Acquiring Big Mechanisms by Reading Primary Literature – Paul Cohen (Information Innovation Office, DARPA and the University of Arizona)
Abstract
The parable of the blind men and their elephant is so banal and shopworn that one hesitates to discuss it in an academic setting, so let’s instead consider blind scientists and the climate, or economies, or social movements, or energy security, or food production, or the signaling networks that regulate cell division and apoptosis. Modern science and academic promotion practices favor deep understanding of tiny fragments of such systems. Indeed, isolating a fragment of a system from its many influences is fundamental to classical experimental science and hypothesis testing. Experimental science often produces causal knowledge about fragments of systems, but causal knowledge encourages interventions, which can be perilous when the systemic effects of interventions are obscure. DARPA’s Big Mechanism program seeks to develop technology for machines to build and maintain causal models of complicated systems. Concretely, the program is developing technology for machines to read the primary literature on Ras-driven cancers, extract fragments of the relevant biology from individual papers, assemble these fragments into causal models of unprecedented scale and fidelity, and reason about the systemic effects of interventions. This talk will be a progress report (nine months in) and a synopsis of the technical challenges of what we call “push” research, where every research result is instantaneously pushed into multiple, machine-maintained, causal theories, or big mechanisms.
Biography
Paul Cohen is a researcher in Artificial Intelligence and Cognitive Science. His PhD is from Stanford University in Psychology and Computer Science. He was a professor of Computer Science at the University of Massachusetts, the director of the Center for Research on Unexpected Events at USC’s Information Sciences Institute, the head of Computer Science at the University of Arizona, and the founding director of the School of Information: Science, Technology and Arts, also at the UA. He has been an IPA to DARPA for 19 months, where he started the Big Mechanism and Communicating with Computers programs.