sarahconstantin's Shortform

sarahconstantin

links, 10/14/2024

https://milton.host.dartmouth.edu/reading_room/pl/book_1/text.shtml [[John Milton]]'s Paradise Lost, annotated online [[poetry]]
https://darioamodei.com/machines-of-loving-grace [[AI]] [[biotech]] [[Dario Amodei]] spends about half of this document talking about AI for bio, and I think it's the most credible "bull case" yet written for AI being radically transformative in the biomedical sphere.
- one caveat is that I think if we're imagining a future with brain mapping, regeneration of macroscopic brain tissue loss, and understanding what brains are doing well enough to know why neurological abnormalities at the cell level produce the psychiatric or cognitive symptoms they do...then we probably can do brain uploading! it's really weird to single out this one piece as pie-in-the-sky science fiction when you're already imagining a lot of similarly ambitious things as achievable.
https://venture.angellist.com/eli-dourado/syndicate [[tech industry]] when [[Eli Dourado]] picks startups, they're at least not boring! i haven't vetted the technical viability of any of these, but he claims to do a lot of that sort of numbers-in-spreadsheets work.
https://forum.effectivealtruism.org/topics/shapley-values [[EA]] [[economics]] how do you assign credit (in a principled fashion) to an outcome that multiple people contributed to? Shapley values! It seems extremely hard to calculate in practice, and subject to contentious judgment calls about the assumptions you make, but maybe it's an improvement over raw handwaving.
https://gwern.net/maze [[Gwern Branwen]] digs up the "Mr. Young" studying maze-running techniques in [[Richard Feynman]]'s "Cargo Cult Science" speech. His name wasn't Young but Quin Fischer Curtis, and he was part of a psychology research program at UMich that published little and had little influence on the outside world, and so was "rebooted" and forgotten. Impressive detective work, though not a story with a very satisfying "moral".
https://en.m.wikipedia.org/wiki/Cary_Elwes [[celebrities]] [[Cary Elwes]] had an ancestor who was [[Charles Dickens]]' inspiration for Ebenezer Scrooge!
https://feministkilljoys.com/2015/06/25/against-students/ [[politics]] an old essay by [[Sara Ahmed]] in defense of trigger warnings in the classroom and in general against the accusations that "students these days" are oversensitive and illiberal.
- She's doing an interesting thing here that I haven't wrapped my head around. She's not making the positive case "students today are NOT oversensitive or illiberal" or "trigger warnings are beneficial," even though she seems to believe both those things. she's more calling into question "why has this complaint become a common talking point? what unstated assumptions does it perpetuate?" I am not sure whether this is a valid approach that's alternate to the forms of argument I'm more used to, or a sign of weakness (a thing she's doing only because she cannot make the positive case for the opposite of what her opponents claim.)
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10080017/ [[cancer]][[medicine]] [[biology]] cancer preventatives are an emerging field
- NSAIDS and omega-3 fatty acids prevent 95% of tumors in a tumor-prone mouse strain?!
- also we're targeting [[STAT3]] now?! that's a thing we're doing.
  - ([[STAT3]] is a major oncogene but it's a transcription factor, it lives in the cytoplasm and the nucleus, this is not easy to target with small molecules like a cell surface protein.)
https://en.m.wikipedia.org/wiki/CLARITY [[biotech]] make a tissue sample transparent so you can make 3D microscopic imaging, with contrast from immunostaining or DNA/RNA labels
https://distill.pub/2020/circuits/frequency-edges/ [[AI]] [[neuroscience]] a type of neuron in vision neural nets, the "high-low frequency detector", has recently also been found to be a thing in literal mouse brain neurons (h/t [[Dario Amodei]]) https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10055119/
https://mosaicmagazine.com/essay/israel-zionism/2024/10/the-failed-concepts-that-brought-israel-to-october-7/ [[politics]][[Israel]][[war]] an informative and sober view on "what went wrong" leading up to Oct 7
- tl;dr: Hamas consistently wants to destroy Israel and commit violence against Israelis, they say so repeatedly, and there was never going to be a long-term possibility of living peacefully side-by-side with them; Netanyahu is a tough talker but kind of a procrastinator who's kicked the can down the road on national security issues for his entire career; catering to settlers is not in the best interests of Israel as a whole (they provoke violence) but they are an unduly powerful voting bloc; Palestinian misery is real but has been institutionalized by the structure of the Gazan state and the UN which prevents any investment into a real local economy; the "peace process" is doomed because Israel keeps offering peace and the Palestinians say no to any peace that isn't the abolition of the State of Israel.
- it's pretty common for reasonable casual observers (eg in America) to see Israel/Palestine as a tragic conflict in which probably both parties are somewhat in the wrong, because that's a reasonable prior on all conflicts. The more you dig into the details, though, the more you realize that "let's live together in peace and make concessions to Palestinians as necessary" has been the mainstream Israeli position since before 1948. It's not a symmetric situation.
[[von Economo neurons]] are spooky [[neuroscience]] https://en.wikipedia.org/wiki/Von_Economo_neuron
- only found in great apes, cetaceans, and humans
- concentrated in the [[anterior cingulate cortex]] and [[insular cortex]] which are closely related to the "sense of self" (i.e. interoception, emotional salience, and the perception that your e.g. hand is "yours" and it was "you" who moved it)
- the first to go in [[frontotemporal dementia]]
- https://www.nature.com/articles/s41467-020-14952-3 we don't know where they project to! they are so big that we haven't tracked them fully!
- https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3953677/
https://www.wired.com/story/lee-holloway-devastating-decline-brilliant-young-coder/ the founder of Cloudflare had [[frontotemporal dementia]] [[neurology]]
[[frontotemporal dementia]] is maybe caused by misfolded proteins being passed around neuron-to-neuron, like prion disease! [[neurology]]
- https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6838634/
- https://www.nature.com/articles/s41467-018-06548-9.pdf inject the bad protein into a mouse and it really does spread!
- https://researchfeatures.com/cell-cell-transmission-proteins-core-neurodegenerative-disease/ something similar might be happening in [[Alzheimer's]] as well
- https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3943211/ the spread of [[ALS]] through the brain is consistent with cell-to-cell transmission of misfolded proteins

Showing 3 of 6 replies (Click to show all)

she's more calling into question "why has this complaint become a common talking point? what unstated assumptions does it perpetuate?" I am not sure whether this is a valid approach that's alternate to the forms of argument I'm more used to, or a sign of weakness

It is good to have one more perspective, and perhaps also good to develop a habit to go meta. So that when someone tells you "X", in addition to asking yourself "is X actually true?" you also consider questions like "why is this person telling me X?", "what could they gain in this situation by maki... (read more)

1MichaelDickens2y

I don't know how familiar you are with regular expressions but you could do this with a two-pass regular expression search and replace: (I used Emacs regex format, your preferred editor might use a different format. notably, in Emacs [ is a literal bracket but ( is a literal parenthesis, for some reason) 1. replace "^(https://.? )([[.?]] )*" with "\1" 2. replace "[[(.*?)]]" with "\1" This first deletes any tags that occur right after a hyperlink at the beginning of a line, then removes the brackets from any remaining tags.

RE Shapley values, I was persuaded by this comment that they're less useful than counterfactual value in at least some practical situations.

sarahconstantin's Shortform

8

sarahconstantin's Shortform

8