i'd be interested if you have a specific example of an ai safety argument that depends on knowing the ai's architecture
-
-
Replying to @VesselOfSpirit @simpolism
as for the aliens below the moon thing: artificial general intelligence seems possible because humans are possible
1 reply 0 retweets 3 likes -
Replying to @VesselOfSpirit @simpolism
and it would help achieve a lot of different goals, which means people will be motivated to try to invent it
1 reply 0 retweets 3 likes -
Replying to @VesselOfSpirit @simpolism
also, sometimes when you don't know what a thing will be like there are a few possibilities and you can study each of them
1 reply 0 retweets 2 likes -
Replying to @VesselOfSpirit
simpolism Retweeted sol
1) re "intelligent thing risk" my new fave take is https://twitter.com/fire__exit/status/1011762984136445953 … 2) humans took millennia to appear and were shaped by long-term environments. we're motivated to make virtual humans but it's gonna be messy & I think Moore's Law is over 3) "study"?
"sound alarm"? 
simpolism added,
1 reply 0 retweets 1 like -
Replying to @simpolism @VesselOfSpirit
I think a major point I deeply disagree with is that I don't think Moore's Law will hold. A lot of AI-related fear seems to assume a continuous creep in computing power. How would AI-risk research change if we were forever stuck with the computing power of our current hardware?
3 replies 0 retweets 0 likes -
Replying to @simpolism @VesselOfSpirit
My current theorizing is that: - the brain is the most efficient physical form for human-like intelligence - the most likely "general AI" is a large brain/"biological machine" - there are deep unknown-unknowns re how this might come to exist and what forms of agency it will have
1 reply 0 retweets 0 likes -
Replying to @simpolism
"most efficient physical form" seems really unlikely to me given how many constraints evolution has been under
3 replies 0 retweets 0 likes -
Replying to @VesselOfSpirit @simpolism
it's like how trains don't have legs
1 reply 0 retweets 0 likes -
Replying to @VesselOfSpirit
by the metric of "speed" a train is more efficient than legs, but if your metric is "turn speed" then legs are more efficient. if your metric is "steps per second" then a train's efficiency is "mu"
1 reply 0 retweets 1 like
the goal isn't to imitate humans but to act/think effectively. more like speed than turn speed or steps per second
-
-
Replying to @VesselOfSpirit
what does it mean for an action or a thought to be "effective"?
2 replies 0 retweets 0 likes -
Replying to @simpolism @VesselOfSpirit
Not speed, nor turn speed, but time through the maze. For some sense of "through" anyway.
1 reply 0 retweets 0 likes - 6 more replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.