Behind today's "big data" buzz lies one of the biggest challenges in artificial intelligence: how can machines automatically learn structural relationships?  Thought is driven by the internal languages you use to represent your perception and conception of the environment.  Chinese speakers conceive of situations differently than English speakers, for example.  Visual, musical, or technical languages that you are fluent in alter the way you think.  Even more strongly, in fact, the human capacity to learn from data is based on your ability to learn the relationships between various alternative languages that you use internally.  How do we learn new kind of languages?  How do we learn to relate these new representations to other previously familiar languages?  The answers form nothing less than the key to intelligence.

Both long-term and near-term applications of these modeling advances are virtually limitless.  Web translation systems, which we pioneered, are one highly visible real world example.  Multilingual data mining is another.  This year, we are pioneering new applications in electronic music.  Foundations for a new computational technology revolution are rapidly emerging from this area of research.