The Volumetric Society is thrilled to be hosting a special meeting with De Kai, a world renown expert in Artificial Intelligence. De Kai is Professor of Computer Science and Engineering at HKUST. He was named Founding ACL Fellow in 2011 by the Association for Computational Linguistics for his pioneering contributions to machine translation and inversion transduction grammars, which are machine learning foundations underlying modern translation technologies like Yahoo Translate, Google Translate or Microsoft Translate. Recruited as founding faculty of HKUST directly from UC Berkeley, where his PhD thesis was one of the first to construct probabilistic machines that learn to understand human languages, he co-founded HKUST's internationally funded Human Language Technology Center which launched the first AI web translator over 20 years ago. His AI research focuses on natural language processing, music technology, learning and cognition.
http://www.cs.ust.hk/~dekai/
https://www.youtube.com/watch?v=eIFBNBGSJnk
De Kai's talk at NYU Magnet in Brooklyn , "How AIs Are and Aren't Kids", discusses the fact that despite explosive progress, our deepest AIs today still can't do what a three year old kid can. Why not? The next big thing in AI is not a thing; it's relating things. Our creative linguistic and musical expression and improvisation abilities are built not in terms of the things we learn, but rather the translations we learn. The key to human level intelligence is our ability to learn the many ambiguous, complex, structural, contextual, creative relationships between things. Categories of things are important, surely, but they are nowhere near as important as categories of relationships between things. Instead of the traditional focus on monolingual representation languages, what matters is the bilingual relationships between the many different representation languages that occur both naturally and artificially. Translating, interpreting, relating, and reframing are central to the uniquely human strengths of music, language, and thought. The key challenges facing true AI lie in explaining how it is that we are able to efficiently and effortlessly learn such complex translations, just as human kids do. The answers lie in the tension between evolutionary selection pressures and computational complexity.
And yet even our weak AIs of today have already become integral, active, influential, learning members of our society. Just like other kids, our AIs are already learning culture from the environment we're raising them in, and the jobs we're giving them. Strong AI, artificial general intelligence, and conscious self-aware AIs may still be many years away, but learning machines have already crept deeply into our social fabric. Today's AIs might be big and dumb, but they're quickly reshaping the culture that each next generation of AIs will learn under. How should we deal with our new AI kids?
