Chrome question for the IT crowd

[SemiSalt]

For the past couple months, I've been getting errors in how Chrome renders text as shown below.

2020-02-17_2005.png

In this particular case, it seems to have been triggered by a dollar sign. The beginning was originally (dollar sign)7500. It's not really a big deal, but it is very odd to have Chrome go walkabout.

[BeauV]

I don't use Chrome, because I don't like giving Google access to that much info on me, but this sounds like a bug in their HTML interpreter. They'd probably like to see what causes it. Do you find in on random web sites, within boards, ???? some old software doesn't conform to the current HTML standards.

[SemiSalt]

It's been going on for several months, plenty long enough for Google to fix it. So I doubt it's widespread.

[SemiSalt]

With the help of my son, I discovered it was caused by this Chrome extension:

[BeauV]

AH! {Loud forehead slap} of course it is!!!

TeX was written by one of my heroes Prof. Donald Knuth of Stanford. He also wrote the multi-volume Art of Computer Programming and played a wicked pipe organ. A great guy who chose to use the "$" as an escape character to delineate expressions to be formatted. There was a discussion about all this. But back in the day it never occurred to use that anyone would try to blend TeX and the not yet invented HTML.

History here.

[kdh]

AH! {Loud forehead slap} of course it is!!!

TeX was written by one of my heroes Prof. Donald Knuth of Stanford. He also wrote the multi-volume Art of Computer Programming and played a wicked pipe organ. A great guy who chose to use the "$" as an escape character to delineate expressions to be formatted. There was a discussion about all this. But back in the day it never occurred to use that anyone would try to blend TeX and the not yet invented HTML.

History here.

To a lot of math guys Knuth wrote the first and basically last word on Computer Science. TeX is pretty neat too.

[BeauV]

Keith, "The Art of Computer Programming" was the awakening that convinced me I absolutely needed to become a programmer. I was a boatyard worker with a philosophy degree who loved logic, The "Sorting & Searching" volume is still a dog-eared beat-up book that I adore. I would probably have wound up as a pro-sailor if it hadn't been for Dr. Knuth. As it is, I'm an aged-out compiler-writer still waiting for Volume 7.

Dr. Knuth is still a towering figure at Stanford. I was lucky enough to stumble into that place in '75, an accident that entirely changed my life. He was very kind to someone who was just starting out in a field he probably did more to define than anyone I can think of.

[SemiSalt]

I remember that I had immediate access to a copy of Knuth's book(s) back in the day, but I don't remember if I actually owned a copy. Long time ago.

Back when I was an honest-to-God Operations Researcher Analyst in the Military Industrial Complex, the book in which I found the most inspiration and the most answers was An Introduction to Probability Theory and Its Applications by William Feller. Giants, these guys.

Just by way of explanation, the TeX widget got installed in Chrome on my computer when I was trying to figure out how to get math formulas to render on my mostly neglected blog (https://phv3773.blogspot.com/). I can't even remember what I was thinking to post but probably it had to do with problem of integrating the error function (https://en.wikipedia.org/wiki/Error_function). This is a problem that goes back at least to Euler and which has had major advances (proving it's impossible) in the last decade or two. I was just trying to get a sense of why such a simple-looking problem was so hard.

But I never did get TeX on render on Blogger. The directions for using MathJax are all over the place, and they just didn't work for me. So, I blamed Google and quit.

[kdh]

I have Feller I and II on my bookshelf.

Great reference for the underappreciated "stable" distributions (those, like the Gaussian distribution, for which sums of stable-distributed random variables are themselves stable-distributed). None other than the Gaussian has a finite second moment. To most this is a big deal for reasons that have always escaped me. I think we'd be a more advanced civilization if we let the whole family into our lives rather than just the Gaussian. All the nonsense about fat tails being "not Normal (Gaussian)" is simply explained.

[Jamie]

I have Feller I and II on my bookshelf.

Great reference for the underappreciated "stable" distributions (those, like the Gaussian distribution, for which sums of stable-distributed random variables are themselves stable-distributed). None other than the Gaussian has a finite second moment. To most this is a big deal for reasons that have always escaped me. I think we'd be a more advanced civilization if we let the whole family into our lives rather than just the Gaussian. All the nonsense about fat tails being "not Normal (Gaussian)" is simply explained.

Would it be fair to say that not normal distributions are more common than normal and that calling a Gaussian distribution "normal" is a cultural projection on the observable world?

[kdh]

I have Feller I and II on my bookshelf.

Great reference for the underappreciated "stable" distributions (those, like the Gaussian distribution, for which sums of stable-distributed random variables are themselves stable-distributed). None other than the Gaussian has a finite second moment. To most this is a big deal for reasons that have always escaped me. I think we'd be a more advanced civilization if we let the whole family into our lives rather than just the Gaussian. All the nonsense about fat tails being "not Normal (Gaussian)" is simply explained.

Would it be fair to say that not normal distributions are more common than normal and that calling a Gaussian distribution "normal" is a cultural projection on the observable world?

Exactly. I make a point of not calling the Gaussian distribution "normal."

Reminds me of a story about Gauss. As a precocious child he was told by a teacher to compute the sum of 1 to 100, i.e., 1 + 2 + 3 + ... + 100, thinking it would keep him out of her hair for a while. He scribbled a bit and came up with the answer practically immediately. Can you guess the trick he used? Hint: the answer is 101*100/2.

[BeauV]

I have Feller I and II on my bookshelf.

Great reference for the underappreciated "stable" distributions (those, like the Gaussian distribution, for which sums of stable-distributed random variables are themselves stable-distributed). None other than the Gaussian has a finite second moment. To most this is a big deal for reasons that have always escaped me. I think we'd be a more advanced civilization if we let the whole family into our lives rather than just the Gaussian. All the nonsense about fat tails being "not Normal (Gaussian)" is simply explained.

Would it be fair to say that not normal distributions are more common than normal and that calling a Gaussian distribution "normal" is a cultural projection on the observable world?

Exactly. I make a point of not calling the Gaussian distribution "normal."

Reminds me of a story about Gauss. As a precocious child he was told by a teacher to compute the sum of 1 to 100, i.e., 1 + 2 + 3 + ... + 100, thinking it would keep him out of her hair for a while. He scribbled a bit and came up with the answer practically immediately. Can you guess the trick he used? Hint: the answer is 101*100/2.

My son's formulation when he was in middle school: 50*101 "You have 50 pairs, it's obvious" I realized then that he was probably a bit better at this stuff than I am.

[BeauV]

Keith, I have to say that "Gaussian Distributions", "Bell-Shaped Curve", and the "80-20 Rule" have all had a terrible effect on people's decision making. They are mental crutches that lead to serious mistakes.

I have spent a lot of my life in business hunting out the opportunities at the profitable end of what folks think is a Gaussian distribution of outcomes (in this case potential upside for a deal). Anyone who does venture investing rapidly learns that the 80/20 rule does NOT apply to deals. Over 50 years of data from the firm I used to work with showed that across all our investments 2% of the deals provided 98% of the profits. The partners would consider you a dolt if you said something like "80/20 rule" when talking about results.

I'm sure you know that one of the other fields where this sort of thinking results in terrible decisions is in evaluating the performance of technical people, specifically programmers. IBM ran a massive study of their programmers' productivity in the '70s and again in the '80s. Both times they discovered that the spread of productivity amongst programmers was over 75:1, as Dr. Thompson of IBM said in a public speech:

A quote from memory: "Because we are unable to hire the absolute best programmers, they typically move into academia (EG: Knuth), and we try hard not to hire the below-average programmer, our research supports an estimate that the range of productivity in the pool of all trained programmers is over 115:1. This means that our ability to forecast the delivery of new software could have a margin of error of greater than 50:1. Said another way, if one doesn't track programmer productivity, then a project could take between one year and over 20 years depending upon the programmer to whom it was assigned. Moreover, we have absolutely no evidence that productivity correlates with any sex or race. The good news is that well documented highly productive programmers tend to continue to perform at breathtaking levels, but that they lose that ability during their later years. "

You could have heard a pin drop as he delivered that report. To this day, I've seen no evidence to show that people's ability to program has become more uniform. There is still at least a 50:1 observable difference, and many believe it's more like 100:1. Understanding how fat the tail is matters a tremendous amount, as does how far from the center of the "bell-shaped curve" that tail really goes.

Of course, that's what makes investing in and managing people so much fun. It's all about being better than the competition at finding those who are truly outstanding.

It also leads you to conclude that we're doing something very wrong if we have a 3:1 difference in pay for a 50:1 difference in performance. Yet, it's socially very hard to have more than a 4:1 difference in pay for the "same job". Sadly, this means that the "pay" is attached to the "job" and not to the "productivity." Thus, the ongoing opportunity for investors to motivate the extreme top end of the distribution to break out on their own so that they can early what they deserve based on productivity vs job title. (probably stating the obvious above, sorry about that.)

[kdh]

I have Feller I and II on my bookshelf.

Great reference for the underappreciated "stable" distributions (those, like the Gaussian distribution, for which sums of stable-distributed random variables are themselves stable-distributed). None other than the Gaussian has a finite second moment. To most this is a big deal for reasons that have always escaped me. I think we'd be a more advanced civilization if we let the whole family into our lives rather than just the Gaussian. All the nonsense about fat tails being "not Normal (Gaussian)" is simply explained.

Would it be fair to say that not normal distributions are more common than normal and that calling a Gaussian distribution "normal" is a cultural projection on the observable world?

Exactly. I make a point of not calling the Gaussian distribution "normal."

Reminds me of a story about Gauss. As a precocious child he was told by a teacher to compute the sum of 1 to 100, i.e., 1 + 2 + 3 + ... + 100, thinking it would keep him out of her hair for a while. He scribbled a bit and came up with the answer practically immediately. Can you guess the trick he used? Hint: the answer is 101*100/2.

My son's formulation when he was in middle school: 50*101 "You have 50 pairs, it's obvious" I realized then that he was probably a bit better at this stuff than I am.

Here's the trick. I think I would describe it as "If you double the sum you have 100 pairs that each sum to 101." I'll explain. Write the sum twice, the second time in reverse order:

1 + 2 + 3 + ... + 100
100 + 99 + 98 + ... + 1
--------------------------------
101 + 101 + 101 + ... + 101 = 101*100

There are 100 elements in the sum of the pairs, and since you've doubled the sum by including the reversed sum the answer is 101*100/2.

[BeauV]

I have Feller I and II on my bookshelf.

Great reference for the underappreciated "stable" distributions (those, like the Gaussian distribution, for which sums of stable-distributed random variables are themselves stable-distributed). None other than the Gaussian has a finite second moment. To most this is a big deal for reasons that have always escaped me. I think we'd be a more advanced civilization if we let the whole family into our lives rather than just the Gaussian. All the nonsense about fat tails being "not Normal (Gaussian)" is simply explained.

Would it be fair to say that not normal distributions are more common than normal and that calling a Gaussian distribution "normal" is a cultural projection on the observable world?

Exactly. I make a point of not calling the Gaussian distribution "normal."

Reminds me of a story about Gauss. As a precocious child he was told by a teacher to compute the sum of 1 to 100, i.e., 1 + 2 + 3 + ... + 100, thinking it would keep him out of her hair for a while. He scribbled a bit and came up with the answer practically immediately. Can you guess the trick he used? Hint: the answer is 101*100/2.

My son's formulation when he was in middle school: 50*101 "You have 50 pairs, it's obvious" I realized then that he was probably a bit better at this stuff than I am.

Here's the trick. I think I would describe it as "If you double the sum you have 100 pairs that each sum to 101." I'll explain. Write the sum twice, the second time in reverse order:

1 + 2 + 3 + ... + 100
100 + 99 + 98 + ... + 1
--------------------------------
101 + 101 + 101 + ... + 101 = 101*100

There are 100 elements in the sum of the pairs, and since you've doubled the sum by including the reversed sum the answer is 101*100/2.

Cool way of describing a way to think about this!!

[kdh]

Beau, your son's 50 pairs of 101 works as well. Just write the sum as:

1 + 2 + 3 + ... + 50 +
100 + 99 + 98 + ... + 51
----------------------------
101 + 101 + 101 + ... + 101 = 101*50

Obvious! Always is once you know the trick.