check in

[Ultimately_Untrue_Thought.git] / content / 2020 / memento-mori.md

diff --git a/content/2020/memento-mori.md b/content/2020/memento-mori.md
index 420908a..13571df 100644 (file)
--- a/content/2020/memento-mori.md
+++ b/content/2020/memento-mori.md
@@ -38,7 +38,7 @@ The parenthetical I deleted was:
 
 It probably wouldn't have mattered either way, with so many messages flying by in the chat. In some ways, Blue Egregore is less like an ideology and more like a regular expression filter: you can get surprisingly far into discussing the actual substance of ideas as long as no one _says a bad word_ like "eugenics".
 
 
 It probably wouldn't have mattered either way, with so many messages flying by in the chat. In some ways, Blue Egregore is less like an ideology and more like a regular expression filter: you can get surprisingly far into discussing the actual substance of ideas as long as no one _says a bad word_ like "eugenics".
 

-—if we even have enough _time_ for things like embryo selection to help, if AI research somehow keeps plodding along [even as everything _else_ falls apart](https://www.unqualified-reservations.org/2007/05/antisingularity/). The [GPT-3 demos](https://www.gwern.net/GPT-3) have been tickling my neuroticism. Sure, it's "just" a language model, doing nothing more but predicting the next token of human-generated text. But [you can do a lot with language](https://bmk.sh/2020/08/17/Building-AGI-Using-Language-Models/). As _disgusted_ as I am with my robot cult as presently constituted, the _argument_ for why you should fear the coming robot apocalypse in which all will be consumed in a cloud of tiny molecular paperclips, still looks solid. But I had always thought of it as a long-term thing—this unspoken sense of, okay, we're probably all going to die, but that'll probably be in, like, 2060 or whatever. People freaking out about it coming _soon_-soon are probably just [following the gradient into](https://www.lesswrong.com/posts/yEjaj7PWacno5EvWa/every-cause-wants-to-be-a-cult) being [a doomsday cult](https://unstableontology.com/2019/07/11/the-ai-timelines-scam/). Now the threat, [and the uncertainty around it](https://intelligence.org/2017/10/13/fire-alarm/), feel more real—like maybe we'll all die in 2035 instead of 2060.
+<a id="if-we-even-have-enough-time"></a>—if we even have enough _time_ for things like embryo selection to help, if AI research somehow keeps plodding along [even as everything _else_ falls apart](https://www.unqualified-reservations.org/2007/05/antisingularity/). The [GPT-3 demos](https://www.gwern.net/GPT-3) have been tickling my neuroticism. Sure, it's "just" a language model, doing nothing more but predicting the next token of human-generated text. But [you can do a lot with language](https://bmk.sh/2020/08/17/Building-AGI-Using-Language-Models/). As _disgusted_ as I am with my robot cult as presently constituted, the _argument_ for why you should fear the coming robot apocalypse in which all will be consumed in a cloud of tiny molecular paperclips, still looks solid. But I had always thought of it as a long-term thing—this unspoken sense of, okay, we're probably all going to die, but that'll probably be in, like, 2060 or whatever. People freaking out about it coming _soon_-soon are probably just [following the gradient into](https://www.lesswrong.com/posts/yEjaj7PWacno5EvWa/every-cause-wants-to-be-a-cult) being [a doomsday cult](https://unstableontology.com/2019/07/11/the-ai-timelines-scam/). Now the threat, [and the uncertainty around it](https://intelligence.org/2017/10/13/fire-alarm/), feel more real—like maybe we'll all die in 2035 instead of 2060.

 
 At some point, I should write a post on the causes and consequences of the psychological traits of fictional characters not matching the real-life distributions by demographic. [The new _Star Trek_ cartoon](https://memory-alpha.fandom.com/wiki/Star_Trek:_Lower_Decks) is not very good, but I'm obligated to enjoy it anyway out of brand loyalty. One of the main characters, Ens. Beckett Mariner, is [brash and boisterous and dominant](https://youtu.be/64obsPsXxkE?t=45)—[friendly, but in a way](https://www.youtube.com/watch?v=7CVJGy0Do5I) that makes it clear that she's _on top_. If you've seen _Rick and Morty_, her relationship with Ens. Brad Boimler has the Rick and Morty dynamic, with Mariner as Rick. ([Series creator](https://en.wikipedia.org/wiki/Mike_McMahan) Mike McMahan actually worked on _Rick and Morty_, so it likely _is_ the same dynamic, not just superficially, but generated by the same algorithm in McMahan's head.)
 
 
 At some point, I should write a post on the causes and consequences of the psychological traits of fictional characters not matching the real-life distributions by demographic. [The new _Star Trek_ cartoon](https://memory-alpha.fandom.com/wiki/Star_Trek:_Lower_Decks) is not very good, but I'm obligated to enjoy it anyway out of brand loyalty. One of the main characters, Ens. Beckett Mariner, is [brash and boisterous and dominant](https://youtu.be/64obsPsXxkE?t=45)—[friendly, but in a way](https://www.youtube.com/watch?v=7CVJGy0Do5I) that makes it clear that she's _on top_. If you've seen _Rick and Morty_, her relationship with Ens. Brad Boimler has the Rick and Morty dynamic, with Mariner as Rick. ([Series creator](https://en.wikipedia.org/wiki/Mike_McMahan) Mike McMahan actually worked on _Rick and Morty_, so it likely _is_ the same dynamic, not just superficially, but generated by the same algorithm in McMahan's head.)
 


@@ -46,6 +46,6 @@ Overall, I'm left with this uncanny feeling that Mariner is ... not drawn from t
 
 I wish I knew _in more detail_ what my brain thinks it's picking up on here? (I could always be wrong.) It's important that I use the word _distribution_ everywhere; I'm at least definitely not being one of those _statistically-illiterate_ sexists. Most men _also_ don't have that kind or degree of boisterous dominance; my surprise is a matter of ratios in the right tail.
 
 
 I wish I knew _in more detail_ what my brain thinks it's picking up on here? (I could always be wrong.) It's important that I use the word _distribution_ everywhere; I'm at least definitely not being one of those _statistically-illiterate_ sexists. Most men _also_ don't have that kind or degree of boisterous dominance; my surprise is a matter of ratios in the right tail.
 

-I wish there was some way I could get a chance to explain to all _my people_ still under the Egregore's control, what _should_ be common knowledge too obvious to mention—that Bayesian _surprise_ is not moral _disapproval_. Beckett Mariner _deserves to exist_. (And, incidentally, I deserve the chance to _be_ her.) But I think the way you realistically _get_ starships and StyleGAN—and survive—is going to require an uncompromising focus on the kind of technical excellence that can explain in mathematical detail what black-box abstractions like "politics" and "empathy" are even supposed to _mean_—an excellence that _doesn't fit past the regex filter_.
+I wish there was some way I could get a chance to explain to all _my people_ still under the Egregore's control, what _should_ be common knowledge too obvious to mention—that Bayesian _surprise_ is not moral _disapproval_. Beckett Mariner _deserves to exist_. (And, incidentally, I deserve the chance to _be_ her.) But I think the way you realistically _get_ starships and full-body StyleGAN—and survive—is going to require an uncompromising focus on the kind of technical excellence that can explain in mathematical detail what black-box abstractions like "politics" and "empathy" are even supposed to _mean_—an excellence that _doesn't fit past the regex filter_.

 
 But I don't expect to live to get the chance.
 
 But I don't expect to live to get the chance.