Подкаст с Константином Воронцовым (ФУПМ'94): AI — это следующее электричество, интернет или атомная энергия?
Каждую технологию сравнивают с электричеством. А что если AI ближе к атомной энергии?
В новом выпуске подкаста «Кабинет Шрёдингера» — Константин Воронцов, выпускник МФТИ ФУПМ'94, один из основоположников российской школы машинного обучения. Д.ф.-м.н., профессор РАН, преподаватель МФТИ и ВМК МГУ, завлаб Института ИИ МГУ, создатель BigARTM, автор Дзен-канала «Цивилизационная идеология» 👉 https://dzen.ru/civideology
Константин — автор метода ARTM и open-source библиотеки BigARTM. Суть прорыва: классический байесовский подход (LDA) требовал переписывать всю математику заново под каждую новую задачу. ARTM заменил это конструктором — любые требования к модели добавляются как независимые слагаемые в одну формулу. Результат: BigARTM обогнал западные аналоги по скорости в разы и стал стандартом тематического моделирования.
📅 21 февраля, 14:00 (МСК) 📍 Онлайн (Zoom)
🏫 В партнерстве с Физтех-союзом
🌍 IS AI THE NEXT ELECTRICITY, THE INTERNET — OR NUCLEAR ENERGY?
Every transformative technology gets compared to electricity. But what if AI is closer to nuclear power — immense potential, existential risks, and an urgent need for governance?
🎙 In the next episode of Schrödinger’s Office podcast, I’m sitting down with Konstantin Vorontsov — MIPT alumni (Faculty of Applied Mathematics, class of ‘94) and one of the founding figures of Russia’s machine learning school.
About the guest
Konstantin holds a Doctor of Sciences in Physics & Mathematics, is a Professor of the Russian Academy of Sciences, and teaches at both MIPT and MSU (Faculty of Computational Mathematics & Cybernetics). He heads a lab at MSU’s AI Institute and chairs the “Machine Learning & Digital Humanities” department at MIPT.
He is the creator of the ARTM method and the open-source library BigARTM — a breakthrough in topic modeling. Where the classical Bayesian approach (LDA) required rebuilding the entire mathematical framework for each new task, ARTM introduced a modular design: any model requirement can be added as an independent term in a single formula. The result outperformed Western analogs in speed by several times and became the standard for topic modeling.
Author of the Dzen channel “Civilizational Ideology” — exploring AI’s role in civilizational development far beyond computer science.
We'll discuss how Russia's ML school emerged in the early '90s, the academic rivalry between MIPT and MSU, and who leads the field today. We'll dig into whether AI coding is killing the junior developer pipeline — and if programming loses value, what gains it. We'll talk about BigARTM and where topic modeling remains irreplaceable even after the LLM revolution, what digital humanities really means, which models could become the next big thing after LLMs, whether ML can generate genuinely new knowledge, and if the world needs an international AI governance commission.
📅 February 21, 2:00 PM (UTC+3) 📍 Online (Zoom) 🗣️ Language: Podcast in Russian