4 points | by 01-_- 9 hours ago
6 comments
How could a company like Stack Overflow reinvent itself in order to survive the crisis it is currently experiencing with regard to AI?
Stack Overflow was in serious trouble well before AI came along; the moderators killed it years ago. LLMs are merely accelerating its decline.
They should have started thinking two years ago
a dead end loop, apparently, applying not just to Stack Overflow but all similar content providers:
A. AI became "Intelligent" thanks to Stack Overflow
B. Stack Overflow collapse due to AI
C. AI cannot update/evolve due to lack of Stack Overflow
Who is going to produce mediated and revised ('good') content?
> C. AI cannot update/evolve due to lack of Stack Overflow
> Who is going to produce mediated and revised ('good') content?
This take is ~3 years old and already obsolete. For the past two years RL flavours (RLAIF, RLVR, etc) have been driving the new models.
Human trainers. There are tens of thousands of people doing it full time for AI companies. Their mode of work isn't much dissimilar from answering SO questions. Just in a closed manner.
How could a company like Stack Overflow reinvent itself in order to survive the crisis it is currently experiencing with regard to AI?
Stack Overflow was in serious trouble well before AI came along; the moderators killed it years ago. LLMs are merely accelerating its decline.
They should have started thinking two years ago
a dead end loop, apparently, applying not just to Stack Overflow but all similar content providers:
A. AI became "Intelligent" thanks to Stack Overflow
B. Stack Overflow collapse due to AI
C. AI cannot update/evolve due to lack of Stack Overflow
Who is going to produce mediated and revised ('good') content?
> C. AI cannot update/evolve due to lack of Stack Overflow
> Who is going to produce mediated and revised ('good') content?
This take is ~3 years old and already obsolete. For the past two years RL flavours (RLAIF, RLVR, etc) have been driving the new models.
Human trainers. There are tens of thousands of people doing it full time for AI companies. Their mode of work isn't much dissimilar from answering SO questions. Just in a closed manner.