Using that mental framework, I take Gabriels main point to be that sometimes factual disagreement is vital and in that case communicating it as clearly as possible is important and should take precedence over optimizing for niceness in form.
@Gabriel: does that match what you wanted to communicate?
Seems to me that "being nice" has multiple possible scopes and this is the generator of the disagreement.
I can be both nice in the form and/or content of my speech, as aphyer points out.
Gabriel seems to include "not inconveniencing others through factual disagreement" in "bing nice", while you (and I think Scott, too) exclude it.
From what I understand I would describe the skill Steven points to as "autonomously and persistently learning at deploy time".
How would you feel about calling systems that posess this ability "self-refining intelligences"?
I think mastery, as Nathan comments above, is a potential outcome of employing this ability rather than the skill/ability itself.