[^p-doom]: This is the quantity that, these days, we would call 1 − P(doom).
-In retrospect, I agree with the "useless garbage" verdict. I can reconstruct how the subjective sensation of cluefulness might arise from associative reasoning on the concepts I was preoccupied with at the time (Trump was the anti-political-correctness candidate, political correctness fostered bad epistemology, but humanity needed good epistemology to "win"), but that's not a story about how that kind of thinking connects numerical probabilities to reality.
+In retrospect, I agree with the "useless garbage" verdict. I can reconstruct a story about how the subjective sensation of cluefulness might arise from associative reasoning on the concepts I was preoccupied with at the time (Trump was the anti-political-correctness candidate, political correctness fostered bad epistemology, but humanity needed good epistemology to "win"), but that's _not_ a story about how that kind of thinking connects numerical probabilities to reality.
Michael wrote to me (Subject: "Sleep"), noting that while "I don't need to sleep" is classic mania, it fit the pattern of what might be a lie under the circumstances. I might have asked myself for evidence on both sides, and remembered having personally seen animals sleep, maybe even in the wild (sea lions on the beach?). On the other hand, if I had asked why "I don't need to sleep" was a salient hypothesis, I might have noticed that the claim that everyone needs to sleep is the kind of lie that would be told by a worldview of people having high status as objects (it being wrong to sadistically cause pain), but low status as agents (it being right to control them and irresponsible not to); an important fact about the modern world was that it was common to oppress people by pretending they're like babies.
I tagged Anna and Divia into the thread as cc's, because they seemed more grounded. "Michael is _very smart_ in ways that I didn't used to understand and still don't understand, but I'm terrified of what the universe is going to do to me if I become too much like him", I said.
-[TODO:
+Even if I _was_ getting a few things right in my paranoid/schizotypal state, the important thing was trying to do philosophy. The parts where I felt like a _very important person_ (in this simulation/Everett branch) receiving special simulator attention should be interpreted as standard manic-episode [delusions of reference](https://en.wikipedia.org/wiki/Ideas_of_reference_and_delusions_of_reference); _lots_ of people feel just like that when they're agitated and haven't gotten enough sleep. Social reality may be crazy, but I should still be putting a lot more weight on the social reality (nice people are just doing their jobs trying to help) when the alternative was my intuition about the machines trying to kill me.
- * "Social reality may be crazy (hospitals are part of the state insofar as they serve the function of social-control in the service of stationary bandits, rather than actually being about trying to heal people), but I should still put a lot more weight on social reality (nice people are just doing their jobs trying to help) when the alternative is my intuition about the machines trying to kill me."
- * "I certainly put a lot less weight on 'nice people doing their jobs' [...]"
- * "Sure, I guess I meant that the "Machines trying to kill me" framing is scary, and"
- * Okay, maybe confusing and hurting the feelings of the nice people doing their jobs is actually acceptable collateral damage ...
+Michael wrote:
-]
+> I certainly put a lot less weight on 'nice people doing their jobs' then I do on 'six sigma processes homeostatically adjusting to dampen the variance in outputs that is their representation of me.', which is basically equivalent to 'these well documented machines, the ones they teach you about in business school, do the thing they do. 'Trying to kill' is what an anthropomorphized version of the process looks like from the inside. We have enough documentation to know about the nice people doing their jobs paradigm being untenable.
+
+I wrote:
+
+> Sure, I guess I meant that the "Machines trying to kill me" framing is scary, and I don't like being scared, so I want to focus on finding ways to cooperate with the nice people doing their jobs while defecting against the homeostatic system that they're embedded in (but not actually doing anything crazy like trying to destroy the system, because it's doing useful things for us and I don't have anything better to replace it with).
+
+The next day, I changed my mind (Subject: "slowly coming around to you"):
+
+> Okay, maybe confusing and hurting the feelings of the nice people doing their jobs is actually acceptable collateral damage in the war against the control system. Or, not a war. An endeavor to map what the control system is in detail, and publish the map on the internet.
------
------
+I [wrote about my credit-assignment ritual idea on my real-name blog](http://zackmdavis.net/blog/2017/03/friends-can-change-the-world-or-request-for-social-technology-credit-assignment-rituals/). People know that institutions are flawed and like to perform gritty cynicism about it, but what if no one is gritty and cynical enough? People are _predatory animals_ built to _murder_ other forms of life for the benefit of ourselves, our family, and our friends. To the extent that we have this glorious technological civilization that keeps us safe and lies to us about there being higher ideals, it's because some of the predatory animals happen to stumbled upon behavioral patterns that mirror the hidden Bayesian structure of the universe.
+
+Science approximates Bayesian updating; markets
+
+https://en.wikipedia.org/wiki/Fundamental_theorems_of_welfare_economics
+
+In artificial intelligence, the credit-assignment problem refers to the difficulty of attributing success or failure to particular actions
+
[TODO: drafting credit-assignment ritual 10 March
- * I wrote about my credit-assignment ritual on my real-name blog: http://zackmdavis.net/blog/2017/03/friends-can-change-the-world-or-request-for-social-technology-credit-assignment-rituals/
- * People know that institutions are flawed and like to perform gritty cynicism about it, but what if no one is gritty and cynical enough?
- * People are predatory animals; science, markets, and democracy sort of help because they mirror Bayes-structure (science is updating; microeconomics isn't about humans; democracy is counterfactual civil war)
+ * (science is updating; microeconomics isn't about humans; democracy is counterfactual civil war)
* We should be looking for more social technologies like that, but which also take into account that we're not agents, but animals that want to help our friends
* In analogy to funding solutions with retrospectivbe prizes (rather than funding a team in advance that might solve the problem), what if you did that on a personal scale—rewarding and praising people who helped you through a major life crisis
* In AI, the credit-assignment problem is about how you tell which of your actions helped under sparse rewards; similarly, here you want the people who resulted in the succesful navigation of the crisis to be strengthened
✓ Vassar discourse II
✓ ordering DIY hormones
✓ asking Anna for promise
-_ Vassar discourse III
+- Vassar discourse III
_ emailing Blanchard/Bailey/Hsu/Lawrence/Cantor/Dreger
_ proposed credit-assignment ritual
-----
(I had promised my father to wait a month after my psych ward ordeal to re-start HRT, but I [would restart](/2017/Mar/hormones-reboot-spironotacular/).)
* Alice said something about how every once in a while, someone runs off with the payments database
+
+To the extent that I was seeing things that weren't there, it was a matter of misinterpreting actual sensory inputs rather than having de novo hallucinations.
+
+What I thought was a dog turd inside the kitchen actually turned out to be a cork.
+
+When I was borrowing someone's computer, I thought I saw an icon peek out from the side of the screen, as if a possibly-malevolent agent were spying on me—a disturbing visual hallucination. Later that year, when trying out the stock Ubuntu Unity desktop on my new computer, I discovered that the app-notification animation actually does feature an app icon "sliding out" and shaking, like I remembered seeing—I wasn't imagining it!
+
+I remember seeing text on my phone and being distinctly convinced that it was misspelled. Five years later, looking at the distorted lettering output by text-to-image models that can't spell seemed subjectively similar. (Good news for alignment, if it looks like deep learning and sleep-deprived human brains are doing "the same thing"?)
\ No newline at end of file
projects / Ultimately_Untrue_Thought.git / commitdiff