‘A Cat That Jumps from the Floor to a Shelf Doesn't Know Newton's Laws’

Anastasia Malashina defended her thesis on a topic related to cryptographic methods of information protection and is now engaged in applied projects in strategic analytics. In an interview with the Young Scientists of HSE University project, she discussed the difficulties she faced during her studies and explained what cryptography is and why large language models will not replace human intelligence.

How I Got into Science

My journey into science began in high school, where I became interested in mathematics. I planned to study a theoretical field, but my set of exams limited my choices to applied mathematics-related majors. At HSE University, I studied applied mathematics and computer security, eventually choosing the latter despite not originally considering information security. After completing my specialist programme, I decided to pursue a doctoral programme immediately. A degree diploma wasn't enough for me; I wanted to obtain an academic status.

What I Studied

During my dissertation, I focused on cryptographic methods for information security. I was offered an opportunity to pursue a specialised direction related to keyless reading during my final years of study. I started researching this topic in my final year of the specialist programme and continued exploring it in graduate school. Eventually, I defended my thesis on this subject.

I found this topic interesting because it allowed for an interdisciplinary approach. Mathematical methods of cryptanalysis could be combined with the analysis of natural language text.

Methods of mathematical linguistics are not typically covered in information security courses. This can create a challenge in terms of terminology, as linguists and cryptographers often use different terms to describe similar concepts. In my research, I sought to bridge this gap by creating a more unified language for discussing these ideas.

As part of my research, I have been working on applying an information theory approach to the analysis of algorithms used for information security. Imagine trying to decrypt intercepted messages or their individual parts by going through all possible combinations. How can we distinguish between chaotic combinations and those that might actually be variations of the original text? To answer this question, we need to take into account the statistical characteristics of natural language texts. These characteristics can be approximated and formalised using probability theory and mathematical statistics, for example.

What is Cryptography?

Cryptography is the science of mathematical techniques used to protect information. For example, WhatsApp communication relies on cryptographic algorithms for encryption, while electronic digital signatures—like those implemented on Gosuslugi—are built upon cryptographic schemes.

In the Soviet Union, cryptography was a highly secretive subject and this word was not even mentioned in public media. Later, some of the cryptographic research became public, but many other studies remained classified.

What Qualities are Important for a Researcher

I once heard that one of the most important qualities for a researcher is the ability to take criticism into account quickly and bring their work up to the required standard. Last year, at the end of April, there were several comments made about my work during a departmental seminar. The committee expressed concerns that I might not be able to make the necessary changes before the fall deadline. However, within a month, I had finalized the manuscript and incorporated several new experiments. These ideas had occurred to me during the research process. Contrary to expectations, I submitted my dissertation proposal for defence as early as June.

I'm still amazed at how I managed to publish my articles in necessary journals and meet the requirements for defence regarding publications. All my major dissertation-related papers were published without co-authors.

What I Do

Strategic consulting and technology analytics are my areas of expertise. I particularly enjoy the project-based approach, as it allows me to focus on specific tasks without being tied down by daily routine. When conducting analytical research, creativity is essential in finding new approaches and solutions. This process resembles doing scientific research, where there is no predetermined methodology, and we must work within the realm of uncertainty.

How Does Analytical Research Compare to Scientific Research?

Scientific research has specific goals, such as acquiring new fundamental knowledge, testing hypotheses, and identifying patterns. A dissertation aims to contribute to the advancement of a specific field of study. Science seeks truth through research and the findings are documented in scientific articles, dissertations, and monographs.

Analytical research, on the other hand, is applied research that answers specific practical questions by turning data into solutions. For example, strategic consulting addresses questions like ‘what is happening and why?’ and ‘how should we respond?’ and provides answers through analytics. The deliverables of business analysis can take various forms, such as reports, summaries, presentations, or white papers, depending on the project duration and the client’s requirements.

But there is another interesting format—popular science texts. These are concise analyses designed for a general audience. These articles provide an overview of the latest trends in science and business, without going into the technical details. This allows readers to understand the broader implications of these trends on their daily lives.

Why Does an Analyst Need a Broad Outlook

Writing about the latest trends in technology requires keeping abreast with scientific achievements in different fields. A person without specialised knowledge in a particular subject will not fully understand the fundamental aspects of what is happening in that field. However, it is essential to have a general understanding in order to navigate the field quickly.

Currently, the hot topic is large language models (LLMs). These models are already used in various applications, improving efficiency in businesses and scientific research. Nevertheless, LLMs have nearly reached their limit. They have been trained on vast amounts of human-generated text, and adding more data would only result in minor improvements.

A cat that jumps from the floor to a shelf may not know Newton's laws, but it performs the jump with remarkable accuracy. It relies on its own empirical experience. Both humans and animals possess the ability of proprioception, whereas language models do not. These models do not understand the world we live in, and this situation cannot be fixed by simply adding more text.

Do I Have Burnout?

I don’t experience burnout in the traditional sense. However, after completing my graduate studies and starting to work on my dissertation before the preliminary defence, I encountered psychological resistance for an extended period of time. When you are constantly facing subjective criticism and dealing with problems on your own, it's easy to become apathetic. The energy of unfinished work can also be overwhelming, as the well-known Zeigarnik effect suggests. This was ultimately what motivated me to finish my dissertation.

Other Interests Besides Science

Recently, I have taken an interest in interior design and lawn tennis.

Advice to Young Scientists

Think ahead about the potential of your research and how it fits into the current context. Young researchers from other fields often face the challenge of choosing a topic that has been well-researched, but my advice is to avoid focusing too narrowly on topics with little known information.

Starting a study on a subject with little practical significance may not be worthwhile. It's not just about defending your research successfully. Your research can be commercialised, receive funding, and be promoted in popular science media.

Therefore, it's important to understand how well-known your chosen topic is within the expert community. Your supervisor should be interested in the subject, as should a few other people at the university. Having access to different experts can help you receive an objective assessment and seek advice as you prepare your work. One person's perspective may be limited.