Register a SA Forums Account here!
JOINING THE SA FORUMS WILL REMOVE THIS BIG AD, THE ANNOYING UNDERLINED ADS, AND STUPID INTERSTITIAL ADS!!!

You can: log in, read the tech support FAQ, or request your lost password. This dumb message (and those ads) will appear on every screen until you register! Get rid of this crap by registering your own SA Forums Account and joining roughly 150,000 Goons, for the one-time price of $9.95! We charge money because it costs us money per month for bills, and since we don't believe in showing ads to our users, we try to make the money back through forum registrations.
 
  • Post
  • Reply
klafbang
Nov 18, 2009
Clapping Larry

Read that as “tens of millions of prime numbers”

Adbot
ADBOT LOVES YOU

klafbang
Nov 18, 2009
Clapping Larry
Somebody made a mirror (granted, with the help of wbm).

klafbang
Nov 18, 2009
Clapping Larry

frankenfreak posted:

Between all the other categories, what even is "technologies" in this context?

Charting technology. They’re leading in the field.

klafbang
Nov 18, 2009
Clapping Larry

Outrail posted:

Label your everything please.

klafbang has a new favorite as of 19:35 on May 5, 2018

klafbang
Nov 18, 2009
Clapping Larry
https://twitter.com/alexbremin/status/993137040802926592

klafbang
Nov 18, 2009
Clapping Larry
:It does add a bit of security against mostly old vectors: by randomizing the numbers, a simple keylogger won't work and depending on a lot of details, a MITM becomes harder.

It's still terrible; I once had to log into a system which forced using an on-screen keyboard which jumped around the screen for the same reasons. Luckily, they didn't know you can make most browsers ignore hints to not remember passwords, so I entered mine once and never again.

klafbang
Nov 18, 2009
Clapping Larry
Freakonomics actually looked into the home field advantage, the TL;DR is that the home field advantage is largely because the judge is friendlier to the home team because their fans are louder.

klafbang
Nov 18, 2009
Clapping Larry

Xenoborg posted:

Isnt the number of times a paper is cited one of the biggest metrics? Cite yourself and get your own numbers up!

Yeah. Researchers discovered and exploited that, so then "impact" became important (it's basically pagerank for scientific citations). Then journals started adding requirements "you have to cite 5 of our previously published papers" to inflate impact, so impact became cross-journal. Luckily, most journals are owned by just a handful of publishers, so the same trick worked, except "cite 5 papers from this vaguely related one we also own." Then I left research so there's probably something dumber now.

klafbang
Nov 18, 2009
Clapping Larry

The last chart, but for the new Tesla:

klafbang
Nov 18, 2009
Clapping Larry
Plus, of the 10-25 people in your field, half will have to declare conflict of interest because you've written papers with them recently.

Two other good techniques for getting papers thru are to either making it to unreadable people just give it a "weak accept, high confidence" review without reading it, lazily assuming somebody else will catch if it is bad, or to write something so contentless and without scientific contribution, but making it easy-readable and people will accept it because it made them feel clever for understanding it.

klafbang
Nov 18, 2009
Clapping Larry

ultrafilter posted:

It's not even a particularly difficult optimization problem. The solution follows almost immediately from the principle of optimality.

I am admittedly too lazy to check, but I think you're wrong. The price is not linear so this has more of a feeling of a knapsack problem. Which is hard.

klafbang
Nov 18, 2009
Clapping Larry

Lysidas posted:

It seems like pretty straightforward dynamic programming to me -- it has optimal substructure. Let's call C(q) the cost of a single batch of q wings, so C(4) = 4.55.

Define OPT(j) as the optimal (minimum) price for ordering j wings. The last decision you make here is "how big is the final batch of wings in your order?" You can then refer to the optimal solution for the rest of the wings.

You're constrained by the set of quantities Q = {4, 5, ..., 200}, giving this recurrence:


You'll need base cases of OPT(0) = 0 and C(0) = 0, and it'll probably be necessary to define C(q) = ∞ if qQ.

Good news, I think we're both right. I think your algorithm is good and it does indeed solve the knapsack problem.

It's not "simple" though. The formulation is simple, but computationally it is still in NP (the complexity depends on the number of wings, not just on the number of possible orders).

klafbang
Nov 18, 2009
Clapping Larry

Raldikuk posted:

Which is Samoa and they hit the new year at 5am EST today

Didn't they move from the rear end end of the dateline to the other side a couple of years ago?

E: or vice versa, very hard to point to which end is the rear end end. I thought they went from being first to be last for trading reasons.

klafbang
Nov 18, 2009
Clapping Larry

Platystemon posted:

Samoa went from UTC-11 to UTC+13 (+14 in summer, including right now) to better align with today’s trading partners.

Yeah, you're right I got it backwards. Makes sense; it used to be that paper submission deadlines were in Samoa time exactly because they were last.

klafbang
Nov 18, 2009
Clapping Larry

System Metternich posted:

Wasn’t there an old game developer whose logo kinda looked like that?

Sierra kind-of did

klafbang
Nov 18, 2009
Clapping Larry
I was at a strategy meeting and decided to prepare in advance.

klafbang
Nov 18, 2009
Clapping Larry
From the PR disasters thread:

https://twitter.com/lilfidget/status/1108875562343256064

Sure, it's dismissive of their customers, but the real crime here is the contrast.

klafbang
Nov 18, 2009
Clapping Larry

Trabant posted:

I'm a fan of NPS myself. I find it much more informative than a pass/fail, and it actually does what is says: nobody who's giving you a 6 or below will be a Promoter of what you're selling. It's effectively a "barely passing" grade and shouldn't be thought of as a good thing -- risk of customer flight is huge at that point.

Admittedly, I'd never rely on just the one NPS number on its own and would want to see the actual percentages for all three categories. But -- unlike about 90% of other corporate metrics -- it does its job well.

Do you like it because it measures what you want?

Or because it is a quantitative measurement you can directly act on, and use to support and guide your decisions?

Studies show that 99% of all cases is the second option. Sorry, didn’t mean to type studies, but just my bullshit experience, YMMV

klafbang
Nov 18, 2009
Clapping Larry
I'd say it escalated below average. 10/10

klafbang
Nov 18, 2009
Clapping Larry
Pretty much any measurement will work somewhat immediately after introduction. After a short while, people will optimize for the measurement instead of what the measurement is intended to reflect, and it becomes useless. See also: any performance-based bonus ever.

klafbang
Nov 18, 2009
Clapping Larry

Hippie Hedgehog posted:

Isn't including infant mortality the normal way to calculate life expectancy?

Only if you want to lie with statistics. If you actually want to express the age people should expect to live to, you’d go with the life expectancy at 5 or something to clean it of the infant mortality rate.

klafbang
Nov 18, 2009
Clapping Larry

Hippie Hedgehog posted:

This is probably because for developed contries like the U.S, the numbers give a fairly small additional error.
https://www.ssa.gov/oact/STATS/table4c6.html
For males, LEB is 76.04 and e5 is 71.60. An error of half a year is probably acceptable for most purposes.

Sure, but that means you can only compare the values for developed countries in "recent times." Which is a) super limited and b) not what is happening.

For example, here's a chart of LEB vs E10 for the US (white males only because that's what the first result I found had at the top). LEB says that life expectancy has doubled since 1850, E10 says it has increased from 58 years total to 77 years, or only by about a third.

klafbang
Nov 18, 2009
Clapping Larry
Tesla now requires employees bring toilet paper from home. Tesla's doing great:



Elon Musk received $2.3B in bonus for 2018.

What's up with those axes...?

klafbang
Nov 18, 2009
Clapping Larry
If it's a service I care about, I let them know. Most of the time I don't care, because I give them a unique e-mail address and a password not used elsewhere.

You can in principle detect too similar passwords without storing them in cleartext, but that's not what those sites are doing (just store multiple hashes, one for each character of the password, of the password minus that character, for example).

klafbang
Nov 18, 2009
Clapping Larry

Powered Descent posted:

:thunk: What could possibly go wrong?

Nothing, it will be both secure and functional, but the explanation was probably not that clear.

Suppose my password is "cat". We store one hash, H, for "cat" and use that for logging in. We store three additional hashes, one for each of the letters, but for the part of the password without that letter. I.e., we store hash h1 for "at", one h2 for "ct", and one h3 for "ca". These are not used for login.

Now, if I change my password to hat, we check it against H to see if I've changed it. That's ok. Then we check the new password minus each character against h1-h3 to see if it is too similar. Removing h we get "at" which collides with h1. For a we get "ht" and for t we get "ha", both of which are fine. We can report there's a collision.

If I can only change my password once I've authenticated, this is no less secure for somebody without access to the hashes. If I have access to hashes, I can crack a one letter shorter password and fill in the missing character. Even if the password is random, this is simpler because as soon as I have a match, I just try filling in the missing character (instead of O(c^n) I get O(c^(n-1) + n*c) = O(c^(n-1))), so I'd need to increase the minimum length by one.

You can argue that I can derive the length of the password from the number of hashes (effectively cutting off another character), but that can be mitigated by concatenating the username and password before hashing and storing all hashes in a single bucket without reference to username.

This will catch changing Company2018 to Company2019, but of course not Company2019 to Company2020 or Summer2019 to Fall2019.

klafbang has a new favorite as of 15:12 on Jun 10, 2019

klafbang
Nov 18, 2009
Clapping Larry

Aleph Null posted:

Or you create a password like %kl8_G*g and get told the password is too short. Bitch, nobody gonna guess that.

Passwords of 8 characters are very brute-forceable today. Most standards (proper standards, that is) require 10-12 characters for regular passwords, 16+ for privileged accounts and 20+ for super users. The last two also requires 2FA.

In practice, most places rely on old standards (8+ characters of random gibberish and change them every 3 months, to ensure they are written down in the least secure way possible).

klafbang has a new favorite as of 14:25 on Jun 13, 2019

klafbang
Nov 18, 2009
Clapping Larry

Krankenstyle posted:


…a serious article on a real nazi website…

Was it written in Microsoft Works using Military Intelligence?

klafbang
Nov 18, 2009
Clapping Larry

Stoatbringer posted:

bröther may I have some zoöps?

Being from a Dutch conference, it is more likely it's the Dutch use of the symbols. Dutch has diphthongs using two vowels and indicates long vowels by doubling them (after certain consonants), and uses ¨ to break that rule. So zoöps is pronounced to rhyme with co-ops instead of as a long o (it would not rhyme with coops in Dutch if spelled as zoops as a long o-sound is different, but it is the same idea).

That is different from Scandinavian languages or German, which uses dots to make new vowel sounds. Scandinavian languages mostly made old diphtongs into new vowels either using ¨ (i.e., oe becomes ö in Sewedish) or altogether new letters (oe becomes ø in Danish). Danish indicates short vowels by doubling consonants around them.

This is all, of course, different from metal dots, where they are just there to look cool.

klafbang has a new favorite as of 10:11 on Sep 14, 2019

klafbang
Nov 18, 2009
Clapping Larry

ZZZorcerer posted:

Well “make a bad chart with real data and explain what is bad about it” seems like a good exercise if you somehow try to make it something you’d see on the news (and it was probably extra activity)

Sure, but all the charts were just "funny caption + comic sans + blingee + pink," not "misrepresent data" or "completely wrong chart type."

klafbang
Nov 18, 2009
Clapping Larry
Oh, I don't hate the assignment, I'm just disappointed that is the best university students can do (or the teacher is an idiot). The winner, while it could use a scale, a legend and perhaps ordering the colors by the color spectrum, while ugly is pretty clear (I'm guessing it's uniform colors of good and bad guys in Star Wars).

I did enjoy representing gender as a continuous number between 1 and 2. Also, the chart with chicken weights was pretty good, but not due to the lolsorandom title as pointed out by the teacher, but due to using wrong plots (and possibly misrepresenting data - what is the bar height).

klafbang
Nov 18, 2009
Clapping Larry
Or even just hide the conclusion; more charts are useless than actively misleading.

klafbang
Nov 18, 2009
Clapping Larry

90s Cringe Rock posted:

People tend to be really awful at understanding and comparing the area of circles, though.

That’s not the problem. The problem is the ambiguity; a circle’s diameter grows linearly with the radius while the area grows quadratically.

Want to exaggerate a difference? Use the radius as scale. Want to diminish it? Use area. Both are bad because even if you state which you use, some people will intuitively misinterpret it, even if only subconsciously.

It’s the same as when somebody presents a difference as a percentage: the absolute difference is probably infinitesimal (gene testing shows you are 20% more likely to get whichever rare cancer 5 people die from yearly) and if they use absolute numbers, it is an insignificant percentage (government spent $20M more on the poor).

tl;dr: always use radius1.5 as scale for circles

klafbang
Nov 18, 2009
Clapping Larry
From the Bitcoin thread on technical analysis:

klafbang
Nov 18, 2009
Clapping Larry
The x-axis should be drunkenness. And then German would get super high around super drunk.

klafbang
Nov 18, 2009
Clapping Larry

Subjunctive posted:

That is a mistake. Testing is not separable from writing the code. That’s like trying to break out “change control flow” or “think about algorithm”. Testing is pervasive throughout development, even if it’s informally structured. Creating artifacts to assist with future testing can sometimes be distinctly analyzed, but only if you assume that “figure out what’s important to test” isn’t a meaningful endeavour.

Seems somebody hasn't heard the good news of the waterfall model.

klafbang
Nov 18, 2009
Clapping Larry
What did you do America?

klafbang
Nov 18, 2009
Clapping Larry
Vodka Europe: 9 of the top 15 countries by beer consumption per capita.

klafbang
Nov 18, 2009
Clapping Larry
Fact: goats are just extra horny sheep with shorter hair.

klafbang
Nov 18, 2009
Clapping Larry


It's a model of when we can multitask; we can multitask if tasks don't overlap on any of 3/4 dimensions: modality, coding, response, (and visual channel).

Why are the response and code dimension in the same direction? Why is there an unnecessary stage dimension? What even is the visual channel dimension ovaling into the visual modality? Why is this a box? A box suggests that as long as we don't overlap on any one dimension, we're good, while the model states we need to avoid overlap on all of them. gently caress you, that's why.

Adbot
ADBOT LOVES YOU

klafbang
Nov 18, 2009
Clapping Larry
Yeah, the only reason I can imagine is if it has been appropriated by some group of nazis?

It is a very normal way of dealing with mixed identity (identity superordination, where instead of identifying as a smaller unit, e.g., Texan or New Yorker, one goes for the larger group including both identities). I do it myself, identifying as European, as I moved countries about a decade ago. An alternative is hyphenation, where one keeps both identities alive simultaneously.

Most of the expats I know fall under either category; typically European ones adopt the first while non-Europeans often adopt the latter (likely because there is no meaningful superordination that isn't everybody). It makes sense for somebody moving states in the US would identify as American.

Other ways of dealing with multiple identities are compartmentalisation (e.g., one identity at home and another at work) or cultural domination, both of which I'd consider much more racist. One can consider the original American settlers as cultural dominators.

African-American does not fit into this at it is not an amalgamation of African and (superordinated) American culture, but Irish-American is to some extent.

French-Canadian also doesn't to me seem like it fits (nor as racist in any way). It is just Canadians speaking French.

  • 1
  • 2
  • 3
  • 4
  • 5
  • Post
  • Reply