My gut sense is that proof-of-work as a spam deterrent can _improve_ the signal-to-noise ratio, but won't bring it up to an acceptable level.
Say you have a simple proof-of-work protecting some action (posting a comment, voting up/down a story, etc.), and you have its difficulty tuned to allow a median-productivity human on a typical desktop computer to do that action at their typical rate.
A spambot doesn't need to sleep, take days off, and can get illicit access to much more computing power than any one human.
It _would_ probably shift spamming activities to focus on more central, high-value venues, though. hmm.
Say you have a simple proof-of-work protecting some action (posting a comment, voting up/down a story, etc.), and you have its difficulty tuned to allow a median-productivity human on a typical desktop computer to do that action at their typical rate.
A spambot doesn't need to sleep, take days off, and can get illicit access to much more computing power than any one human.
It _would_ probably shift spamming activities to focus on more central, high-value venues, though. hmm.