bkfunk

> I get people are sad if they get forced out of their homes for this

Wow, what a way to dismiss the humanity and worth of millions of people.

First of all, in some cases it actually IS a surprise: historical flood data was often quite limited and cities and communities formed before there were accurate estimates of flooding probability, let alone enough understanding of climate change to predict 21st century flooding probability.

If you’re thinking “this side of the street floods, but not the other side, just move across the street”, in many places that is not at all the situation. Miami (avg elevation: 5.9 ft) and New Orleans (elevation from -6.6 to 19.7 ft) are large, dense (by American standards) cities that have existed for over a hundred years. Giant swathes of them are in flood plains. You cannot move a third or more of a city’s residents in any reasonable time—there aren’t enough housing units, never mind the schools and utility capacity and businesses and everything else needed to support a community of people. Plus, most Americans, if they own a home, the vast majority of their net worth is in their house. You pull the insurance, that value vaporizes. Sure, from the perspective of the overall economy, that may seem more efficient. But individuals can use economic efficiency for a down payment. Congratulations, you just vanished billions of dollars of assets! Hope that doesn’t affect the banking system or the local economy!

And they can’t move across the street—to get out of the flood plain they may have to move a significant distance, and people already live in the higher areas of the city: effectively (especially for low and moderate income families) they have to move > 50 miles away. That means they lose their social network, support system, and likely their job. Moving a long distance is expensive, extremely stressful, and can break the few remaining social bonds that we now know are a huge factor in everything from recovering from bankruptcy to whether someone is too disabled to work, to how long people live.

> I feel zero outrage here.

I’m not sure anyone is trying to get you to feel outrage? Like it’s a hard policy problem that greatly affects millions of people. Maybe you could just…care?

If moving millions of people—functionally including entire metro areas like…oh I don’t know…Miami and New Orleans and large swathes of NYC—by forcing them to abandon their homes (with no insurance they are now unsalable) and then buy/rent (using what money?) new homes far away from their communities and jobs (if they bought close, they’d still be in the flood plain!)—if that’s “simple”, I don’t want to know what policy implementations you’d call “hard”!

Once I started learning more about biology, I realized that everything is just so complex. The body repurposes chemicals a lot, so you have things like serotonin being a key neurotransmitter in the brain, but also in the gut. And you have enzymes that are coded in genes, but then there are also networks of genes that are up- or down- regulated by hundreds of other genes, and sometimes only in certain types of cells or certain physiological environments. And then of course there are epigenetic and immune-modulated effects at the genome, gene network, and individual gene levels. Not to even mention all the feedback mechanisms and meta-feedback mechanisms (the drive toward homeostasis is POWERFUL), and effects of countless chemicals in our environment.

There are certainly clear-cut cause-effect relationships in biological systems, but even they will have edge cases and random chance to muddle the picture.

I would posit that the human body is far more complex than even the largest codebase, not least because it was jury-rigged together with no architect or style guide.

Also, in general, the more common the exposure, the harder it is to find a link; try finding a control group of people who have never been exposed to PTFEs, or HSV, and who also aren’t like hunter gatherers.

> off-patent

Well there’s your problem: no one can make money off of it, unless they develop a new delivery mechanism, etc.

Patents encourage developing new medicines, but not developing new knowledge about (never mind use of) old medicine.

The solution (in the US) is obvious: federal funding of research that stands to help lots of people but not make lots of money. Since most of these patients (in the US) are going to be on Medicare, there could be huge potential cost savings to the taxpayer: memory care is EXPENSIVE, so even the paltry amount covered by Medicare racks up (and the opportunity costs of people paying for private memory care is enormous).

But instead of increasing funding for this kind of life- AND MONEY-saving research, this administration is freezing and slashing research funding, and specifically targeting Columbia for political/Trump’s-petty-grudge reasons.

If you have termites, you don’t just light the house on fire.

So many tech people try to solve all the problems of Gov tech in the executive branch, which is intentionally slow and conservative. And yet, watch any Congressional hearing about a tech topic, and it’s painfully obvious that Congress has very little expertise in tech issues on staff.

Instead of going 12 rounds with OIRA about the PRA (which I hate as much as the author does), what if we…changed it?

The Judiciary also has no idea how to think about tech issues.

Don’t blame the executive branch for the perverse constraints and incentives created by the Legislative and Judicial.

“That today Microsoft is a giant company is irrelevant...”

I am not too young to remember the old Microsoft. To say that Microsoft is “irrelevant” is so myopic. Despite Tesla, GM is still relevant. Despite AWS, DB2 mainframes are still relevant. Heck, I have to work with EBCDIC data, a format designed to not produce holes in punchcards that are too close together. Even when we eventually move to a modern db, decades of archival data is not going to be converted from EBCDIC.

Windows might be irrelevant to FAANG or MANGA or GAMMA or whatever, but how many Fortune 500 companies don’t have a significant Microsoft presence?

Apple computers are pretty nice, but they’re expensive, and the vast majority of employees do fine with a cheap PC and Microsoft 365—why would a company pay more for unnecessary hardware that also requires rebuilding a bunch of IT systems, not to mention retraining thousands of employees.

The number of EOs is a poor measure of the extent of executive power. There are many other ways to concentrate executive power; for example OMB (part of the Executive Office of the President) issues “M-memos” (“M” for “management”) to the heads of all departments, instructing them on how to implement EOs, laws, and White House priorities.

There is a consensus among researchers of the workings of the US government and the legal context thereof that, since FDR and especially since Reagan, that the “imperial presidency” has been gaining ground.

Unless you want to “write a Shakespearean sonnet about Bayes’ rule”. Here’s what Claude sonnet (ahem) came up with:

"Bayes's Revelation"

  What knowledge held before new facts arrive,
  Our prior thoughts, though incomplete they be,
  Must yield and change as evidence does strive
  To shape posterior probability.

  Through multiplicative steps we find our way,
  As likelihood with prior beliefs combine,
  The normalizing constant holds its sway,
  While truth emerges, clearer line by line.

  O Reverend Bayes, thy theorem divine
  Shows how our minds should learn from what we see,
  Each observation helps our thoughts refine,
  Till posterior wisdom sets uncertainty free.

  In statistics' realm, this rule holds fast:
  Each present builds upon what's known from past.

  The sonnet weaves together the key components of Bayes' theorem - prior probability, likelihood, and how they combine to form the posterior probability. Would you like me to explain any of these statistical concepts in more detail?                

Although it would be an easier change now that Bard is Gemini…

This is a study of people in their 70s. The vast majority of people with hearing loss in their 70s lost it late in life; they have no Deaf/HH community, they almost never learn to sign, and they often struggle to adjust for their loss of hearing.

The study you linked talks about reduced stimulation, and in particular _social_ stimulation:

> when an individual suffers from moderate to severe hearing loss, they are less likely to participate in social activities. Perhaps they are embarrassed about their hearing loss. Or they may simply find it unrewarding to attend a social event when they cannot hear what is going on.

People who are born deaf/hh , or who lose their hearing early in life, if they are allowed to access and participate Deaf/HH communities and spaces, simply do not have any of these difficulties in social contexts within those communities.

Martha’s Vineyard had an unusually high rate of congenital deafness for centuries [1]. It became a place where everybody, deaf and hearing alike, used sign language regularly. In such a society, being deaf was not a significant impediment to participating in social society at all; I am aware of no evidence that would suggest the dementia rates would be higher for the deaf residents just because of their deafness.

A disability is only a disability in a given context; for some conditions (eg advanced ALS), they are disabling in almost all contexts, while for others (eg a food allergy), they are disabling in a relatively narrow set of contexts. The relationship to dementia is caused by the hearing loss mis-fitting the individual’s context; people with the same condition but different contexts would not be deprived of stimulation and therefore not susceptible to dementia in the same way.

[1]: https://en.wikipedia.org/wiki/Martha's_Vineyard?wprov=sfti1#... (Martha’s Vineyard sign language is a major source for what became American Sign Language. The other was French Sign Language, which is why British Sign Language and ASL are quite different despite sharing the same local spoken language)

Or you could read the article and use common sense?

> At first, AirWayBill’s managing director, Khaled Sehly, declined to discuss whether the app could be breaking any TSA regulations, asking to go off the record. (I refused.) He then asked me to wait four weeks until they had real users, saying that the reviews left on the app were actually from peers rather than customers. In our next series of broken calls, Sehly directed me towards the app’s terms and conditions and privacy policy, which he says were designed by a reputable law firm. Section 8 of the terms and conditions states that users must comply with technical and legal obligations and restrictions, including customs rules. However, it also indemnifies AirWayBill for liability related to shipments, stating that “any request will be made or accepted at the Members’ own risk” and that unless it’s explicitly specified otherwise within the platform, “AirWayBill’s responsibilities are limited to the correct functioning of the app and its service to the interested parties.”

> Unfortunately for wary couriers, AirWayBill is not planning on running background checks on anybody using their services. Sehly says that delivering packages for strangers—or at least friends of friends—is something that he’s seen done in an unregulated way in the Middle East and parts of Europe, with people asking if someone could deliver items ranging from documents to baby food to their families. Still, the possibility of inadvertently smuggling who knows what still remains. “We always urge people not to carry anything that they are not super confident about,” Sehly says, pointing out that deliverers can inspect the items they plan on transporting.

So the company is obviously taking the caveat emptor aka “f around find out” approach. Drug smugglers already have a lot of skill hiding drugs in seemingly innocuous items (sewn into teddy bears, etc.), so it is not at all an “aggressive assumption” to think being a courier with this service has potentially life changing (i.e. incarceration) risks.

But this wouldn’t be in orbit; it would be in what NASA calls “deep space”, which relies on the Deep Space Network [1]. The DSN is severely bandwidth constrained, due primarily to a lack of ground antennas. Indeed, for instruments that are located outside Earth’s orbit (e.g. SOHO, which is at Sun-Earth L1 [2]), bandwidth is often a limiting constraint in the design.

My understanding is that some newer instruments do both compress and select data to be downloaded (i.e. prioritizing signal over noise), and that there is more and more consideration of on-board processing for future missions, as well as possibly introducing the capability within DSN itself to prioritize which instruments get bandwidth based on scientific value of their data.

Source: A presentation from people at NASA Heliophysics last week, where this very topic came up.

[1]: < https://www.nasa.gov/communicating-with-missions/dsn/> [2]: < https://science.nasa.gov/mission/soho/>

It’s very rare in the West, but speakers of tonal languages seem to be much more likely to be able to do it https://www.sciencedaily.com/releases/2009/05/090519172202.h...

Suppose:

1. An AI model, which works simply by predicting the best next word given a particular context (context which obviously includes whatever prompts it was given), becomes sufficiently sophisticated that, given some particular context, it can produce the text (so, normal language but also, say, shell commands and, like, python) needed to hack into, let’s say, a natural gas pipeline company’s systems and fiddle with the pressurization such that it causes a catastrophic explosion that at the very least stops gas from going to millions of homes in the dead of winter. 2. The AI system’s representation of “context” is sufficiently sophisticated enough that it can, when given some simple linguistic prompts, can represent (even if it does not “understand”) the intent or latent meanings of the human providing the prompts (or rather, the multiple humans, including those who built it and primed it with some initial set of rules and goals, as well as the perhaps unwitting or perhaps nefarious end user). 3. The AI system’s produced text can evoke additional context, either in a conversation where the AI can lead a human down a path of providing additional prompts, or where the AI system can interact with systems that might give responses to the original text (including erroneous responses). 4. The AI is trained on the corpus of human culture (and not just the parts we like!), including both fictional and nonfictional accounts of and tutorials on hacking, war, terrorism, revolutionary mobilization, heroic tales of overthrowing oppressive regimes, etc. (That is, for every conflict, every actor’s glorification and vilification is included in the corpus, and there is no way to ensure that the “correct” versions, that is the versions that we as people steeped in Western ethics and systems of morality, are labeled as such or weighted more 5. The AI system’s text can be ported to a shell, servers on the internet, or even just to humans who have access to such methods of turning text into real world interactions.

What might happen? Is it conceivable that a bad actor could use the AI to do malicious work in the real world that the actor did not have the skills to execute themselves? Is it conceivable that a jokester could play around, but that the AI could misunderstand the intent and/or psychologically influence them into providing prompts that lead to disastrous outcomes? Could someone asking it to fix climate change cause it, correctly IMO, to develop a contextual representation that we have to move away from fossil fuels with great urgency even if it causes some short term economic harm, and then, incorrectly IMO, develop a contextual representation that economic ecoterrorism is the best way to do that?

In other words: If it can, through benign intent, malicious intent, or simply through misunderstanding or error, cause great harm, who cares it it is just an algorithm, with no sentience or “beingness”? If it’s below freezing and you don’t have heat for weeks, who cares if the AI can be properly said to be “thinking” or is just responding to stimuli?

Note that TFA says that the man in question lived in England at some point in his 20s, so not that “far away”, in terms of amount of exposure to various Irish accents. Also, it doesn’t say whether it was an accurate or consistent accent. He could have just been doing the accent of the Lucky Charms guy.

“Biological” does not mean “genetic”. I believe parent is musing whether there is a neurodevelopmental basis, which of course there is, by all evidence; if you do not hear certain sounds by a young age, you will never be able to speak a language using those sounds with an accent indistinguishable from that of a native speaker. Obviously there is, indeed, a neurological component to that phenomenon.

In the US, the “Irish accent” (scare quotes because there is of course no single accent of Ireland) is fairly well known (that is, watching American TV and movies, you encounter Irish native English speakers more than, say, South African English speakers). And in TFA it says the patient in question lives in England, where of course there is even more exposure to Irish accents.

There must be, I posit, some circuitry that can interpret the phonemes of differently accented English and translate them into whatever latent representation of words or concepts is used for thought and speech production; so it’s interesting to wonder whether what is happening is the wires are getting crossed, as it were, and receptive circuits start connecting to productive circuits that control the speech organs to produce different sounds. What’s more, TFA doesn’t say if it was a good or accurate or even geographically consistent and coherent accent.

I believe parent is saying that PTSD is precisely a situation where memories act like perceptions; EMDR as a therapeutic modality with at least some evidence for efficacy is premised on the idea that it helps to make traumatic memories feel like memories rather than perceptions

The issue is, can myapp know that it was set that way, vs being set somewhere else? The command line argument uses syntactical constraints to ensure that the value of the argument was set specifically with the intention of using it in myapp.

If Random House declines to publish your book because they don’t think it will sell, is that censorship?

If I run a sci-fi bookstore, and I choose not to stock your book about political philosophy, is that censorship?

If it write an article that reviews your book (wherein, necessarily, I pick and choose what parts of your book I talk about, and also paraphrase [is that the same as “manipulating information”]), is that censorship?

When there is simply too much information for any person to consume, and even too much to be able to _evaluate whether To consume_, what does _not having censorship_ look like?

Don’t know why you’re getting downvoted, this is literally a theory discussed in TFA:

> "This kind of quirky idea was that their arms were small enough that they didn't get in the way of all these guys fighting over food with giant mouths, so they weren't basically biting their own arms," says Krumenacker.

Not cheap, but check out https://www.levelshealth.com/

People (understandably) often lament how hard it is to make an impact in the world. Being a juror can have far more impact than every vote you’ll ever cast.

They do this. It’s called Airbnb Plus [0].

[0]: https://www.airbnb.com/plus

On Mythbusters, every now and then they would say something like "We would like to do X [e.g. something crazy like jump out of a plane with a duct tape parachute], but our insurance company wouldn't let us."

I've always how that worked. Like, who is the employee at the insurance company who takes that call, and how do they decide what's just too risky, when there can't be much historical data on risk?

I generally agree about the pros you list of Bayesian, but I think in companies that are doing a lot of experiments and are not just optimizing a checkout page, they don't hold as well. For example, you often are testing out features that really are different from any feature before, and so a prior is harder to get alignment on. There are also frequentist methods like CUPED to mitigate the effect of confounding variables, and then usually you're analyzing the results across user segments so that you can try to look for heterogeneous treatment effects. (FWIW, having analyzed a lot of experiments, I've been surprised at how often there is NOT a heterogeneous treatment effect across user segments; the baselines are often very different, but the direction and general magnitude of the lifts are typically similar. And, the confounder that is most often relevant is simply some broad "engagement" dimension: users who use the product a lot might behave differently from users who do not.)

In my experience, the biggest blocker to using Bayesian approaches has been about data and computation requirements; there are some closed-form solutions (e.g. https://www.evanmiller.org/bayesian-ab-testing.html , also https://www.chrisstucchio.com/blog/2014/bayesian_asymptotics... ) but even those are computationally difficult at scale even if you don't have to do a lot of MCMC sampling.

But what is really interesting to me about your comment is:

> A/B Tests are not really controlled experiments, at least not in the same way clinical trials are. User behavior is always observational even if you have a control group

Could you expand on that? Isn't clinical data also observational: you observe the patient after some time, and see what their symptoms are or what various endpoint measurements are?

Power is important if you want to distinguish between noise and true signals. In other words, if you care about P(real effect | significant).

From good ol' Bayes we get:

P(real|sig) = P(sig|real) x P(real) / P(sig)

P(sig|real) is the power; so if you have more power, all other things being equal (a bit of a weaselly caveat), the likelihood that your stat sig result is real is higher.

Mentioned earlier in the series: https://en.wikipedia.org/wiki/Ship_of_Theseus

More like a median of 47k, according to BLS, with 90% earning less than 70k: https://www.bls.gov/oes/current/oes533032.htm

The Hippocratic oath is not required in the US, and many schools that do use it have significantly modified it. Eg:

> In a 2000 survey of US medical schools, all of the then extant medical schools administered some type of profession oath. Among schools of modern medicine, sixty-two of 122 used the Hippocratic Oath, or a modified version of it. The other sixty schools used the original or modified Declaration of Geneva, Oath of Maimonides, or an oath authored by students and or faculty. All nineteen osteopathic schools used the Osteopathic Oath.

https://en.m.wikipedia.org/wiki/Hippocratic_Oath

Really interesting company! From the article:

> Facit Homes is a relatively small builder - they build fewer than 10 homes a year, all of them custom. And though they’re using advanced manufacturing techniques, they don’t seem especially interested in any sort of mass production.

I’m wondering if this is true? Or do you think starting high end and super custom might lead to bigger scale?

This looks interesting! Can I ask, what's the origin of the name? Both "siu" and "siuba"? Would be a useful addition to the docs, IMO.