Stanford Seminar - Computing with High-Dimensional Vectors

Explore the concept of computing with high-dimensional vectors in this Stanford University seminar presented by Pentti Kanerva. Delve into a computing approach that bridges the gap between symbolic AI and artificial neural networks, using 10,000-bit words as basic objects. Discover how this method encodes information across all components of a vector, allowing for superposition of data fields. Learn about the algebra of high-dimensional vectors, including operations analogous to addition and multiplication, as well as permutation of coordinates. Understand the advantages of high dimensionality in providing nearly orthogonal vectors and its applications in encoding various data structures. Examine the architecture's compatibility with emerging nanotechnology and its potential in machine learning. Witness a practical demonstration of high-dimensional computing through a language identification algorithm. Gain insights into the speaker's background and research journey in understanding brains in computing terms.