrong, but so will SO answers. :) And it is amazing at how clarifying it can be. Even the wrong answers are sometimes a clue that the questions themselves are a little off.