Statement by Eric Horvitz

June 2005

 

We are coming up on the 50th anniversary of a meeting organized by John McCarthy, Marvin Minsky, Nathaniel Rochester, and Claude Shannon to discuss the challenges of using computing devices to replicate “higher functions of the human brain.” The meeting helped to weave together several threads of activity into a young field that came to be called Artificial Intelligence.

 

I am honored to have been nominated to help lead our community forward at this milestone in our history.  Exciting times lay ahead.  There is a palpable acceleration in the generation of core results across our subdisciplines, and interest has been growing in the work that we do, among scholars across a spectrum of disciplines, as well as the public at large. 

 

I will work to ensure that AI continues to extend its growing reputation as a community of researchers known for creativity, high-quality scholarship, leadership, and relevance across disciplines—and as a community of scientists that can be assumed to understand, extend, and to create the best modeling methods for tackling the grand challenge of understanding intelligence. This effort will include strengthening our outreach to related fields, such as the management and decision sciences, statistics, economics, theoretical computer science, control theory, and the behavioral sciences. 

 

The founding leaders called for a multidisciplinary approach, expressing interest in concepts and methods from logic, statistics, cybernetics, information theory, and pattern recognition to address difficult challenges.  However, as AI set out to define itself as a new and fresh endeavor, distinct from its predecessors, the call for building upon teachings from related disciplines was not always heeded.  At times, this led to the perception that AI researchers were overlooking tools and innovations from related fields.  This perception has evaporated over the last 15 years, as the community vigorously embraced the interdisciplinary thrust and openness recommended by its early leaders. 

 

While the AI community continues to be known for fresh perspectives and innovations, today, it is easy to cite numerous cases where the leading expertise in methods that were originally introduced in other communities now sits squarely within AI.  For example, AI is now home to some of the world’s brightest statisticians who have developed an impressive array of tools for inducing predictive and explanatory models from observations.  The best new work on Markov decision processes, a methodology developed within the decision sciences, occurs within AI.  Decision analysts have been accessing the AI literature to track the latest innovations with influence diagrams and representations of preferences.

 

Beyond engaging disciplines with a center of mass outside of AI, we need to promote a greater sharing of ideas and results among the strong subdisciplines within our community.  It is truly a great sign of the vibrance of AI research that strong subcommunities have evolved in such areas as knowledge representation, uncertainty and decision making, machine learning, computational neuroscience, planning, economic methods and distributed agents, vision, natural language and message understanding, speech recognition, and hardness and complexity.  Nevertheless, much can be done to bring the family of descendant communities closer together so as to share advances more quickly.

 

Have we made progress on the challenges articulated in 1955—those of developing “machine methods of forming abstractions from sensory and other data,” “manipulating words according to rules of reasoning and rules of conjecture,” “carrying out activities which may best be described as self-improvement,” and developing “a theory of the complexity” for various aspects of intelligence?  We most definitely have.  Our community has accrued a great body of expertise, results, and intuitions about modern formulations of those challenges, and about many other challenges and subchallenges that have come into focus with research.  On the horizon, I see new scientific results and insights and an array of impressive applications that will no doubt have dramatic, positive influences on the world.  I look forward to working with you, and with colleagues in related fields, as we embark on the next 50 years of AI research.