skzv

Awesome. To take this a step further, I used this math to turn 2D photos into 3D scenes with depth maps: https://blog.skz.dev/3d-reconstruction-from-public-photos

Cool, I worked on indoor localization, particularly with RTT, at Google for many years :)

From the wiki article:

> Scholars believe that the Novgorod Republic had an unusually high level of literacy for the time, with literacy apparently widespread throughout different classes and among both sexes.

I bought my father a RM-BIV3 ribbon microphone [0] to record violin. It's a beautiful piece.

[0] https://micpedia.com/microphone/bashaudio-rm-biv3/

Yes, you are totally correct, but I believe this term is omitted from the cross-entropy loss function that is used in machine learning? Because it is a constant which does not contribute to the optimization.

Please correct me if I'm wrong.

To bring things full circle: the cross-entropy loss is the KL divergence. So intuitively, when you're minimizing cross-entropy loss, you're trying to minimize the "divergence" between the true distribution and your model distribution.

This intuition really helped me understand CE loss.

Increasing efficiency often causes an increase in consumption, not less [0]

[0] https://en.wikipedia.org/wiki/Jevons_paradox

I was concussed and suffered from persistent headaches for 2 years.

It was really tough. I was suicidal. My only reprieve from pain was falling asleep.

I saw a neurologist and he told me that two most important things for your brain are:

- consistent sleep schedule

- regular exercise

Once I got those two under control, the headaches finally went away.

The Vice documentary mentioned in the article is really great [0].

Best of luck to Agafia.

[0]: https://www.youtube.com/watch?v=tt2AYafET68

I exploited one of these regional naming conventions so that I only have a first name on Facebook.

I go to the office almost everyday by choice. Free food, snacks, and coffee, gym, and medical clinics on campus. And it's just nice to get dressed and leave the house.

But it's really nice to have the flexibility to WFH when I need to, especially just mornings to skip traffic.

Aren't you describing quantum field theory (QFT)?

Anyway, what exactly is a field besides a mathematical object? What is it made of?

Sorry, it's actually 49". It's the Samsung Odyssey G9. I love it.

I've been happily using Magnet for a long time for this purpose. Good window tiling is critical when working on a 57" monitor...

I grew up in Canada, but my family and I are first generation immigrants from Ukraine. The privilege of my peers left a huge, almost traumatizing, impression on me. My parents were constantly working overtime at their minimum wage jobs at a plastics plant to support us, while my brother and I were raised by my grandmother. Most of my peers had wealthy parents and it was hard to relate. The silver lining is that it really motivated me to work hard and escape that. I live in California now and make many times more than my parents ever made in their lives. My mom still works as a janitor for minimum wage.

Most people I know back in Canada get lots of support from their parents. Everyone I know who bought real estate had huge help from their parents. Canada is just too expensive otherwise and local jobs don't support the cost of living. It definitely feels unfair to those that don't have wealthy parents. They fall behind, and as housing prices rise, it becomes harder to catch up. Those definitely won demographic lottery, and I don't think many of them understand how lucky they are.

All that being said, given a global perspective, I am very lucky, too.

I slaughtered about 40 chickens once. I was probably around 19 at the time. My ex-girlfriend's dad just told me to get in the truck, asked his kids to stay at home, and took me to a chicken coop where him and his friend were raising chickens. I took them out of the coop, a few at a time, hung them upside down, and then killed them with a knife. After the first few, the rest in the coop began panicking even though they couldn't see what was going on. That really stood out to me. I'm sure they could hear or smell death. They knew what was coming.

Yes, but it also makes a huge difference whether we are asking the model to interpolate or extrapolate.

Generally speaking, models perform much better on the former task, and have big problems with the latter.

One of my favourite economic paradoxes. It changes the way you think about efficiency and consumption.

My colleague introduced me to this idea. He had been studying ways to increase computing efficiency out of concern for the environment. Making programs more efficient would reduce energy consumption, right?

His advisor introduced him to Jevons paradox and he realized such efforts could have the exact opposite effect. So he dropped that research entirely. If you're worried about energy consumption, you need to make energy production more green, not machines more efficient.

Making data centers more efficient will probably cause us to build more data centers and use more power overall, not less.

Interesting, I don't remember it getting hot.

Last time I went into an MRI, I realized half way through that I had forgotten to take off my metal necklace. Thankfully it was silver - it was an extreme purity test.

Thank you for all your sacrifices, the laboratory mouse! [0]

[0] https://en.wikipedia.org/wiki/Monument_to_the_laboratory_mou...

One of the most useful things I learned in mathematics.

Gauss-Newton algorithm, GNSS (GPS) solution solving, financial calculus (Ito calculus), and more.

The Google acquisition of Fitbit was only approved by the EU under the stipulation that they can't offer any 1st party APIs for Fitbit/Google wearables that aren't available to 3rd party wearables.

Apple Watch doesn't have to play by the same rules.

Of course, but that's not the issue that's being described here. The issue is 1st party APIs that only Apple or Google have access to, that 3rd party apps can't use, regardless of whether the use would like to grant them permission.

You could secure the parked car so that it doesn't move, and its crumple zone will still slow the deceleration.

Consider this: would you rather drive head-on into a parked car at 100mph or a brick wall?

> Also, the GNSS software in most phones is sadly unable to accept the correction data from any of these systems, regardless of whether it's a nationwide network or your personal setup. This is purely a software limitation on the vendor GNSS stack, but sadly there is not enough demand for this. (An app will not fix this, we're talking vendor specific low level system code here.)

I don't think that's true. Android surfaces raw GNSS measurements including carrier phase (sub wavelength measurements) to do centimeter level positioning through the raw measurements API [0].

There's even an API to specify the phone antenna phase pattern to correct the carrier phase measurements (source: I implemented it [1]). For those that aren't familiar, the idea is that the antenna pattern on phones isn't perfectly symmetrical, and depending on the direction of the incoming signal, it may appear longer. Knowing the antenna pattern, you can correct for this.

[0] https://developer.android.com/reference/android/location/Gns...

[1] https://developer.android.com/reference/android/location/Gns...

The raw measurements API provides raw GNSS measurements so that somebody can implement their own position engine, and do things like RTK.

The corrections mentioned in the blog post are calculated by the phone (in Google Play Services, using Google's 3D building models) using the raw GNSS measurements, and provided back to the chipset to improve its position estimate (correcting for reflections and interference).

The waste is one issue, but personally I am also worried about the consequences of the constant and inescapable exposure to plastics on our health.