Data Source: BigQuery public dataset (bigquery-public-data.stackoverflow), Stack Exchange API (api.stackexchange.com/2.3)
Tools: Pandas, BigQuery, Bruin, Streamlit, Altair
Trollercoaster101 on
It is funny how the LLMs still needed stackoverflow to get training and then killed it as a thank you gift.
TOO_MUCH_BRAVERY on
Actually a big problem. Soon troubleshooting knowledge will all be proprietary training data accessible though an LLM subscription.
sssarel on
How can a decline from 200k to 110k questions before the event be called a plateau? At best the plateau should end at 2018, there is a clear downward trend from there, that accelerates further in 2023.
whaaatcrazy on
Curious if this will reduce overall questions to ones that aren’t easily answered making more complicated ones get more visibility
TheGacAttack on
So people couldn’t ever find the Search feature, but they can find Chat GPT.
I guess people just enjoy the act of asking.
Brighter_rocks on
The decline clearly started years before ChatGPT – 2022 just accelerated an already downward trend
Weshtonio on
It looks like it is directly correlated to the Ubisoft stock price.
Faraway-Fire on
It’s not even about asking AI the questions any more. No longer need to. GitHub Copilot (trained on millions of repos) now just corrects / predicts your code directly.
Cut out this middleman need to even ask a question.
Will be interesting to see a version of this graph for Software Engineering stack exchange, where it’s more about the discussions around the edges rather than raw code.
Justryan95 on
For something like a „plateau“ it sure looked like a decline.
DManeOne on
SO is one of the web’s most toxic sites for being a user. It is much more effective to speak with an agent than a bunch of passive aggressive neckbeards
HaroerHaktak on
There are examples showing how stack overflow is a toxic environment and asking even a simple question will get you instantly banned lol.
Leave A Reply
Du musst angemeldet sein, um einen Kommentar abzugeben.
12 Kommentare
Data Source: BigQuery public dataset (bigquery-public-data.stackoverflow), Stack Exchange API (api.stackexchange.com/2.3)
Tools: Pandas, BigQuery, Bruin, Streamlit, Altair
It is funny how the LLMs still needed stackoverflow to get training and then killed it as a thank you gift.
Actually a big problem. Soon troubleshooting knowledge will all be proprietary training data accessible though an LLM subscription.
How can a decline from 200k to 110k questions before the event be called a plateau? At best the plateau should end at 2018, there is a clear downward trend from there, that accelerates further in 2023.
Curious if this will reduce overall questions to ones that aren’t easily answered making more complicated ones get more visibility
So people couldn’t ever find the Search feature, but they can find Chat GPT.
I guess people just enjoy the act of asking.
The decline clearly started years before ChatGPT – 2022 just accelerated an already downward trend
It looks like it is directly correlated to the Ubisoft stock price.
It’s not even about asking AI the questions any more. No longer need to. GitHub Copilot (trained on millions of repos) now just corrects / predicts your code directly.
Cut out this middleman need to even ask a question.
Will be interesting to see a version of this graph for Software Engineering stack exchange, where it’s more about the discussions around the edges rather than raw code.
For something like a „plateau“ it sure looked like a decline.
SO is one of the web’s most toxic sites for being a user. It is much more effective to speak with an agent than a bunch of passive aggressive neckbeards
There are examples showing how stack overflow is a toxic environment and asking even a simple question will get you instantly banned lol.