this week i

published health coverage by race and ethnicity, 2010-2022 and key facts about the uninsured population.  cinnamon looks like this rn


dined on cod curry.  i'm done with other beers, they've had their chance.  jeopardy round 4/20/23 ken: "and the final clue.."  contestant #3: "bring it"


wonder would dragon sneeze snot or fire? wmata ceilings look like tmnt breastplates. no no no here's all you have to do. type, "hey [colleague], {comical typing motion} describe problem describe problem describe problem" and hit send // if i just put a photo of you, i don't have to describe it


listened and read..

(1) hallelujah, i love her so


(2) christine lagarde also


(3) the national experimental wellbeing statistics

we address biases from (1) unit nonresponse through improved weights, (2) missing income information in both survey and administrative data through improved imputation, and (3) misreporting by combining or replacing survey responses with administrative information.  reducing survey error substantially affects key measures of wellbeing: we estimate median household income is 6.3 percent higher than in survey estimates, and poverty is 1.1 percentage points lower.  these changes are driven by subpopulations for which survey error is particularly relevant.  for householders aged 65 and over, median household income is 27.3 percent higher than in survey estimates.  for people 65 and over, poverty is 3.3 percentage points lower than the survey estimate.  we do not find a significant impact on median household income for householders under 65 or on child poverty

transparency about our methods, code, and estimates is required for an outsider to understand the implications of those kind of detailed data choices 


(4) irony punctuation

the modern question mark (? u+003f) is descended from the "punctus interrogativus" (described as "a lightning flash, striking from right to left"), but unlike the modern question mark, the punctus interrogativus may be contrasted with the punctus percontativus - the former marking questions that require an answer while the latter marks rhetorical questions

ironic statements..printed in italics that lean the other way from conventional italics, also called sartalics

scare quotes..placed around a word or phrase to indicate that it is not used in the fashion that the writer would personally use it  


(5) einstein's why socialism

the payment of the worker is not determined by the value of his product

how is it possible, in view of the far-reaching centralization of political and economic power, to prevent bureaucracy from becoming all-powerful and overweening?



(6) david foster wallace's laughing with kafka

it's not for nothing that kafka spoke of literature as "a hatchet with which we chop at the frozen seas inside us"

the really central kafka joke - that the horrific struggle to establish a human self results in a self whose humanity is inseparable from that horrific struggle.  that our endless and impossible journey toward home is in fact our can tell them that maybe it's good they don't "get" kafka.  you can ask them to imagine his art as a kind of door.  to envision us readers coming up and pounding on this door, pounding and pounding, not just wanting admission but needing it, we don't know what it is but we can feel it, this total desperation to enter, pounding and pushing and kicking, etc.  that, finally, the door opens...and it opens outward: we've been inside what we wanted all along.  das ist komisch

(7) cormac mccarthy's the kekulé problem

 trying to arrive at the configuration of the benzene molecule and not making much progress when he fell asleep in front of the fire and had his famous dream of a snake coiled in a hoop with its tail in its mouth - the ouroboros of mythology - and woke up exclaiming to himself: "it's a ring.  the molecule is in the form of a ring."  well.  the problem of course - not kekulé's but ours - is that since the unconscious understands language perfectly well or it would not understand the problem in the first place, why doesnt it simply answer kekulé's question with something like: "kekulé, it's a bloody ring."  to which our scientist might respond: "okay.  got it.  thanks."

 the unconscious is a machine for operating an animal

breathing, for instance, is not controlled by the unconscious but by the pons and the medulla oblongata, two systems located in the brainstem.  except of course in the case of cetaceans, who have to breathe when they come up for air..the first dolphin anesthetized on an operating table simply died

i've pointed out to some of my mathematical friends that the unconscious appears to be better at math than they are.  my friend george zweig calls this the night shift

the unconscious is a biological operative and language is not.  or not yet.  you have to be careful about inviting descartes to the table

the unconscious is just not used to giving verbal instructions and is not happy doing so.  habits of two million years duration are hard to break


(8) sec chair gary gensler's concerns about artificial intelligence crashing markets

financial crises often arise in one sector, region, or market.  history is replete with shocks emanating from one corner of finance in which the pulling of one thread undoes the financial knitting across an economy.  consider the 2008 financial crisis, the epicenter of which was the u.s. subprime mortgage market

early stage technologies often outpace the development of requisite monitoring capabilities leading to periods of regulation gaps

history and economics have shown that following early phases of competitive diversity, finance often recedes to a more technological uniformity with concentrated actors and interconnected systems

mark carney highlights the likelihood of increased procyclicality in the financial sector due to uniformity.  larry wall and the world economic forum raise the risk that economies of scale in data aggregation will concentrate data sources, perhaps also leading to additional herding behavior

 reinforcement learning helps solve problems that can be modeled as games with rules and incentive structures.  the human-beating chess machines and google deepmind's alpha-go rely on reinforcement learning models.  reinforcement learning may have particular usefulness for capital market trading and investing.  after all, there may be no bigger multi-party competition in the world than the global stock and debt markets

a simple multiple linear model with three input variables will have four parameters - one for each explanatory variable and one for the y-intercept constant.  the neural network shown in figure 1 despite also having three input variables has 37 parameters.  gpt-3, the natural language generator engine of openai, has 175 billion parameters.  the cutting edge of finance neural networks are not as deep as gpt-3, but can still have thousands to millions of parameters.  more modeling parameters increases the likelihood that a model overfits, especially when there are orders of magnitude more parameters than input data points to train on

like forrest gump reaching into his box of chocolates, each time a neural network is trained, the developer does not know what they are going to get

this lack of traditional explainability poses diverse challenges at various levels within organizations and regulatory bodies

the financial system as a complex, adaptive network with similarities to both tropical rainforests and populations during the spread of disease

obscuring risks

multiple sectors deploying deep learning have already seen a coalescence around large, critical datasets.  imagenet is a dominant dataset for academic research in the field of vision recognition research.  google maps, google earth and their affiliate waze dominate the route optimization business and related traffic datasets.  in the field of autonomous vehicles there are the waymo and level 5 datasets, among others.  breast cancer researchers often use the breast cancer wisconsin (diagnostic) data set.  nlp models for text processing and generation commonly use common crawl, a dataset with snapshots of all websites from the last 20 years - effectively representing all of the text on the internet

the inherent characteristics of deep learning, including the challenges of explainability, complexity, and robustness would leave significant gaps for regulators using insufficient tools - akin to using foggy wave glasses - to supervise these models' behavior

existing financial sector regulatory regimes - built in an era of data analytics technology - are likely to fall short in addressing the risks posed by deep learning 


(9) weights in statistics

there are roughly three and a half distinct uses of the term weights in statistical methodology, and it's a problem for software documentation and software development

what's different is variance estimation

only ratios of precision weights matter

with frequency weights..having ten copies of each observation is different from having one copy of each observation

with sampling weights, the randomness is in the sampling..the mean is more precisely estimated when the weights are all similar and less precisely estimated when the weights vary a lot 

(10) retrieval augmented generation

give users insight into llms' generative process

"think of the model as an overager junior employee that blurts out an answer before checking the facts," said lastras.  "experience teaches us to stop and say when we don't know something.  but llms need to be explicitly trained to recognize questions they can't answer"

(11) where did other human species go

they also had chins; something that no other homo species has had (although we don't know why only h sapiens has the protuberance)

at their most abundant there were about 52,000 neanderthals in eurasia before they began to decline

(12) mountain girl

she quickly joined the inner circle of pranksters and became romantically involved with kesey, having a daughter by him named sunshine..she took up with jerry garcia, but they did not marry for many years after.  while living together over the next decade, they raised sunshine along with their two daughters


(13) approximating the size of your social network

there are more than three million americans named michael, around 1% of the population, according to the social security administration, so michaels should also make up 1% of your acquaintances.  if you know eight michaels, you probably know about 800 people

(14) crime statistics

this rate is back to its prepandemic level, which itself came near the end of a multidecade decline.  from 1991 until 2014 violent crime in the u.s., like much of the world, fell sharply, from 783 incidents per 100,000 people to 362.."one of the least remarked upon but most important social phenonema of our lifetimes"

the september national crime victimization survey showed only about 40% of violent crimes were reported to the police in 2022


(15) van gogh's publicist

the key player in the growth of vincent's posthumous fame

the amsterdam social-democratic women's propaganda club

wikipedia for the stage

(16) the latest pisa

globally, students lost the equivalent of three-quarters of a year of learning in math, which was the primary focus of the 2022 test

(17) the excel olympics

the lebron james of excel

the dreaded #ref error

to the uninitiated, the event was a little like watching people sweat while taking the sats 


(18) the economic census of william the conqueror

the king sent his agents to survey every shire in england, to list his holdings and dues owed to him

under the feudal system, the king was the only true "owner" of land in england, by virtue of his allodial title

tenant (from the latin verb
tenere, "to hold")

in total, 268,984 people are tallied in the domesday book, each of whom was the head of a household

the crown was entitled to archaic dues in kind, such as honey

a descriptio (enrolling)

from it, as from the last judgement, there is no further appeal

(19) the prussian officer by d. h. lawrence

and from that time an undiscovered feeling had held between the two men

what cattle were you bred by, that you can't keep straight eyes?

once he flung a heavy military glove into the young soldier's face.  then he had the satisfaction of seeing the black eyes flare up into his own, like a blaze when straw is thrown on a fire.  and he had laughed with a little tremor and a sneer

his blood seemed to be corroding

"yes, sir," he replied, between shut teeth

the bodies of the two men lay together, side by side, in the mortuary

(20) r studio community feels about code written by artificial intelligence same way ronald reagan felt about negotiating with the soviet union