I'm the ethics exec for the University of Victoria AI Club. Most of our members are more concerned with short term / non-RSI / non-SAGI issues, but I am trying to build community focus both on short & "long" term issues.
#Victoria, BC, Canada
quick saffron
We mostly communicate on discord, you can find a link at uvicai.ca
Btw, does anyone have a good alternative to "short and long term" to mean relating to RSI/AGI/SAI or not?
flint wigeon
Btw, does anyone have a good alternative to "short and long term" to mean relati...
(I'm not in your area, but I want to answer your question.)
When speaking with a mixed audience, I would say "immediate risks" and "existential risks" (with a bonus of "current harms" for the things that are already happening).
If I'm speaking just with x-risk-aware people, it's a little easier to get away with "mundane risks" and "x-risks."
It's also technically correct to say "known risks" and "theoretical risks," but the public may parse "theoretical" to mean "not real."
(In the weird timeline we are in, though, strategic deception, situational awareness during training and testing, and power-seeking behavior are all known risks that have been observed in lab settings. 
)
quick saffron
"current harms", "immediate risks" and "existential risks"... yeah, that describes it really well. Thanks : )
" "... yeah... I'm not sure if I should be updating to "maybe we can control deceptive sub-super-critical AI & the situation is better" or "oh no ppl are really messing this up badly & the situation is worse"
flint wigeon
Yep. I for one am grateful that we don't appear to be in a fast takeoff. Or, if the takeoff will be fast, there appears to be significant room for bad things to happen first before anyone triggers an RSI doom loop.
So 
because instrumental convergence appears to be correct, but 
because we are still alive to watch it happen, and it's one more thing to sound an alarm about.