Report: Machine Learning Driving AI
Artificial intelligence research continues to accelerate as human and machines collaborate to solve more complex problems. A new survey by the National Academy of Science identifies the frontiers of AI research that include “augmented cognition” along with “integrative” AI.
In a workshop report on IT innovation just released by the National Academy’s science, engineering and medicine branches, a section on “Developing Smart Machines” describes ongoing AI research efforts along with machine learning, a discipline some experts consider a subfield of AI. “Machine learning is what lets computers discover patterns within data and then use those patterns to make useful, and ideally correct, predictions,” the report states.
The authors quotes Jaime Carbonell, a professor of computer science at Carnegie Mellon University, as noting” “Machine learning essentially is the engine that is driving modern artificial intelligence.”
Machine learning “often deals with unbalanced data sets in which the ultimate focus of decision making is precisely the outlier cases,” that is, the extreme cases that contrast sharply with typical data and provide opportunities for the “most important learning opportunities.”
Ignoring the rare cases, or outliers, means “you cold miss everything [in a data set] that is interesting,” Carbonell asserted.
The report also attributes recent advances in machine learning to the rapid growth of big data, particularly data generated by networks of connected sensors. While big data has driven advances in machines learning, Carbonell noted the “flip side” is that machine learning is needed more than ever to cope with huge data volumes.
While much work has focused on areas of machine learning like natural language processing, other researchers emphasized emerging fields such as augmented cognition and integrative AI. Eric Horvitz, managing director of Microsoft Research (NASDAQ: MSFT), described augmented cognition as machine learning complementing human cognition in areas such as memory, attention or judgment.
Examples of human-machine collaboration in which a problem is divided between a human operator and a machine include new surgical approaches, the report notes.
Meanwhile, integrative AI could be used to create systems capable of dealing with real-world complexity. The approach ultimately “could be the key to transforming computers, which currently have deep but very narrow intelligence, into broader, more human-like thinking machines.”
The report also stressed that much of the commercial development of AI technologies ranging from Siri voice-recognition systems to grammar checking apps is based on several decades of government-sponsored research. U.S. research funded by the Defense Advanced Research Projects Agency and other federal agencies provided the foundation for today’s facial recognition technology, Horvitz noted.
Meanwhile, machine learning experts stressed in the report that new approaches would seek to design machine capable of asking questions or identifying missing data, then actively seeking answers or data needed to fill in the gaps. Researchers stressed this is a continuous cycle in which the human operator allows the machine to refine its knowledge and understand nuance.
In the case of an application like machine translation, a computer would identify a missing piece of information and ask the operator to supply it. This “active learning” approach would allow the machine to incorporate new data into its model to improve translations over time.