Regarding 2: Eliezer’s original reply to Roko’s post was ambiguous. The current WP article version claims the correct interpretation is X; Eliezer disputes this, and says he meant Y instead.
The actual meaning, in Eliezer’s telling, is roughly: ‘Roko’s argument is silly, but there seems to be non-negligible probability (here in 2010) that there’s some structurally analogous argument out there that would work, and the field hasn’t had time to explore the space and confirm that there’s no such argument; when we haven’t done that basic due diligence, we shouldn’t go around posting about this on the public Internet.’
Which is easy to round off via a game of telephone to ‘Roko’s basilisk is dangerous’, and thereby to ‘Eliezer thinks Roko’s argument is sound’. But Wikipedia isn’t supposed to be in the business of contributing to games of journalistic telephone. It aspires to be more like the ‘final draft’ of humanity’s knowledge—an extremely vetted encyclopedia, more carefully fact-checked than Britannica, with large conservative boundaries in place to ensure it errs on the side of excluding any information it’s even slightly uncertain about. It obviously hasn’t reached that ideal, since it’s hard to wrangle thousands of volunteer editors spread across millions of articles; but that is what the editorial policies are trying to create.
In this case, both interpretations are perhaps consistent with the text (neither is silly or obviously false). I think a case can be made that Eliezer’s claim about what he had in mind makes sense, there’s no special reason to doubt it, and it can just be reported on, if the topic is noteworthy enough to justify that level of detail.
… But if there is serious published disagreement about whether Eliezer is making stuff up here, then: In cases like this, as a matter of policy, Wikipedia avoids doing original research / detective work into sussing out that Eliezer’s lying and his real meaning was something totally different from what he claims it was. Instead, if the topic is important enough to cover in detail at all, you quote the source text and then report on noteworthy claims about what that text means.
Regarding 3: This wouldn’t necessarily be out of place in a long Wikipedia article that includes a bunch of other quotes, so long as the article is detailed enough to provide any context for why Auerbach thinks any of those things. (As is, someone reading WP to learn about LW for the first time won’t know what any of this has to do with ‘transcending conventional morality’ or ‘messianic ambitions’; bringing all this up without context is just bad writing.)
However, on its own, this is an obviously axe-grindy choice of quote that makes the Wikipedia article look like an op-ed that’s trying to support a critique of LW, rather than looking like an encyclopedia.
Regarding 2: Eliezer’s original reply to Roko’s post was ambiguous. The current WP article version claims the correct interpretation is X; Eliezer disputes this, and says he meant Y instead.
The actual meaning, in Eliezer’s telling, is roughly: ‘Roko’s argument is silly, but there seems to be non-negligible probability (here in 2010) that there’s some structurally analogous argument out there that would work, and the field hasn’t had time to explore the space and confirm that there’s no such argument; when we haven’t done that basic due diligence, we shouldn’t go around posting about this on the public Internet.’
Which is easy to round off via a game of telephone to ‘Roko’s basilisk is dangerous’, and thereby to ‘Eliezer thinks Roko’s argument is sound’. But Wikipedia isn’t supposed to be in the business of contributing to games of journalistic telephone. It aspires to be more like the ‘final draft’ of humanity’s knowledge—an extremely vetted encyclopedia, more carefully fact-checked than Britannica, with large conservative boundaries in place to ensure it errs on the side of excluding any information it’s even slightly uncertain about. It obviously hasn’t reached that ideal, since it’s hard to wrangle thousands of volunteer editors spread across millions of articles; but that is what the editorial policies are trying to create.
In this case, both interpretations are perhaps consistent with the text (neither is silly or obviously false). I think a case can be made that Eliezer’s claim about what he had in mind makes sense, there’s no special reason to doubt it, and it can just be reported on, if the topic is noteworthy enough to justify that level of detail.
… But if there is serious published disagreement about whether Eliezer is making stuff up here, then: In cases like this, as a matter of policy, Wikipedia avoids doing original research / detective work into sussing out that Eliezer’s lying and his real meaning was something totally different from what he claims it was. Instead, if the topic is important enough to cover in detail at all, you quote the source text and then report on noteworthy claims about what that text means.
_____________________________________________________
Regarding 3: This wouldn’t necessarily be out of place in a long Wikipedia article that includes a bunch of other quotes, so long as the article is detailed enough to provide any context for why Auerbach thinks any of those things. (As is, someone reading WP to learn about LW for the first time won’t know what any of this has to do with ‘transcending conventional morality’ or ‘messianic ambitions’; bringing all this up without context is just bad writing.)
However, on its own, this is an obviously axe-grindy choice of quote that makes the Wikipedia article look like an op-ed that’s trying to support a critique of LW, rather than looking like an encyclopedia.