Artem Kirsanov

Yes. Artem Kirsanov's YouTube channel explicitly describes him as a Harvard neuroscience researcher, and his content reflects deep familiarity with peer-reviewed computational neuroscience and mathematical modeling. His channel sits at the intersection of neuroscience, computer science, and mathematics — a profile consistent with active graduate-level or postdoctoral research.

Backpropagation requires error signals to travel backwards through a network in a way that has no clear biological mechanism in real neurons — a problem researchers call the weight transport problem. Artem's channel argues that the brain likely relies on something closer to predictive coding or local Hebbian-style learning rules, which better match what we know about synaptic plasticity. The video challenges the assumption that deep learning and brain learning are doing the same thing.

Predictive coding is a neuroscience theory proposing that the brain continuously generates predictions about incoming sensory data and only propagates the difference — the prediction error — to higher brain regions. It's a recurring theme on Artem's channel because it offers a biologically plausible alternative to backpropagation and is a leading candidate for a unified theory of how the brain learns and perceives. His hashtags and video titles consistently return to it as a central framework.

The 2024 Nobel Prize in Physics was awarded to John Hopfield and Geoffrey Hinton for foundational work on Hopfield networks and Boltzmann machines — generative models that store and retrieve patterns in ways loosely inspired by physics, and that laid the conceptual groundwork for modern deep learning. Artem covered this award through the lens of his channel's core focus: how neural systems, biological or artificial, learn to represent the world.

The framing refers to cortical columns — repeating, functionally self-contained processing units in the neocortex, first described by Vernon Mountcastle and later expanded by researchers like Jeff Hawkins. The argument is that rather than being one unified system, the neocortex is composed of many nearly-identical modular units each running similar computations. Artem uses this to challenge the top-down, monolithic picture of brain function still common in introductory neuroscience.

The video most likely covers Karl Friston's Free Energy Principle, the most mathematically ambitious attempt to unify perception, action, learning, and attention under a single framework. The core claim is that the brain minimizes surprise about sensory inputs by constantly refining its internal generative model of the world. Artem's channel has built a following in the computational neuroscience community specifically for making theories like this accessible without sacrificing rigor.

The video argues that the standard least-squares framing taught in introductory statistics and machine learning courses obscures what curve fitting is really doing — namely, performing maximum likelihood estimation under an implicit probabilistic model. Artem's signature style is to take a concept students already think they understand and reveal the richer mathematical structure underneath it. The video is aimed at people who learned curve fitting procedurally but never saw the Bayesian or probabilistic interpretation.

Artem Kirsanov has worked with Brilliant, the interactive STEM learning platform, as well as Shortform, a nonfiction book-summary service, and Squarespace. These sponsors align closely with his audience of analytically-minded students, researchers, and engineers who are actively looking to deepen their technical education.

Artem Kirsanov is based in the United States, which is also where the largest portion of his YouTube audience is located. His academic connection to Harvard places him in the Boston and Cambridge, Massachusetts area.

Artem Kirsanov's YouTube channel has grown to well over 350,000 subscribers — a substantial audience for a channel covering graduate-level neuroscience and mathematical theory. More notably, his engagement rate runs far above the category average for educational channels, suggesting his viewers are deeply invested rather than casual browsers.