Keynote: People, Machines and Intelligence
By Violet Chen
Back in the 1960s, Dr. J. C. R Licklider plotted a long-term dream of man-computer symbiosis, “people will set the goals, formulate the hypotheses, determine the criteria, and perform the evaluations. Computing machines will prepare the way for insights and decisions in technical and scientific thinking.” On Wednesday afternoon, in his keynote talk, Dr. Eric Horvitz, Technical Fellow at Microsoft and Director of Microsoft Research Labs, enlightened the audience with several current opportunities to continue realizing this dream. Since the beginning of his career as a post-doc consultant at NASA, he has been passionate about “harnessing computing advances to enhance the quality of people’s lives” and “leveraging the complementarities of human and machine reasoning.” He discussed four key opportunities at fostering a tighter coupling between human and machine.
The first one is to develop joint reasoning models of cognition and a state of the world. Proper integration of human knowledge and machine knowledge has great potentials in leading to better information inferences. Dr. Horvitz used a traffic forecasting service provided by Microsoft as an example of practical joint models. Taking advantage of integrated knowledge, the system can bring attention to anomaly situations, which may surprise even human experts.
Leveraging complementarity between machine perception and human perception is the second crucial opportunity. Increasing research efforts have been dedicated to building computing mechanisms (algorithms, optimization models) to aid real problem solving. Examples of expected improvement include reduction in human error and increase in cost efficiency. Another aspect of complementarity that Dr. Horvitz addressed is the need to ensure compatibility. With an example of potential incompatibility arising during updates of a healthcare decision support system, he stressed that compatibility does not come for free.
Next, Dr. Horvitz discussed open opportunities in the coordination of initiatives between human and machine. He shared a robotic surgical system, studied at Johns Hopkins University, as a practical use case of coordinating human actions and machine decisions. A key of integrative intelligence is to utilize decision theory to guide human-machine dialogue in an effective way. From several fun clips of human interaction with robot and avatar assistants at Microsoft, the audience were given a glimpse of the ongoing research efforts to construct a computer system capable of facilitating fluid human-centered conversations.
Finally, he discussed augmenting human cognition with machine intelligence. Dr. Horvitz proposed that leveraging and extending results from cognitive psychology can be highly impactful. With machine intelligence, we may develop approaches to fill gaps in human attention, memory, and judgement. Such machine-enhanced tools could lead us to much better solutions to persistent and important challenges in real-life problems.