Sarah Constantin Profile picture
Writes @ https://t.co/R5P3YYtUwT Married to @oscredwin
4 subscribers
Aug 15 5 tweets 1 min read
@strangestloop there is no Female Experience (TM) and lots of women feel like they're not having the "real" one

i guess read female authors

Anthropology of an American Girl is my favorite novel that most people haven't heard of @strangestloop Jane Eyre for a perspective on romance and coming-of-age

Joy Luck Club for mother/daughter issues

the Vorkosigan series for what motherhood is like

The Female Man is weird and TERFy but has some very sharp insights
Jul 26 17 tweets 2 min read
there's this concept of "easy mode" I've been rolling around in my head lately.🧵 easy mode is:

you can find a small group of people (friends, family, fans) who will applaud your {blog post, art, speech} even if it's mediocre
Jun 17 86 tweets 20 min read
Liveblog/commentary of that Situational Awareness Leopold Aschenbrenner piece: here goes!

situational-awareness.ai

I'm going to quibble with this graph -- the "preschooler"/"smart high schooler" stuff doesn't really describe what's going on at all. situational-awareness.ai/from-gpt-4-to-…
Image
May 12 5 tweets 1 min read
Politics is now about Expressing Who You Are.

It didn't use to be; voting was more a matter of where you lived, what your family background was, etc.

Now it really is part of a coherent personal aesthetic. (And I don't mean "other, dumber people's politics", I mean mine and probably most of yours!)
May 8 31 tweets 6 min read
Had a "viewquake" for the first time in a while

Previously my paradigm was ~informed by history

"think about what usually happens across millennia, everything is temporary, almost no times and places are like our own, expect regression to the mean" with a corollary that markets and technology are basically fragile & culturally contingent

and 90s-era transhumanist predictions were naively optimistic

and just in general buying into "My Tribe are creative and good but not Worldly"
Mar 22 25 tweets 4 min read
So here’s the deal on “artificial wombs”
🧵 The stuff you’ve seen in the news is basically plumbing.

It’s for circulating and oxygenating the blood of extreme preemies, and it’s going to be tried in humans any day now.

link.springer.com/article/10.100…
Jan 29 7 tweets 2 min read
Off topic but I find that I’m very drawn to the premodern “just have rich people fund things they like” model, partly for the embarrassing reason that I *understand* patron-client relationships much more intuitively than more complex & formal institutional dynamics It’s not that I necessarily think Medici-style patronage is best for science! Clearly the people who have thought about it more than I think it isn’t!
Jan 29 6 tweets 1 min read
@kanzure There were supplement shills in extropian circles since the 90’s and they might have set the tone?

Also “normal” people are very miscalibrated about the fact that developing drugs is hard. Not just in aging/longevity contexts. @kanzure You see this in the recurrent claim that “pharma just takes discoveries from academia and hoards the profit”…as though industrial R&D and clinical trials are trivial???
Jan 3 19 tweets 3 min read
This is a good rhetorical counterpoint piece to the common sentiment that the comfort and safety of the modern world has made us shallow. It’s a little more complicated, though.

As long as “you couldn’t withstand adversity” is an accusation that stings because it might be true, the Adversity Fandom will remain compelling.
Dec 23, 2023 14 tweets 5 min read
Favorites 🧵:

reply with a category & I’ll tell you my favorite thing in that category Favorite Christmas carol:

Hark! The Herald Angels Sing. (The tune is Mendelssohn!)
Dec 21, 2023 9 tweets 2 min read
Apparently “income tax withholding makes people not mind high taxes” is a much, much bigger deal than I realized? To the point that:

US income tax was virtually impossible to collect until employers were deputized to collect it in the form of withholding

Republicans think making tax filing easier is a stealth tax increase

The IRS thinks the same thing with opposite valence
Dec 7, 2023 9 tweets 2 min read
Oh I love this.

So, context:

Biology has *dynamics*.

Genes, neurons, and chemical signals like hormones, “upregulate” and “downregulate” each other. You gotta look at them as differential equations (or state machines) not static “levels”. Uri Alon is one of the masters of this approach. (And wrote one of my favorite textbooks of all time. )

This paper applies it to hormones and depression.amazon.com/Introduction-S…
Nov 10, 2023 11 tweets 2 min read
🧵on "niceness."

To be nice is to be, roughly, a person who means no harm, a nonaggressive person. To be nice is not to be good.

You can be nice and cowardly. You can be nice and never do anything nontrivial to help others.

You can be nice but really incompetent, to the point that you do more harm than good.

Still, being nice is, all else equal, a positive thing.
Aug 27, 2023 27 tweets 4 min read


The psychology studies that got debunked in the replication crisis weren’t random; they were part of a theory or worldview that’s still pervasive, and we need to consider what it means for that worldview to be false.carcinisation.com/2023/08/22/aga… The blog post calls the worldview “automaticity”.

Automaticity says that we are “automatically” primed, nudged, subliminally or hypnotically suggested, by small cues in our environment, and that these subconscious cues are more powerful than conscious reasoning.
Mar 29, 2023 7 tweets 2 min read
Everything in this rant is 100% true.

The way math is usually taught (in the US at least) teaches fear, obedience, and faking comprehension.

Real math is the exact opposite. Yep.
Mar 28, 2023 4 tweets 1 min read
marginalrevolution.com/marginalrevolu…

I have mixed feelings on Tyler Cowen's AI piece.

"Progress Good, Stasis Bad"? Yep, agree.

"The Future Is Radically Uncertain"? Yep, also agree.

"Arguing about the details of AI is a mistake"? WTF no stop red flag bad no no no. There's a whiff of "tl;dr" about the rhetoric here:

"Especially when you hear a nine-part argument based upon eight new conceptual categories that were first discussed on LessWrong eleven years ago."

that sounds like it's the...length? or the forum? that bothers him
Mar 28, 2023 4 tweets 1 min read
What I took away from this poll series:

People overwhelmingly think they prefer entering bad-SOTA fields, but significantly fewer are ok with the usual problems of bad-SOTA fields (worse data, worse tools, less prior work, less appreciation of the need for improvement). I’m pretty strongly on the “bad-SOTA is more fun/interesting/tractable” side, both at work and at play.

Eg I like ancient over modern history, small literatures over big literatures, don’t mind hand-generating small datasets, think observational sciences are extra cool, etc
Mar 27, 2023 5 tweets 3 min read
Generated in Midjourney 5:

♈️♉️♊️♋️ ♌️♍️♎️♏️
Feb 15, 2023 6 tweets 2 min read
Nostalgebraist has some of the best writing out there thinking critically/technically about LLM potential.

eg this 2021 post arguing that LLMs won't improve as fast as boosters believe *because they already had most core capabilities at GPT-2's scale*.
lesswrong.com/posts/pv7Qpu8W… OTOH I sort-of disagree with this post arguing that GPT-4 "won't be very useful and won't see much practical use" because OpenAI hasn't done much "productizing" of LLMs to date. nostalgebraist.tumblr.com/post/705192637…
Feb 15, 2023 6 tweets 2 min read
@Ben_Reinhardt is awesome and his new organization looks very cool.

They’re working on two natural “paths to real nanotech”: molecular additive manufacturing (with covalent bonds) and nanomodular electronics (an approach to custom microelectronics without silicon etching). I’m thrilled that he’s chosen materials research for his “private DARPA-style R&D” organizational model.
Feb 13, 2023 13 tweets 4 min read
Yeah, this is fairly legit.

AAV36 is associated with obesity. mdpi.com/1999-4915/7/7/… A bunch of studies have found that obese and overweight humans are more likely to be seropositive for AAV36 -- but the studies are smallish (almost all <1000 subjects) and the seroprevalence rates are all over the place.