I have little idea why Eliezer et al think they can give people anything more than social connections and typical self-help improvements...

Less Wrong has been created with the goal in mind of getting people to support SIAI.

Less Wrong is mainly a set of beliefs and arguments selected for their persuasiveness in convincing people that creating friendly AI is of utmost importance (follow the link above if you think I am wrong).

The two hallmarks of LessWrong are H&B and Bayesian probability: the latter is often abused...

I believe that the most abused field is artificial intelligence. The ratio of evidence to claims about artificial intelligence is extremely low.