Text compression as proxy for AI

A very interesting challenge:

http://cs.fit.edu/~mmahoney/compression/rationale.html

The goal is to compress Wikipedia losslessly. Intuitively, some
semantics aware compressor would do really well here. The problem is
that no one seems to know how to build one. The best entries so far are
all string-based (e.g., http://www.compression.ru/ds/).

Comments are closed.