Friday, June 25, 2010
Why We Need Vector Symbolic Architectures
A couple months ago I gave a talk at Haskins about Vector Symbolic Architectures, a new approach to knowledge representation and computing that is very robust to even large amounts of noise. This blog post describes how a classical (von Neumann) computing architecture can fail from the random flipping of even a single bit. Whatever organizing principles the mind/brain is using, they can't be anything like this!