Orange site
Stack Overflow questions are being flooded with answers from ChatGPT
Use of ChatGPT generated text for posts on Stack Overflow is temporarily banned
It will take Days for me to just read the SO post
Orange site
Stack Overflow questions are being flooded with answers from ChatGPT
Use of ChatGPT generated text for posts on Stack Overflow is temporarily banned
It will take Days for me to just read the SO post
Now playing: Blurry Flurry (Tropical Freeze).mp3
Jump in the discussion.
No email address required.
I hate chatgptcels. It kinda works for things that it was trained on, but the minute you ask it something it doesn't know about (like obscure libraries) it just starts lying and returning nonsense.
Jump in the discussion.
No email address required.
This Technology Review article is a little
, but the premise is pretty funny. I like the idea that all future language models have to either limit their training data or participate in an incrementally growing ouroboros of bullshit in, bullshit out.
Jump in the discussion.
No email address required.
I'm not buying this. In what situation would this really happen?
Anyone who has used ChatGPT for more than 10min will know it's dumb as heck and guessing half the time. Personally I think it just needs few disclaimers and on-boarding help screen explaining what it is - and what it isnt (as opposed to banning wrongthink and 'dangerous' prompts).
I'm very skeptical people will see it was an expert at everything and start jamming pills down their throat cuz it told them to. And I work in software, so I know how dumb the average person is.
These critiques always present FUDy scenarios with zero context, that it will somehow be used in a vacuum without other signals / information the user knows going into it (let alone common sense).
Jump in the discussion.
No email address required.
Also I feel like "misinformation" holds the same sort of cultural bugbear status today as "communist influence" did in the 50s. It's a buzzword-turned-useful spook for riling up tards to increase your political or social cachet.
Jump in the discussion.
No email address required.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
That's what I mean by
. They're really stretching for an irl consequence, but if you seek health advice from the internet, let alone a LLM chatbot, you deserve what you get.
The argument I found more compelling is that publicly accessible LLMs like ChatGPT are somewhat self defeating in the sense that they're tainting the data that could be used to train better future models. I don't know enough about machine learning to know how much of a problem this actually is, but I love the idea that proliferation of early versions like we have now could potentially impede future development. I'm not enough of a utopian idealist to imagine any future where outsourcing our problem solving abilities to a neural network works out well, so anything that makes the responses less accurate is purely funny and good imo.
Jump in the discussion.
No email address required.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
which makes it a pretty darn normal redditor
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
Yes, exactly, I never perceived it as a threat. but SO has the same questions being asked daily so I thought it would be helpful there but they seem to be very very mad
Jump in the discussion.
No email address required.
I'm going to get chatgpt as a browser extension so I can get better search results hopefully.
Jump in the discussion.
No email address required.
Large language models fine tuned for a conversational style + augmented with traditional search engine fact checking is gonna be the future of search
Jump in the discussion.
No email address required.
No, it's going to be the end of intelligent dialogue on the internet outside of well-curated walled gardens. Those AI models can produce text at a rate far greater than intelligent and informed humans can.
The result is going to be AI engines training on AI engine generated content and spewing out tons more AI generated content which is going to very rapidly replace human generated content.
It's going to very rapidly descend into 99% of the indexable written content being absolute nonsense which is basically what we're seeing now with google results, if you want anything other than SEO spam you have to refine your search to avoid all commercial sources.
This has made driving a narrative virtually free in terms of human labour costs. One person with access to GPT-3 can create a whole subreddit of imitation consensus about a product, political party, etc.
Jump in the discussion.
No email address required.
Oh yeah for sure, AI generated content is gonna dwarf human stuff way sooner than people would have expected a couple years ago
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
More options
Context
Those mfs get mad about everything
Jump in the discussion.
No email address required.
More options
Context
In an ideal world there would just be a ChatGPT service for people to ask easy programming questions, leaving the more difficult and niche questions for SO. Of course, in this shit world of ours SO just gets spammed with garbage ChatGPT answers, instead.
Jump in the discussion.
No email address required.
That's what Github Copilot is. I use it every day for that
I can see it getting much better in the future as it copies ChatGPT model
Jump in the discussion.
No email address required.
I thought GitHub Copilot just outputs code.
Jump in the discussion.
No email address required.
That's all you need lmao.
Jump in the discussion.
No email address required.
I go to StackOverflow to get answers to specific technical questions/problems about languages, frameworks, systems, etc, not to have code written for me.
Jump in the discussion.
No email address required.
Literally the code is the answer. If you can't figure it out by just the code and output, maybe you should learn to draw or something.
Jump in the discussion.
No email address required.
If you need StackOverflow for general code then you're either a codemonkey or still in college.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
More options
Context
More options
Context
More options
Context
More options
Context
More options
Context
It also fails to produce anything of value outside its data set, which is disappointing. Other than codecel work, it can’t even properly translate across domains.
Ask it to assess the performance of something, even the most obvious stuff and it just completely shits the bed.
Jump in the discussion.
No email address required.
For my line of work, I tried to get it to talk about the latest CVV or hack and it will just talk about how harmful they are but can't say anything of substance.
Krayon sexually assaulted his sister.
![https://i.rdrama.net/images/17156480765435808.webp](https://i.rdrama.net/i/l.webp)
Jump in the discussion.
No email address required.
It can spew out a lot of meaningless babble that makes it obvious half its database is journos
Jump in the discussion.
No email address required.
More options
Context
More options
Context
I use it to respond to student emails. If they want a good answer they should learn not to ask me![:marseyhomofascist: :marseyhomofascist:](/e/marseyhomofascist.webp)
Jump in the discussion.
No email address required.
More options
Context
More options
Context
It's good if you just want fluff with nothing new. If you need to talk about something new out there, then it just craps out.
Krayon sexually assaulted his sister.
![https://i.rdrama.net/images/17156480765435808.webp](https://i.rdrama.net/i/l.webp)
Jump in the discussion.
No email address required.
More options
Context
It's just like me fr
Jump in the discussion.
No email address required.
hella no cap?
Jump in the discussion.
No email address required.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
It's funny when codecels talk about SWE engineers being automated out by GPT because it's all projection. Anyone with real expertise can see instantly where chatGPT lacks.
In fact this should probably be a new interview question: "what do you think of coding being automated by GPT?"
If there's any hint of GPT putting jobs in danger, or using it for anything other than for some boilerplate or autofill, immediate red flag.
Jump in the discussion.
No email address required.
Literally the one thing I like it for so far is generate HTML place holders. Just ask it to spite out an html login page with insert javashitlibs and boom.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
But it knows that it might be wrong. Are you saying you'd ban humans from replying unless they're 100% sure that they're right? Sounds kinda racist, chud.
Jump in the discussion.
No email address required.
More options
Context
proof i go outside: i thought "obscure libraries" was talking about the library of alexandria or the new jersey public library or some shit and you were saying the bot would just make shit up
Jump in the discussion.
No email address required.
lmao that too. If you ask it about incomplete works sometimes it just invents the missing lines
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context