“It’s better to learn new and complex things in ‘combat’ conditions”: AI Junior winners about their projects and plans for the future

AI Journey Junior is an online conference for middle and high school students. With its help, children can improve their skills in artificial intelligence and data analysis, listen to master classes and lectures from industry leaders and create their own projects.

Participants in a recent session under the age of 18 talked about their journey into Data Science and Machine Learning, their competition and their goals for the future.

“It’s better to learn new and complex things in” combat “conditions”: AI Junior winners about their projects and plans for the future

Maria Peredok

Alexander Goptar

I started my journey in Data Science and Machine Learning three years ago by participating in the NTI Olympiad on Artificial Intelligence. It was both a cool get-together and an opportunity to develop in the field of advanced IT technologies. I started doing this more rationally and consciously last year. The lockdown influenced, there was a lot of time for self-development.

Now I work as a data analyst at the NTI 2035 University, as well as on launching a startup Сommon Data in the role of ML-engineer. We are making a platform for collecting data into projects using artificial intelligence. I am directly working on a project pre-moderation system using machine learning algorithms to detect illegal use of personal data.

The experience gained in Natural Language Processing in AI Journey helps me a lot in this, namely, solving the problem of generating an essay. Initially, a basic solution was given.

There are always two ways:

• finalize the existing project;
• start from scratch.

I supplemented the proposed essay base, then manually selected the most correct ones, processed their structures according to the USE criteria. I created special labels in them, where it is written: introduction, argumentation, opinion and conclusion. RuGPT can be effectively trained to highlight these places in the text. Further, after training the neural network for generating essays, using additional functions, he brought the final product into a readable form.

In addition, I had an additional hypothesis that separate models should be taught for each subject, but this division gave the worst result. The data is reduced by about four times, and the neural network begins to get confused in the structuring of the text.

In AI Journey, I pumped myself into text generation and NLP, and also discovered a cool technology that I really want to use in some of my projects.

Daniel Zimmerman

Now I am a first-year student at ITMO. I entered there thanks to the joint profile “Artificial Intelligence” of the NTI Olympiad and the “Academy of Artificial Intelligence” for schoolchildren. I got into Data Sceince and Machine Learning in grade 10. Before that I was engaged in web development, but decided to try to participate in the “Big Challenges” and began to study neural networks on my own. As a result, I passed the regional selection from my native Primorsky Territory with a prototype program that analyzes costs and gives advice on how to save money.

After that, I began the stage of a more systemic and in-depth study of artificial intelligence. I took courses on Coursera, improved my knowledge of mathematical analysis, probability theory and linear algebra. Then there was a shift at Sirius. This made it possible to understand how machine learning works, to see mathematics in action.

In parallel with this, I tried to compete, I wanted to try and test myself, but until the beginning of 2020, everything was unsuccessful. Then came the first results: tenth place in the Olympiad in artificial intelligence NTI and the “AI Academy”, other competitions where it was possible to take prizes, and, finally, victory in the “adult competition” from Sberbank to form time series on the spread of coronavirus. Since then, I have tried not to miss interesting AI competitions.

AI Journey attracted me with large prize money and an interesting applied task – I needed to make a model that can decipher the manuscripts of Peter I in the picture into a digitized text.

Before that, I have never tried working with optical character recognition. Now I have experience with a problem that is common in the real world, and I can use it.

In it I see myself as a tech entrepreneur with a package of profitable startups. AI technologies are best suited for this because of their applicability in all areas. There is no need to worry about the market value of what you can do, as your knowledge and skills will be applicable in any area.

Ivan Konyushenko

Artificial intelligence came to my attention a year and a half ago, in the ninth grade. In 2019, our school opened the IT-Class (Moscow educational project). While participating in the project, I studied data analysis and machine learning, as well as mathematics with a focus on linear algebra. This helps to better understand this area.

Part of my free time outside of school, I studied Data Science on courses from the leading companies in the country. We used them to solve many different problems. They were educational in nature. I understand that they are needed to master new knowledge, but I really wanted to rise to a higher level. I became curious about tasks from different hackathons, in which you need to show your ingenuity and research skills. In addition, in them I was attracted by the connection with reality, with practice.

When I saw the description of the NoFloodWithAI problem on the development of an algorithm for short-term forecasting of water levels in the Amur River to prevent emergencies in the regions of Russia, I fired up and decided to participate in the AI ​​Journey. It was especially exciting that it could benefit people and reduce flood damage.

While working on the solution, I started by studying the theory: I had to understand the analysis of time series. Then I managed to write the code and take the junior track prize.

Through my participation in AI Journey, I became convinced that machine learning and data analysis can be used for a wide variety of research tasks. This is what attracts me the most to these technologies.

My immediate plans are to win the joint profile “Artificial Intelligence” of the Olympiad of the NTI CD and the “AI Academy” for schoolchildren and enter the university according to its preferences.

Krasnoyaruzhsky Sergey

Artificial intelligence became really interesting to me thanks to my teacher Vitaly Prakhov. He is not yet 30 years old, and he can communicate with us in the same language. Despite this, he is a very cool specialist and very passionate about his subject. It motivates and energizes us.

We worked with him on many projects. We made neural networks that can generate news headlines and be used for text analysis. He prepared me for the AI ​​Journey, as well as for the Olympiad in the field of artificial intelligence NTI and the “AI Academy” for schoolchildren, in which I won last year and still participate in this. Participation in such competitions allows you to make a big breakthrough – sometimes it is better to learn something new and difficult in “combat” conditions.

I participated in two AI Journey tracks:

• transcript of the manuscripts of Peter I;
• generation of complex texts.

You just had to sit down and write the algorithms. The second task was more practical and interesting for me. Solving it, I got acquainted with the ruGPT-3 technology, which can “write” meaningful texts with just one request in natural language.

In the end, I managed to create a solution that generates an essay on social studies at the USE level in the 11th grade. The resulting compositions were difficult to distinguish from those written by people. This earned me a prize in the junior special category.

Results like this are very cool. I hope that in the future I will be able to make a breakthrough in the development of AI at the global level and make it qualitatively stronger. In general, the topic of strong artificial intelligence is especially interesting to me. People need such ambitious goals and will distract them from unimportant things. I like not only the process of solving problems and working with neural networks, but also reasoning on the topic of AI. Without this, machine learning becomes something common or, as they say, “loses its charm.”

Cover photo: Unsplash