I'm an ex-particle physicist from CERN specializing in heavy ion collisions. From 2013 to 2024, I conducted research while also being a science communicator - public speaker, blogger, and LHC guide. For several years, I was responsible for data selection and outreach coordination in the SHINE collaboration at CERN.
Three years ago, I became interested in community formation and the evolution of social norms. A year ago, I lost my research position at St. Petersburg State University (Russia) due to my anti-war and anti-government stance. As a result, I shifted my focus to connecting liberal anti-war communities in Russia while exploring large language models (LLMs) for fun—fascinated by their ability to grasp and generate knowledge.
Currently, I'm in the process of shifting my research interests from physics to collective AI systems.
to predict, explain, or discover phenomena of interest.
Explain and (not or) predict*. There is no point in a model that only describes experimental data. We say, "With enough parameters, you can fit any data in any model".
* Also, it has to predict something that can be measured. This is the reason why particle physicists do not consider string theories as something serious (more like a math exercise).
What do you think about the possibility of emerging collective behavior when AI agents will interact on the web in large numbers?
thank you, fixed.