The Allen Institute for Artificial Intelligence in Seattle is making an ambitious bid to give AI common sense, a major factor in taking the technology beyond its current limitations.
Now the institute has hired a new leader for the project: University of Washington Professor Yejin Choi. Choi will take the helm of Project Alexandria, the “common sense AI” initiative backed by $125 million from Paul Allen, the Microsoft co-founder and founder of the institute.
“Common sense is what makes the fundamental difference between human intelligence and machine intelligence today,” Choi told Geekwire. “Our research will help [enable] AI technologies that are significantly more intelligent and robust than what are practically possible today.”
Choi is a longtime AI researcher with an expertise in natural language processing, particularly the contextual and non-literal understanding of language that is so important to developing “common sense” for an artificial intelligence system. She will also maintain her position as an associate professor at the University of Washington Paul G. Allen School of Computer Science.
Her addition comes amid a fierce talent war over top AI researchers and engineers. Just in the past few months, Facebook poached a top AI2 researcher for its growing Seattle office. AI2, in turn, hired a key AI leader from Amazon’s Alexa division.
Choi’s addition gives Project Alexandria a shot in the arm as it gets off the ground. Her vision for the project is to bring about a fundamental shift in how artificial intelligence and machine learning systems work — not just in how each system functions but also how they relate to and teach each other.
“The existing machine learning paradigm tends to focus on learning models that are specialists on some narrowly defined tasks such as speech recognition or machine translation,” Choi said. “While we have seen dramatic advancements with deep learning models for some of these tasks, they do not yet have the capabilities to abstract away generalizable knowledge (about the world) that can be re-used for other tasks. Project Alexandria aims to create universal representations of commonsense that can be shared and used by other AI systems.”
The revolution Choi describes will be no easy task — but she’s come prepared with a game plan on how to get there.
“Over the next few years, Project Alexandria will introduce standard measurements for the common sense abilities of an AI system, develop novel crowdsourcing methods to acquire common sense knowledge from people at an unprecedented scale, and develop applications that utilize common sense to boost performance for a broad range of practical AI challenges, from machine reading and translation to robotics and vision,” Choi said.