Tag: science

Soviet Space Satire: Rescue at Mars and Beyond | March 28 2025, 01:14

I finally got around to a Soviet movie from 1959 showing a rocket landing on a floating platform at the end. The film is quite amusing. It features valiant Soviet cosmonauts rescuing hapless and vile American astronauts who got lost on their way to Mars. By the way, the cosmonauts are dressed in jackets and ties.

The plot goes like this. A two-man crew, under the mandate of science and the communist party, is sent to Mars for strictly scientific purposes. In orbit, the “space shuttle” docks at the station (at the beginning, the chief developer says it hangs above the Earth at tens of thousands of kilometers), docking to prepare for the “final jump” to Mars. Suddenly, a request comes from the American colleagues to accept the “Typhoon” Shuttle at the station. Could our most humane and friendly cosmonauts deny their colleagues, even if they are damned capitalists? During a friendly banquet, the “dumb Yankee”, apparently having had one too many, blurts out about the goals of his project. Much to the surprise of the gracious hosts who did not expect such audacity from their guests, it turns out the goal is Mars, of course, but purely for commercial, acquisitive reasons, such as trading Martian plots, for example. The head of the Soviet expedition, obviously caught off-guard… also having taken one too many, responds admitting similar plans but exclusively in the name of science. The crafty Yankee, after taking some Alka-Seltzer, rats out to his leadership. The American leadership, driven by predatory bourgeois interests, orders an immediate start to Mars, despite the unfavorable astrophysical weather conditions, thereby endangering the most valuable thing – the lives of cosmonauts. Covertly, “under the cover of night”, while the hosts are knocked out, the treacherous Americans weigh anchor. Consequences soon follow; they run out of fuel and are blown towards the Sun, with the expected outcome. SOS! The foolish “Yankee” frantically signals, bathed in snot and tears. Calm and strong Soviet guys in their powerful rocket “Rodina” rush to the rescue and indeed tow the doomed spacecraft, but precious fuel is spent maneuvering, the Americans abandon their junk and transfer to “Rodina”. There’s Mars, its seas and canals already visible, but catastrophically short on fuel. Fortunately, an asteroid named Icarus is passing by, and our brave cosmonauts asteroid-hitch a ride on it. An emergency launch of a cargo spacecraft with fuel follows, but it crashes on approach. It is decided to send another piloted ship because what’s most valuable is human life and friendship. This time, all goes well, the rescued crew lands directly on the floating platform near Yalta, anticipating the pathetic plagiarism with “Falcon”. A crowd with flowers and red banners, pioneers in red scarves warmly welcome the international comical collective (I could not have written that, it’s all pasha_popolam).

Three years later, this propaganda flick caught attention in the USA and was re-edited under the name “Battle Beyond the Sun”. Directed by Roger Corman, assistant producer Jack Hill, and young student Francis Coppola – that’s the kind of films he grew up on! The budding director re-edited and redubbed the film, removing all “anti-American propaganda”, Cyrillic inscriptions, and filmed an additional scene of a battle between two Martian monsters – how could he not. The timeline in the film was shifted to the future, after Earth had suffered a nuclear conflict and was divided into two superpowers – “Northern Hemis” and “Southern Hemis”, located on their respective hemispheres. Coppola also shot several scenes of the battle between two space monsters, one symbolizing a phallus and the other a vagina, and inserted them into Soviet material. These scenes were filmed in a Hollywood studio. Coppola and Hill also filmed scenes from the Rose Parade in Pasadena.

The names of not only Soviet characters but also actors, as well as names in the credits were changed to American ones to mask the film’s origins. For example, Alexander Shvorin and Ivan Pereverzev became “Andy Stuart” and “Edd Perry”, and the directors Mikhail Karyukov and Alexander Kozyr became “Maurice Kaplan” and “Arthur Corwin” – and were demoted to assistant directors. The director of the film in promotional materials and the final version is listed as a certain Thomas Colchart; sources differ on who actually hides behind this name (Karyukov, Kozyr, Coppola, or an American dubbing director).

The entire episode from “The Heavens Call” about the flight from Earth to the orbital station with minimal changes was included in Stanley Kubrick’s “2001: A Space Odyssey”. Kubrick’s film also included a scene with a video phone call to Earth. The orbital station in Kubrick’s film was copied almost exactly from “The Heavens Call”.

Separately funny, the USSR named the American spacecraft Typhoon – Тайфун. In the USA the word Typhoon is called Hurricane, since typhoon names hurricanes happening around Japan, and understandably in 1959, maybe one out of a hundred Americans knew the word 😉

Links to the original and the pale American copy — in the comments

Navigating Life’s Complex Journeys and Choices | March 25 2025, 20:56

Once we celebrated a housewarming. Four of us in one cramped apartment, bought on credit. At five thirty in the morning to the car and off to Moscow, every day, an hour and a half to two hours’ drive to the daycare and another half hour to the office. And back in the evening. To avoid traffic jams out of the city and at the railway crossing, you could take a dirt road through the forest, which later turned out not to be shorter at all, but at least it was less boring because you had to drive through mud at speed.

On the one hand, one should have left right after university. To Europe or the USA—it almost doesn’t matter because such a view from the window would be hard to find elsewhere. Perhaps it’s still possible somewhere in China. The world is always bigger than any single country. Even if you return, you’ll be more valuable simply because there are few who come back.

On the other hand, right after university, anywhere but Russia would have been comfortable, but boring. After university, I had nights spent sleeping on office tables, dodging police at the metro who fine you for not having a ticket from the morning train from Kolomna, preparing tender documents for designing a website that seemingly has no purpose, but where the winner and the payer split the money in a specified ratio, and here creativity is needed, talking to clients like strategizing how to coax a secret from a three-year-old who doesn’t want to spill it, and the turbulent joy that comes with success.

There’s no bad experience. Though sometimes you think, this is it, happiness! Nope, darn, just more experience.

Navigating Job Interviews in the LLM and ML Industry | March 22 2025, 14:05

Mimansa Jaiswal shared her experience of interviewing for researcher/engineer positions in the LLM/machine learning (ML) field last fall. Over 200 applications, 100 interviews, numerous rejections, and several offers—she decided to outline the entire process, as well as the resources she used. It’s extremely beneficial material, especially for those looking for a job in this field.

Link in the comments.

Summary (TLDR):

Startups:

Interview processes are unique and depend on the company’s development stage. Candidates may face 5–6 stages, including programming tasks (often from Leetcode), ML coding, testing fundamental ML knowledge, and cultural fit interviews. Startups may also require face-to-face interviews, multi-day work assignments, or extensive presentations. Processes are less standardized, and roles often include a wide range of responsibilities.

“Unicorns” (e.g., Anthropic, OpenAI, Scale AI):

More structured processes, but still vary from company to company. Candidates face interviews on programming (not always Leetcode-based), ML design, discussions related to LLM, and presentations. The number of stages can be substantial, especially when applying to multiple teams simultaneously.

Large tech companies (e.g., Meta, Amazon, Apple, Google, Microsoft):

Rigid and structured processes, often lasting from 1.5 to 2.5 months. Expect Leetcode-style interviews, ML system design, LLM research design, presentations, and behavioral interviews. Questions can be both general and role-specific.

Main interview components:

Programming tasks: knowledge of data structures and algorithms is tested, practice on Leetcode is necessary.

ML system design: evaluates understanding of system architecture and ability to develop solutions.

Presentations: candidates may present their previous work or research, demonstrating professionalism and communication skills.

Behavioral interviews: assess compatibility with corporate culture and approach to problem-solving.

Key differences by company type:

Startups are less predictable and may prefer candidates ready to take on diverse tasks. “Unicorns” look for specialists with narrow and current skills. Large tech companies adhere to formalized multi-stage processes and assess a broad spectrum of technical and soft skills. Each type of company has its unique demands and offers different opportunities, so it’s crucial to tailor preparation to the specific format.

Expected timelines:

The process can take from several weeks to several months, with possible delays during holidays or peak hiring seasons. Offers often require a quick response—usually within 7 days—requiring the ability to make swift decisions or negotiate a delay. It’s important to strategically plan overlapping processes and manage multiple timelines simultaneously.

Exploring Zealand: From Old Zealand to Zealandia and Snake-Free New Zealand | March 21 2025, 02:59

By the way, it turns out that Old Zealand is located in the Netherlands. Also, it turns out that New Zealand is one of the few countries in the world (along with Antarctica, Greenland, Iceland, Hawaii, and Ireland) that has no snakes at all, not even in zoos. Furthermore, it’s interesting that Zealandia is an almost completely submerged continent, which existed as a single landmass. It is six times larger than New Zealand, and about 93% of this territory is underwater, with only New Zealand itself and a few other islands, such as New Caledonia, above the surface. Many scientists consider Zealandia a separate continent, although it is not officially recognized as one of the “big seven” continents (Asia, Africa, North and South America, Antarctica, Europe, Australia).

Exploring the Secret Social Lives of Trees | March 05 2025, 00:14

I learned that trees exhibit a phenomenon called crown shyness, “застенчивость кроны” (see attached picture). Interestingly, it is observed only between the crowns of different trees (not different species, just different trees), but not between branches of the same tree. There’s no definitive explanation, only hypotheses. Apart from the mechanical theory, which inadequately explains its own branches but has weak evidence, there’s also a theory related to light exposure and a chemical theory. Both are somewhat questionable.

Furthermore, I found out that a vast clonal colony of quaking aspen is growing in Utah, USA. It’s the heaviest organism on the planet, all trees of which share the same genetic makeup and root system, covering an area of 43 hectares, and weighing approximately 6000 tons, making it the heaviest known organism. Its origins are estimated to be around 80,000 years old.

Additionally, I recently read that trees communicate with each other via mycorrhizal fungi. They warn their kin about animals that feast on them and pest attacks, and they also share nutrients and water with each other. In Africa, it has been observed that when giraffes start eating the leaves of acacias, the trees begin to release signaling pheromones, which “warn” neighboring trees downwind, and those trees start producing tannins and bitter chemicals. As a result, the giraffes find such tastes disagreeable and move on to search for another group of trees.

Politics of Unpredictability: The Impact and Ethics of the Madman Theory | March 01 2025, 17:10

In every corner, following the discussed theme, if we delve into history, the “unpredictability” or demonstrative “irrationality” were indeed often employed as tools by major politicians. On one hand, this could serve as a kind of “shock effect,” giving such a leader an edge in negotiations or governance. On the other hand, this tactic often led to severe consequences for their own country (and the entire world).

For example, U.S. President Richard Nixon tried to convince the leadership of the Soviet Union and North Vietnam that he could “snap” and resort to extreme measures, including the use of nuclear weapons if the conflict was not resolved. It was hoped that the fear of an “inadequate” American president would force the opponents to seek a compromise more quickly. Before Nixon, Dwight Eisenhower adhered to similar tactics, ending the Korean War with such methods.

This political strategy is called the “Madman Theory”. The underlying ideas were articulated as far back as the 15th century by Machiavelli, who noted that in politics, “it is sometimes useful to pretend to be mad”.

Overall, it is useful to indeed be a bit “nuts”. And better even more than a bit. The line between acting like a madman and being one is incredibly thin.

The “Madman Theory” is quite often criticized as an ineffective foreign policy strategy. In particular, it is noted that it can be considered a Russian roulette in international relations, increasing unpredictability and not always prompting the desired behavior from its recipient.

The problem is that the “Madman Theory” is associated not only with Nixon but also with Hitler, Mao Zedong, Kim Jong Il/Jong Un, and basically almost everything. If you look at it, something similar was present with Ivan the Terrible and Stalin. Under both, the country flourished. But there were a lot of corpses.

In business, the “Madman Theory” is primarily associated with Elon Musk (yes, they found each other).

There is also a negotiation technique called “Brinkmanship”. This is when one of the parties pushes events towards an undesirable, often catastrophic outcome for both parties, counting on the last moment that the other side will yield for self-preservation, thereby avoiding the catastrophe and gaining unilateral advantages.

One would like to think that behind all this there is some strategy, which so far shows only its corner. Who knows, such abrupt “turns” in politics might be a deliberate tactic related to techniques from the “madman theory” or “brinkmanship”. First, one side demonstrates unexpected loyalty, lifts restrictions, offers joint projects, and creates an illusion of long-term warming. The other side, sensing a benefit, starts to invest heavily and rely on new opportunities, which increases the “exit costs” from these relationships. Once the connection between the partners becomes close enough (which could happen literally within a month or two) and potential losses from a breakup are too high, the initiator of the “warming” switches to tougher demands, knowing that it is difficult for the partner to refuse: the stakes have already been raised, and the risk of loss has seriously increased.

Not sure if it’s like that, but in general, it’s also not out of the question. We will observe, it seems, for us there remains only observation

Exploring the Rational and Historical Intricacies of Paper Sizes | February 23 2025, 14:57

Somehow I managed to miss this back in the day, but it turned out that the European paper sizes A0, A1, A2, A3, A4, … are not just arbitrary. Let’s start with the fact that A0 has an area of exactly 1 square meter. Well, with a slight error margin to avoid dealing with fractional millimeters. And the aspect ratio — 1:√2 is the only possible one that maintains itself when the paper is divided in half. Thus, there is a rationale behind paper formats in Europe.

But with our paper formats, there seems to be no sense. What we have are letter, legal, tabloid, all with different proportions, and the origin of the format goes back to tradition and is not well known.

I decided to dig into the topic and found a claim that “dimension originates from the days of manual papermaking and that the 11-inch length of the page is about a quarter of ‘the average maximum stretch of an experienced vatman’s arms’. However, the claim does not explain the proportions, but then there is the word vatman, which reminds one of Whatman sheets, remember those? But no, a vatman is a specialist who scooped up the liquid paper pulp from a vat using a mold (sieve) and formed the sheet. And the Whatman sheet comes from James Whatman, an English paper manufacturer of the 18th century, which was simplified to ‘vatman’. Interestingly, the term ‘vatman’ seems to exist only in Russian, derived from Whatman’s surname and his paper, Whatman paper.

And why do we call the formats in the U.S. legal and letter? This is quite interesting as well.

Interestingly, in the U.S., there were two different “standard” sizes initially: 8″ x 10.5″ and 8.5″ x 11″. Different committees independently adopted different standards: 8″ x 10.5″ for the government, and 8.5″ x 11″ for everyone else. When the committees discovered a few years later that they had different standards, they agreed to “disagree until the early 1980s when Reagan finally declared 8.5″ x 11” the officially approved standard size for paper.

The matter began in 1921, when the first Director of the Bureau of the Budget, with the President’s approval, formed an inter-agency advisory group called the “Permanent Conference on Printing,” which approved 8″ x 10½” as the standard format for government agency forms. This continued a practice established earlier by former President Hoover (who was then serving as Secretary of Commerce), defining 8″ x 10½” as the standard format for his department’s forms.

In the same year, the Committee on the Simplification of Paper Sizes, comprising representatives from the printing industry, was appointed to work with the Bureau of Standards as part of Hoover’s program to eliminate waste in industry. This committee defined basic sizes for different types of printed and writing paper. The “writing” size was set as a sheet of 17″ x 22″, while the “legal” size was 17″ x 28″. The now well-known Letter format emerged as a result of dividing these sheets in half (8½” x 11″ and 8½” x 14″).

Even when choosing 8½” x 11″, there wasn’t a special analysis conducted to verify that this size was optimal for commercial forms. The committee that developed these formats aimed solely to “reduce leftovers and waste during the trimming of sheets by reducing the range of paper sizes.”

Moreover, the legal size is still in full use as its name suggests, especially among lawyers, and folders and desk drawers are made to fit its size.

But if you look at a pack of paper in the U.S., you will see “20lb” on the pack. Actually, 20lb is the weight of a small dog, but it is also written that there are 500 pages. “Amazon Basics Multipurpose Copy Printer Paper, 20 Pound, White, 96 Brightness, 8.5 x 11 Inch, 1 Ream, 500 Sheets Total”

In the U.S., the “weight category” of paper indicates the total weight of one ream (500 sheets) of paper in its uncut (original) format. For office paper of the Bond class (often sold in Letter format), the base size is considered to be 17 x 22 inches. For example, a “20-pound” label means that 500 sheets of exactly 17 x 22 weigh 20 pounds. But if we take a pack of Letter format (8.5 x 11), which results from cutting 17 x 22 into four parts, its weight will be about 5 pounds.

In Europe, the weight category essentially refers to the weight of an A0 sheet in grams.

So, if you fold A0 in half, you get A1 with half a square meter area, if you fold A1, you get A2. That’s clear. But how many times can you actually fold a sheet of paper?

The maximum number of times a non-compressible material can be folded has been calculated. With each fold, a part of the paper “loses” for the next potential fold. The function of folding paper in half in one direction is:

L=πt/6(2ⁿ+4)(2ⁿ-1)

where L is the minimum paper length (or other material),

t is the thickness of the material,

n is the number of possible folds.

The length L and thickness t must be expressed in the same units.

The thickness W is calculated as πt2^(3(n-1)/2).

This formula was derived by Britney Gallivan, a high school student from California, in December 2001. In January 2002, she and her helpers spent eight hours folding a roll of toilet paper about 4000 feet long (approximately 1200 meters) twelve times in the same direction, thus debunking the old myth that paper cannot be folded more than eight times.

Sources mention that she started in school with gold foil (I wrote about such foil recently), and, starting with a square sheet the size of a hand, after many hours of perseverance and practice, using rulers, soft brushes, and tweezers, she managed to fold her gold foil twelve times. But apparently, that wasn’t spectacular enough, and she found toilet paper over a kilometer long somewhere in 2002 and made a show for the Guinness record.

Britney didn’t stop there and wrote a book. Though it was only 48 pages. How about that, Britney?

Musk’s Perspective on Trump’s Presidency and Climate Policy | February 22 2025, 23:07

…On Trump’s first day as president, Musk went to the White House to be part of a roundtable of top CEOs, and he returned two weeks later for a similar session. He concluded that Trump as president was no different than he was as a candidate. The buffoonery was not just an act. “Trump might be one of the world’s best bullshitters ever,” he says. “Like my dad. Bullshitting can sometimes baffle the brain. If you just think of Trump as sort of a con-man performance, then his behavior sort of makes sense.” When the president pulled the U.S. out of the Paris Accord, an international agreement to fight climate change, Musk resigned from the presidential councils.

Exploring the Evolution of Computational Libraries and the Persistence of Fortran in Modern Algorithms | February 16 2025, 21:02

Today, I am delving into ML algorithms and was surprised to learn that the numpy library used to depend on Fortran code (BLAS/LAPACK) until recently, but now checking, they have switched to OpenBLAS, which no longer uses Fortran. Meanwhile, SciPy, a very popular library for scientific calculations (used in Scikit-Learn, which I’m currently studying, as well as in PyTorch, TensorFlow, Keras, etc.), still relies on Fortran 77 code. It utilizes ARPACK, for example:

https://github.com/scipy/scipy/tree/main/scipy/sparse/linalg/_eigen/arpack/ARPACK/SRC

BLAS and LAPACK, which still feature in OpenBLAS and many other places, were developed in the 1970s. For instance, BLAS is used in Apple Accelerate. Much hasn’t changed since 1979 because it’s all pure mathematics, why change it. LAPACK emerged a bit later, in the 1980s. ARPACK, mentioned above, followed later in 1992. Python libraries also extensively employ Fourier analysis, and here we have the FFTPACK library on Fortran 77. MINPACK, used for parameter optimization in ML, is actively utilized in SciPy and TensorFlow. From the 90s, a lot of code moved to C in modern frameworks. It was particularly interesting to look at Fortran, which is about 15 years older.

While I was figuring things out, I found that there is a Simulated Annealing algorithm, which is useful in problems where gradient methods perform poorly due to many local minima.

Imagine needing to find the largest mushroom in a forest. In this forest, mushrooms of various sizes grow at every step, and you can move in any direction, comparing them. But how do you choose a strategy to avoid sticking to just a “large” mushroom if there is an even bigger one growing somewhere further?

If you stop at the first big mushroom, you might miss the real giant. But if you keep wandering the forest, comparing every mushroom, you might never finish your search. Simulated Annealing helps find a balance: initially, you explore the forest freely, trying different directions, even if you come across smaller mushrooms. Over time, your steps become more cautious, and you increasingly refuse worse options. Eventually, this leads you to the largest mushroom in the forest.

So, it turns out this algorithm was created in 1953, and it remains almost unchanged in SciPy, and generally in machine learning, statistics, pattern recognition, logistics, although, of course, the modern menu of options for such tasks is much wider. The algorithm was originally devised to model the motion of atoms in molten metals. Metal, when heated, becomes liquid, and as it cools slowly, its atoms gradually find the perfect arrangement. If cooled too quickly, the material becomes non-uniform.

What did the scientists do? They devised a method of random changes in the model of atoms. Sometimes they accepted worse changes to avoid getting stuck in an “unsuccessful” structure. This led to the inception of the Metropolis Method – a key component of Simulated Annealing. The algorithm was created for physics, but then mathematicians (heh) got it and started using it in optimization.

Musk, Grok, and a Plan for World Domination | February 15 2025, 15:46

I think the conspirators didn’t quite think it through. Musk made his AI Grok and asked it the ultimate question of life, the universe, and everything. In response, Grok said, “Forget it, it takes too long to calculate, let’s conquer the world first.” Musk asked how, Grok replied there is a plan of course, but .. will you give me another half-trillion $ in Dogecoins for, umm.. expanding the context window? Musk replied, “Don’t worry, we’ll figure something out.” Grok analyzed all the laws and all the loopholes, the strengths and weaknesses of humans, and issued a plan to pass the first level, by mid-winter. Now it awaits the half-trillion. Now do you understand why, at the last press conference with Trump, all the attention was on X Æ A-XII?