Get Mystery Box with random crypto!

Pathetic low-frequenciers

Logo of telegram channel pathetic_low_frequenciers_eng — Pathetic low-frequenciers P
Logo of telegram channel pathetic_low_frequenciers_eng — Pathetic low-frequenciers
Categories: Technologies
Language: English
Country: Russia
Subscribers: 156
Description from channel

That's my personal channel of some crazy stuff. Daily I see a lot of strange things across the internet, so I decided to publish some of them here. Beware of: weird math, crazy pics, cybernercophilia, nerdish humor.

Ratings & Reviews

1.67

3 reviews

Reviews can be left only by registered users. All reviews are moderated by admins.

5 stars

0

4 stars

0

3 stars

0

2 stars

2

1 stars

1


The latest Messages

2021-12-12 17:38:56 Recently passed NaNoGenMo 2021(the "national novel generation month" I wrote about previously).
I didn't have time to participate this year, but I skimmed through the works and share with you my personal top:

1. John Lambert put together a system that cuts music videos into separate frames, runs them through a description generator, and collects text that retells what is happening in the video. The result is a book of 22 such retellings.

2. Kevan Davis sliced Moby Dick's text into pieces (mocking Moby Dick is a long-standing NaNoGenMo tradition) and put them together in an interactive game book-style adventure. It turned out quite well.

3. Mark Sample used an ageless trick — translated "Pride and Prejudice" by auto-translators in the following cycle: English->Russian->Chinese->Portuguese->English. The result is a book called "Pride and Injury."

4. John Ohno, impressed by the Dark Shadows TV series, put together a generator called Shark Dadoes, which produces dialogues consisting of evasive answers and question-to-question answers.

There were other funny things, like generating synopsis for films by title, director, and actors; replacing all matching words in Moby Dick with emoji; a neural network trained on the texts of Phillip Dick.

And if you don't like this, you can read the work on 58 ways to visualize structures in the text of Alice in Wonderland or our recent post on how to select and check validation metrics in classification problems (based on our paper from NeurIPS 2021).
56 views14:38
Open / Comment
2021-11-27 18:50:20 Sometimes I write digital archaeological posts, for example, about the origin of keyboard layouts, how r-pentomino was invented, or about April Fools' RFCs. Today I will write a little about the etymology of various computer terms.

Everyone knows that the word modem comes from a combination of modulator and demodulator - two devices used to convert digital information into a form convenient for transmission over analog networks and back. The word codec ([en]coder + decoder) and the less widely known slang terms like balun (balanced + unbalanced) and serdes (serializer + deserializer) have a similar origin.

Similar to a codec in spelling and sound, the name of the Kodak company, registered in 1888, has a different nature. The company's founder, George Eastman, wanted to invite a new word — short, easily recognizable, and pronounced in different languages. According to legend, he used a set of letters from the Anagrams game (the grandmother of the Scrabble game). One of the criteria for George was the use of his favorite letter K, which accounts for 40% of the result. The idea to make a new word was not entirely successful: in 1896, on the pages of the Amateur Photographer magazine, readers made a dispute, trying to find out the word's origin. It was found, for example, that in Hindustani (it came there from Persian), this word means "boy," and one of the readers pointed out the similarity with the Hebrew Kahdak.

The word bit in the sense of a minimum amount of information was first publicly used in Claude Shannon's 1948 article" Mathematical Communication Theory." Claude himself referred to the authorship of the mathematician John Tukey, who used bit as an abbreviation for binary [information] digit in internal documents of Bell Labs. The word byte (distorted English bite as piece) stands for the minimum amount of information processed at one time or directly addressed. Werner Buchholz first used it in 1956 in the design documentation for the IBM Stretch system. On different systems, bytes come in various sizes, for example, 4, 6, or 9 bits (the size of a byte can even be variable). To accurately indicate the size of a byte of 8 bits, it is common to use the term octet.

For engineering reasons, it is more efficient for computers to work with numbers that are powers of two. Therefore, engineers often understand the word kilobit as 1024 bits (2^10), but in some cases, it means 1000 bits (10^3, as with other measures, such as meters). For example, the 1968 year's edition of the Encyclopedia of Library and Information Science, on the same page, states that 1 kilobit is 1000 bits, and 1 kilobyte is 1024 bytes. A similar story with the prefixes mega, giga, and so on. All this confusion continued until the end of 1998 when the International Electrotechnical Commission finally came in and fixed it (no). Since then, according to international standards, kilobits should mean 1000 bits, and for 1024 bits, the term kibibit should be used. However, not everyone agrees with this: according to the Russian "Regulations on the units of quantities" from 2009, the term kilobyte is fixed anyway in the value of 1024 bytes.

To measure the data transfer rate, in addition to any kilobytes per second and kilobits per second, engineers sometimes use terms based on the word baud (for example, kilobaud). Baud in modern communication usually means the number of changes in the carrier frequency per second, so if, for example, the carrier uses two signal levels, then 1 baud is 1 bit per second. But this is not accurate because the bits are considered gross here, i.e., include any overhead information, such as error correction. These bauds are named after Jean Maurice Émile Baudot, a French engineer who, in 1870, invented the basic encoding for the telegraph (aka International Telegraph Code # 1).

Also, it turns out that the word android is almost three times older than the word robot (which turned 100 years old last year).
102 viewsedited  15:50
Open / Comment
2021-10-14 13:51:11
I already wrote about the self-organization of different creatures in swarms and algorithms for modeling such behavior. I'll write a little more:

1. GIF above is from an experiment by a Harvard Self-Organizing Systems Research Group; they made many very simple identical robots and tested swarm algorithms on them, forcing them to form the desired configurations. Video.
2. A team from the Max Planck Institute of Animal Behavior made the DeepPoseKit library, which uses object and pose recognition with neural networks to track the swarming behavior of animals and insects. Code, article.
3. The team of Alexander Mordvintsev (author of DeepDream) is studying differentiable cellular automata, where each cell is a small neural network interacting with neighbors, and all together, they are able to form a global configuration and restore it from damage. Interactive demo, short video.
161 viewsedited  10:51
Open / Comment
2021-09-02 13:48:40
Almost 100 years ago, Wolfgang Koehler conducted his famous experiment on sound symbolism. People were shown two pictures (the top row) and were asked to choose which of them was "baluba" and which was "takete." The majority of people chose a rounded baluba and an angular takete.

Since then, the experiment has been repeated with people who speak different languages, with two-year-olds, and so on. Researchers also tried changing the words, for example, to buba/kiki. In all cases, the effect was preserved.

Since multi-modal models have become very popular this year (I periodically write about them), Nearcyan from Austin decided to see what the CLIP model thinks about these words. In the second row, there are examples of generated images for kiki and buba, in the third — for the forms of "maluma" and "takete."

More details, pictures, and other words are in the original blog post.
50 views10:48
Open / Comment
2021-08-23 15:24:13
I recently wrote about the neural network generation of pixel graphics by Tom White.

Last weekend, I got to play with the code a bit and added a couple of optional features: palette enforcement and an additional loss for smoothing. It turned out unexpectedly well: check out the picture above with several results. You can find more images and a link to google colab are in my Twitter thread.
74 views12:24
Open / Comment
2021-08-18 17:25:38
Top left — the sculpture Trinity by Frank Haase, a translucent cube whose three projections are three different QR-codes. Bottom left — the QR Rubik's Cube with six different messages on different sides; I once made it as a birthday gift. Top right — my QR code, made using the approach described by Russ Cox in the excellent article QArt Codes. Bottom right — the three-layer code invented by Eckart Schadt: depending on the distance, the contrast of some pixels changes, and the code is read differently (it works very poorly from the screen, try the printout.)

Also, for the Internet connectivity: my old post on mirror QR codes generation.
108 views14:25
Open / Comment
2021-08-16 14:05:53
Tom White, an AI Artist from New Zealand, came up with an idea of how to generate pixel art images with VQGAN+CLIP networks. For the second week now, he is posting the neuro-pixel-art alphabet in this Twitter thread (he got to the letter W yesterday). I suspect a huge amount of cherry-picking; anyway, he promised to publish a colab soon, so you can experiment on your own.

If you like this, also pay attention to the 8-bit fan art episode of Rick and Morty, drawn by Australian animator Paul Robinson. And if not, check out how much of the Hitchhiker’s Guide to the Galaxy you can fit in a single QR Code, or read our recent article with Max Ryabinin about the cross-lingual neural networks solving Winograd schemas.
97 views11:05
Open / Comment
2021-08-04 20:15:34 As usual, there is not enough time for anything, so here is just another selection of strange/exciting things without any special comments (sorry if I missed some sources):

* Article which tries to describe Escher patterns in mathematical language.
* HumBugDB - the paper presents a large-scale acoustic mosquito dataset (submitted for NeurIPS 2021, but it will most likely be rejected).
* Archaeological investigation of the origin of the bubble sort algorithm (the authorship, the origin of the name, etc.).
* Video of restoration of the incredibly cool fractal vise.
* A DIY project for assembling a VGA card via Dima Kuznetsov.
* Video of the design of a spherical gear.
* Game mechanics of one-dimensional chess, via @backtracking channel.
* Announcement of the talk on the creation of a Tibetan typewriter (the talk will be in a week).
* JPEG XL graphic format is almost Turing complete (via Wolfram's Rule 110 automaton).
* Emoticons that are valid javascript code.
* Japanese Circular Forest Experiment.
* Doom Captcha.
95 views17:15
Open / Comment
2021-07-28 17:11:23 The other day, I found the colab with a TPU to run a shiny GPT-J-6B (this is kinda an open-source version of GPT-3). The env is brewed for 15 minutes in total, but then processing one prompt is fast enough, like, in 10 seconds.

I tried it to generate jokes about mathematicians in a bar, so I share the results with you (almost without cherry-picking):

A mathematician walks into a bar. A mathematician and a philosopher stand by the bar, in conversation, and a bartender is doing his rounds. The mathematician says to the philosopher, “You know that art and philosophy are branches of mathematics, don’t you?” The philosopher says, “Yes, I knew you’d eventually ask that question.”

A 0 of mathematicians walk into a bar, and they are immediately thrown out because "0 is used for zero" or some other way of saying "0 is a number".

An infinite number of mathematicians walk into a bar, and just one gets his research published. Who is that?

An negative number of mathematicians walk into a bar, and say "I'm actually going to be less than zero".

A transcendental number of mathematicians walk into a bar,
and hear one mathematician say, "Look, I've found the best beer in town."
This mathematician says, "And look, I've found the best steak in town."

A quaternion number of mathematicians walk into a bar,
A pair of discrete logarithms, A prime gap,
A clash of protocols, A geometrical proof,
A product of exiles, A cosmic lie,
A shape-changing goblet, A parallelogram of regular solids,
An obstruction course, A normal subgroup,
An impulse of bits, An irrational number of settings
...
123 views14:11
Open / Comment
2021-07-12 15:42:02
The boom around VQGAN+CLIP generated images and videos continues to increase. This video I found earlier today on Twitter, but that more interesting, is a fresh post on the Berkeley ML blog (the whole blog is good, actually). This post formulates an exciting idea that we see the formation of a new demoscene here.

Besides, read a scientific paper on searching for evidence of time travelers on the Internet and watch a video from an ancient computer game from Timothy Leary.
103 views12:42
Open / Comment