This might not be that bad if LLMs didn’t hallucinate. At least SO had a community to audit fraudulent answers.
pull down to refresh
121 sats \ 1 reply \ @cointastical 16h
That measures "questions asked" (i.e., new question posts). Why ask a question when a search returns the answer?
It's not just AI that made Stack Overflow die, it's
- having to press the "Accept Cookies" prompt,
- the "Are you a human being?" CAPTCHAs
- the Questions showing the Green Checkmark (i.e., Answered) and the answer was: "Figured it out, guys .... thanks" (and nothing more).
reply
42 sats \ 0 replies \ @k00b OP 16h
Great points. I had similar thoughts about the "corpus" of questions/answers already being formed and naturally leading to a tapering of new ones.
reply
226 sats \ 4 replies \ @kepford 9 Jun
I have a feeling the answer to LLM making stuff up is the same as how StackOverflow. People correcting it. These companies and chatbots have to use feedback to fix the mistakes. I have watched different chatbots make stuff up and when I call it out it admits it. It's pretty comical to watch.
reply
98 sats \ 1 reply \ @SimpleStacker 9 Jun
The problem with this is that the chatbots can be gaslit into agreeing with you even if they're right and you're wrong
Some crusty old dude on stack overflow would never be gaslit like that
reply
0 sats \ 0 replies \ @istealkids 9 Jun
exactly, they are better manipulators than Geobbels
reply
132 sats \ 1 reply \ @k00b OP 9 Jun
I have a friend that asks ChatGPT stuff and goes into a rage based on the answers. When I’m like “I’m not sure that’s true” he’s like “well it’s hallucinating less now” and goes on being upset.
reply
31 sats \ 0 replies \ @kepford 9 Jun
Ha
reply
36 sats \ 1 reply \ @sox 22h
Something that's really sad is that not only LLMs are decreasing Stack Overflow's popularity, they're also decreasing its posts quality.
You have to search before 2023 to have more chance to find quality answers.
reply
10 sats \ 0 replies \ @teemupleb 19h
The GPT-fication eats and dumbs down all online content and it’s like a cancer trying to kill original creativity but in the end, it’s all cyclical and humanity will prevail.
reply
21 sats \ 0 replies \ @WeAreAllSatoshi 9 Jun
This makes me very sad
reply
10 sats \ 0 replies \ @istealkids 9 Jun
ngl, stackoverflow does have some meritorious people who do have potential.
But earlier, I used to see gpt answers in only english, story and bug fixes sites, now noobs are getting into places like infosec and AI, with their gpt-fied answers
don't know whom to trust and whom not to.
the only space I know that these AI-noobs have still not captured fully is Quora probably.
reply
84 sats \ 0 replies \ @nathanael 9 Jun
To be fair, the decline had already begun before ChatGPT launched. LLMs likely accelerated this decline, and COVID-19 only paused it temporarily.
I think it's more important than ever to ask ourselves: Is this true? How do I know? What are my sources? However, most people probably don't have time to ask these questions and must rely on trusted sources—as we all do at some point.
I hope SN doesn't meet the same fate as SO. I'm here on SN for now, though who knows what the future holds.
reply
15 sats \ 0 replies \ @realBitcoinDog 9 Jun
“The good old days”
reply