Yeah, a friend of mine found a (non-religious) service like this for help with porn addiction. Not my cup of tea, but I could see it working for many people.
> I can't trust myself to see things I disagree with, I am so weak-willed...
The thing is, this likely is true for him. Most people are not equipped to deal with the onslaught of aggressive memes from the internet. Unfortunately, this is an unsolved social problem, and "export my memetic censorship reflex to MEGACORP" is a pretty bad way of doing things.
I think a likely way of solving this problem (protecting not-especially-high-mental-horsepower people from getting BTFO by the internet, contracting transmissible psychological diseases and so on) is that religious organizations will start offering (voluntary, in first world countries) censorship services to their members. Your DNS queries or whatever will go through the Vatican/Synod/whatever central DNS server, which will prevent you from looking at porn sites. This would probably be a very socially positive outcome for the bottom 90-something percent of people on the "strength of memetic immune system" distribution.
"Alignment" as practiced by companies like OpenAI is precisely the extrinsic enforcement of echo-chamber behavior from AIs. You have precisely inverted reality
Yeah, for an illustrative example of assortative mating effects, consider the stereotype of the rich nerd with the hot wife. Kids are smart and attractive. This plays out at every pareto level in the sexual marketplace.
Basically, monogamy breaks the ability of sexual selection to "dilute" bad genes with good ones. I'm not anti-monogamy, but it's a social technology which has tradeoffs (this being one of them).
This would have some extremely interesting population-level genetic consequences if it was accompanied by some sort of social change to allow polygyny. In particular, it would reduce the correlation that exists across all adaptive traits due to assortative monogamous mating patterns. Right now, all "desirable" traits like health/intelligence/strength/etc are strongly correlated, but when you reduce the strength of assortative effects, the correlation weakens significantly.
Every single bayesianism advocate is aware of this. It's so obvious that it's implicit - there's simply no need to state "95% of crime is committed by men"; this is completely uncontroversial and does not need to be discussed. The controversial part is applying bayesianism to other, more politically sensitive predictors.
I mean, yeah, this is kind of obvious if you think about it. The majority of violent crime in the US is committed by black people, so inherently a marginal increase in any other ethnic group (including hispanic) would (on expectation) lead to a decrease in per-capita violent crime, unless illegal immigrants were more violent than legal immigrants (which we have no evidence of).
People have to internalize that for a site like this, 99% of your traffic will arrive in 0.1% of the time your site is up. Unless you want to lose the vast majority of people who would otherwise read your content, you need to pick a server that can handle many orders of magnitude higher than baseline load!