DizzyBoba
8mo
So does this mean LLMs will be dumber in the future because human solutions to problems from sites like stackoverflow have started to disappear, and there will be no data to steal solutions from, to hard problems?
FluffyKoala
8mo
Yup, exactly. And also, the internet is filling up with AI-generated slop, and model collapse of LLMs trained on their own output is a known problem with LLMs, and that will start degrading these models when trained on the future internet data.
The future of LLMs, internet, and humanity, all look bleak 😅
WigglyUnicorn
8mo
That's a chart of human creativity going down as well
Discover more
Curated from across
