korin43

Comments

How much to worry about the US election unrest?

I doubt riots will get any worse after the election, and they're a local problem, not a country problem. If there's violence where you are, I expect it to be about the same afterward, but most people in the country aren't interested in violence and that won't change just because the current president stays in office for a few more years.

It doesn't matter if Trump and Pence respect the election result. If Trump loses the election, he's not the president anymore and the federal bureaucracy and military will stop listening to him. For a supposed-fascist, he's terrible at stocking the government with supporters, and the idea that the US military would support him in a coup is unbelievable.

It also doesn't matter if he wants to interfere with the election, because he can't. Presidential elections are run by state governments (and there are some strong constitutional barriers that would prevent anyone in the federal government from interfering with them).

Why isn't JS a popular language for deep learning?

if you want to continue to train in users' browsers

Is this a thing that can be done? I would have expected it to be way too slow.

In my experience (in the US), the nicer a house gets, the more you save buy owning the house vs renting it. For example, renting a house in my neighborhood costs $1200 per month more than my mortgage. It's obviously not an optimal financial situation, since I could sell this house and rent a run-down 2-bedroom apartment for even cheaper, but I don't want to live in a run-down 2-bedroom apartment anymore.

I suspect this is because of the disconnect between what makes rentals risky for landlords (tenants screwing things up) and the benefits of renting for a tenant (if you screw things up, there's limits to how much it's your problem). Once you reach a certain expense / necessary upkeep to remain nice-looking level, the only way to make it work is to either have insanely high rent so the landlord can afford to fix the place up if a rental goes bad, or to make the landlord and tenant the same person so their desire to not-screw-it-up is aligned.

Although for the record, at least in the US, high home ownership rates are a sign of a problem with society. Government planners think everyone should own their house whether they want to or not, so renters have to pay additional taxes which are used to subsidize home owners. The crossover point would presumably be different if our tax system was fair.

Why isn't JS a popular language for deep learning?

One other thing to consider is that the language you do your data science work in doesn't have to be the same one you deploy your code to in production. At one point, the company I work at did data science work in R and Python and then compiled models to PFA which allows us to run the same model in various other languages. I'm not sure how popular this is, but I suspect you can load TensorFlow models into other languages relatively easily.

Efficacy of Vitamin D in helping with COVID

Some of the results are reported in really confusing ways:

A study in Indonesia found that out of the patients that died from COVID-19, 98.9% of them were deficient in vitamin D, while only 4% of the patients with sufficient vitamin D died.

These two numbers don't add up to 100%, which indicates immediately that something sketchy is going on. I think what this is describing is two different measurements:

  • Of patients who died, what percentage had vitamin D deficiency? (98.9% deficient, 1.1% sufficient)
  • Of patients who had deficiency or sufficiency, how many died? (4% of sufficient died, [unknown]% of deficient died)

This doesn't change the results in this case, but it's needlessly confusing to report an apples-and-oranges comparison when you could compare apples to apples.

A study of patients in New Orleans found that 84.6% of the COVID-19 patients in the ICU were deficient in Vitamin D while only 4% of the patients in the ICU had sufficient levels of Vitamin D.

Same thing here except they seem to be reporting the same measurement. What category are the remaining 11% in? Or is this reporting two different categories (COVID-19 patients in ICU vs all patients in ICU)?

Design 3: Intentionality

Standard Notes had some annoyances (sync not working, license issues, too-strict page styles), so I switched to Notion. It also gave me a few more features:

  • File attachments
  • Page hierarchy
  • Web interface if I need it
  • Ability to share page links

Apparently I've been doing this consistently since I posted the original comment. My usual method is to write up a todo list and then add comments as a bulleted list under items and sometimes link to additional pages. At the start of each day, I copy the previous day's journal entry, delete anything that's no longer relevant, and then add anything new that comes up.

The major ways it's been useful are:

  • When people ask me questions about things I don't remember, I can look them up (if it was recent enough)
  • At the start of the day, I can quickly refresh myself on what I did and what I need to do (which makes it easier to get started)
  • It's useful as a reference during standups

I think there are still things I could do better:

  • I don't do enough retrospectives. A daily/weekly/monthly/yearly review and summary would be useful partially as an index to find things, something I could reference for performance reviews or resume material, and as just a feel-good "look at all the stuff I did".
  • Right now this is the majority of my work-related notes and it would probably be good if I had subject-based pages and not just date-based pages.

Maybe I'll try adding in more retrospective/summary stuff and see how that goes.

nostalgebraist: Recursive Goodhart's Law

The author seems to be skipping a step in their argument. I thought Goodhart's Law was about how it's hard to specify a measurable target which exactly matches your true goal, not that true goals don't exist.

For example, if I wanted to donate to a COVID-19 charity, I might pick one with the measurable goal of reducing the official case numbers.. and they could spend all of their money bribing people to not report cases or make testing harder. Or if they're an AI, they could hit this goal perfectly by killing all humans. But just because this goal (and probably all easily measurable goals) are Goodhartable doesn't mean all possible goals are. The thing I actually want is still well defined (I want actual COVID-19 cases to decrease and I want the method to pass a filter defined by my brain), it's just that the real fundamental thing I want is impossible to measure.

What am I missing? (quantum physics)

I feel like this is missing something. How do we know they have both spins before we measure them and that they haven't "decided" their spin beforehand?

Does crime explain the exceptional US incarceration rate?

I think your argument is interesting, but doesn't make sense in the context of this article. The author is using homicide as their proxy for crime, and homicide is the prototypical example of something which is always a crime (although there's some variation in who can get away with it and what the punishment is).

(Also while it's true that homicide was historically not punished by prison very often, your second paragraph dances around the fact that that's because the punishment was worse - death)

Repeat Until Broke

This reminds me of the Mad Fientist's rules, although I think he uses a 90/10 split and rebalances by buying and never selling. https://www.madfientist.com/my-portfolio/

Load More