Alexander Mordvintsev is holding a research scientist position at Google. His current work is focused on Artificial Life and principles of Self-Organizing Systems Design. The very first, and probably the most advanced skill that every living creature masters from the moment of inception is the ability to build and maintain its own body. This happens through collective behavior of countless tiny locally integrating agents pursuing their own goals. Alexander is looking to apply lessons from Differentiable Programming (aka Deep Learning) to create systems based on these principles that are able to act according to a provided specification or objective. This effort was triggered by the presentation “What Bodies Think About” that Prof. Michael Levin gave at NeurIPS 2018.
Previously Alexander worked on understanding deep neural networks by inspecting the computational circuits and their dynamics emerging during training. This work was started after joining Google in 2014 when he got introduced to the modern generation of differentiable machine learning models. DeepDream was probably the most known artifact produced by this line of his research that flooded the internet with psychedelic dog-slug images in summer 2015.
Before joining Google Alexander worked in St.Petersburg, Russia on various 3D computer vision and simulation applications. He studied computer science in St.Petersburg State University of Information Technology, Mechanics and Optics. During the late university years he participated in Google Summer of Code program twice, working on Python integration for OpenCV computer vision library.
As a part of his research, Alexander created a number of artistic projects exploring the themes of self-organization and the beauty of inner mechanics of artificial neural systems.
Featured works
(2023) SwissGL minimalistic web graphics library
Self-Organising Systems
(2023) Isotropic Neural Cellular Automata
(2022) Particle Lenia and the energy-based formulation
(2021) Self-Organising Textures (tweet)
(2020) Self-classifying MNIST Digits (tweet)
(2020) Growing Neural Cellular Automata (tweet)
DeepDream, Neural Network Visualization and Interpretability:
(2018) The Building Blocks of Interpretability
(2018) Differentiable Image Parameterizations
(2017) Feature Visualization
(2015) DeepDream (code)
Featured Art
Hexells (SIGGRAPH’2021, Leicester AI Art festival)
deepdream.c
Neverendeing story music video (with Perforated Cerebral Party)
Featured mentions
(2020) DeepDream: How Alexander Mordvintsev Excavated the Computer’s Hidden Layers by Arthur I. Miller (MIT Press)
(2016) How computers are learning to be creative by Blaise Agüera y Arcas (TED talk)
(2015) Inside Deep Dreams: How Google Made Its Computers Go Crazy by Steven Levy (Wired)
