/lit/ - Literature

>>22051003
The good news is we're about 60 years behind on interpretability, let alone robust alignment. By far the most likely outcome is that everyone dies a quick and mostly painless death.

>>20567460
If we create a superintelligence, it's possible that the alignment problem is gotten slightly wrong. It could implement religious hells, or implement the opposite of humanity's utility function.

https://reducing-suffering.org/near-miss/