Hey Eliezer, your explanations are incredible but I don’t understand why we are folding the amplitude distribution over that diagonal line. I do understand that the folded distribution is unfactorable, but what is the motivation behind the folding operation?
Liron
Karma: 4,080
Even though the decision is overdetermined, I’ve been cryocrastinating. I’ll schedule it with more urgency.
I expect it is an easy distinction for most people whom Eliezer describes as “highly intelligent”.
Eliezer’s post focuses on the distinction between two concepts a person can believe (hereby called “narratives”):
“God is real.”
“I have something that qualifies as a ‘belief in God’.”
Either narrative will be associated with positive things in the person’s mind. And the person, particularly with narrative #2, often forms a meta-narrative:
3. “My belief in God has positive effects in my life.”
But: Unlike the meta-narrative, our analysis should not proceed as if the relationship between narrative and effects is a simple causal link.
The actual cognitive process that determines the narrative might go something like this:
Notice that the desirable aspects of life enjoyed by religious people in the community conflict with undesirable properties (e.g. falsehood, silliness, uselessness) of religious beliefs.
Trigger a search: “How do I make the undesirable properties go away while keeping benefits?”
Settle on a local optimum way of thinking, according to some evaluation algorithm that is attracted by predictions of certain consequences and repulsed by others.
The search can have a very different character from one individual to another. For example, if the idea of not having a defensible narrative isn’t repulsive, then the person says: “I’m happy in my religious community, so I don’t think too hard about my religion.” The kind of thing they are actually repulsed by would be “for me or my peers to believe that I am not a fully committed member of my in-group”.
Or, if the person is given to conscious reasoning, then it would be extremely repulsive to not have a defensible narrative. What their search evaluation algorithm is actually repulsed by might be something like, “the self-doubt that I am not a capable reasoner”, or “the loss of respect and status among other intellectuals”. So the quick fix is: Add more layers of justification and arguments surrounding religion, so that both you and your peers can plausibly feel that you are a capable reasoner occupying a justifyable stance on a complex issue.
So regarding Eliezer’s post, it’s not surprising that someone with narrative #2 can get a “placebo” version of the positive effects that come with narrative #1. The narrative doesn’t independently cause the positive effects; the narrative is shaped by a cognitive algorithm that predicts the benefits of believing it.
This post is also a followup to Beware “I Believe”. Here is what I’ve learned.
Thinking about “believing in X” triggers positive affect, so one says “I believe in X”. The process that forms the “I believe in” thoughts is separate from the process that analyzes the content of propositions about the territory.
The “I believe in” process can really mess with one’s map. This happens in two ways:
It sticks post-it notes over sections of their already-formed map that sever those sections’ coherence with the rest of the map, e.g. a Christian who believes evolution happened but in various points God came in and did stuff that’s responsible for morality.
It tampers with bias-ridden fragile belief-forming methodology. They might do some transformation to incoming explanations of events so as to reconcile with “I believe in X”, which probably won’t increase its entanglement with the territory.
And as Eliezer pointed out in a comment to Robin’s post, the interference between the separate belief and “belief in” processes is paralleled by the confusing English word “believe” which refers to both processes. And there is no adequate synonym for saying something like “I believe in democracy.”
You should attack the bad links in the causal chain that lead to absurd conclusions, not the conclusions themselves.
So yes, it’s worth attacking the bible’s nonchalance about contradictions—but don’t bother dwelling on the contradictions themselves.
We know.
Sure, when the dust settles, it could turn out that apples don’t exist, Earth doesn’t exist, reality doesn’t exist. But the nonexistent apples will still fall toward the nonexistent ground at a meaningless rate of 9.8 m/s2.
This is shameless Karma whoring. We should ban this user.
You’re thinking of “optimization”.
I read The Phantom Tollbooth as a 10-year-old, and thoroughly enjoyed it, but all such lessons went over my head at the time.
Unlikely. I really didn’t think about stories on an abstract level back then, and “Demons of Ignorance” doesn’t ring a bell now as being part of the story.
Indeed! I pay attention both to gender pronouns and to Eliezer’s writing patterns, and I never noticed this. (Eliezer_2000 used “ve” a lot though.)
I had previously decided on “he” in order to optimize for flow, but I am happy to accept this well-made point and switch to “they’.
Are buses safer than cars? For one thing, they don’t have seat belts.
The “textbooks” link is broken.
Yes we know “right”’s relationship to the territory has to do with the complexites of the brain. But the entire “right” module can still be part of a reduction.
Tabooing is useful for:
Making sure your reduction attempt isn’t circular
Figuring out what a speaker actually means when they use a word whose referent is ambiguous (like “make a sound”) or just points to their own confusion (like Searle’s “semantics”)
No need for it ere.
This is a really good post. I particularly like the suggestion that we don’t have to infer and cache conclusions about ourselves when we screw up and don’t return a library book. (Of course, other people would be rational to cache a conclusion about us because thinking differently wouldn’t be a self-fulfilling prophecy.)
Eliezer two-boxes on Newcomb’s problem, and both boxes contain money.
Angering Eliezer Yudkowsky is a global existential risk
Where’s the punch line?
Hey Eliezer, this is a great post. I just have one question: HOW ARE YOU SO AWESOME?! Seriously, these posts are incredible.