Bulatov Questions Linear Algebra Need in Learning Algorithms

Yaroslav Bulatov posted on X about his habit of questioning why learning algorithms rely on linear algebra and arithmetic. He referenced Stephen Wolfram's tracing of certain conventions to the 1943 McCulloch and Pitts paper. Wolfram attributed the choice to McCulloch's personal preference for linear algebra. Bulatov has held roles at Google Brain on TensorFlow, Meta on PyTorch, early OpenAI, and now works at Together AI on scaling infrastructure. The post presents the historical detail without further analysis of alternatives.