Research:
Over the last four decades Prof. Crutchfield has worked in the areas of
nonlinear dynamics, solidstate physics, astrophysics, fluid mechanics,
critical phenomena and phase transitions, chaos, and pattern formation.
His current research interests center on computational mechanics, the
physics of complexity, statistical inference for nonlinear processes,
genetic algorithms, evolutionary theory, machine learning, distributed
intelligence, animal behavior, and quantum computation. He has published
over 250 papers in
these areas; many are available from his website:
csc.ucdavis.edu/~chaos.

Computational Mechanics

Dynamics of Learning

Evolving Cellular Automata

Evolutionary Dynamics 
The unifying theme of my research is
patterns—what they are,
how nature produces them, and how we discover new ones. The origins
of this interest date back to the 1970s, when the advent of powerful
and interactive computers stimulated much work on nonlinear
dynamics—deterministic chaos and bifurcations between distinct
behaviors. This early work raised a number of questions on how the
properties of nonlinear systems bear on the foundations of statistical
mechanics, including the existence of nonequilibrium states and how
one distinguishes molecular chaos—required to derive macroscopic
properties from microscopic dynamics—from the mechanisms of
deterministic chaos.
Progress during the 1980s in analyzing increasingly more complex
nonlinear systems eventually showed that these foundational questions
were special cases of broader issues: How is it that nature spontaneously
generates macroscopic order and structure? What mechanisms support
the production of structure? How does nature balance randomness and
order as structure emerges? And, perhaps most important of all, what
do we mean by structure, pattern, order, and regularity? Can there
be a theory that allows us to measure patterns as concretely and
workably as we measure randomness using thermodynamic entropy and
temperature?
This focus on patterns led to an even more central question, How
do we (or any agent moving through the natural world) discover
patterns in the first place? I call this pattern discovery to
distinguish it from pattern recognition—familiar in engineering,
where one designs systems with a builtin palette of templates, and
familiar in the natural sciences, where one analyzes data in terms
of an hypothesized representation, such as with Fourier transforms.
In these cases, a pattern is recognized when data most closely
matches one of the stored templates. Pattern recognition, however,
begs the question of discovery, Where do these representations come
from in the first place?
Answering these questions led me to develop a generalization of
statistical mechanics that explicitly defines structure and connects
structure in natural systems to how they store and process information.
In short, one asks, How does nature compute? The theory—unsurprisingly
called computational mechanics—attempts to answer three quantitative
questions (i) how much historical information does a system store,
(ii) where is that information stored, and (iii) how is it processed
to produce future behavior? These computational properties complement
the questions we typically ask in physics: How much energy is stored,
in what form is it stored, and how is it transformed over time and space?
In its approach to patterns computational mechanics uses the basic
paradigm of statistical mechanics to synthesize nonlinear dynamics
with information and computation theories. Over the last decade
it has been used in a number domains, some well outside physics—in
learning theory, evolutionary biology, and neuroscience, for example. My current
research focuses on applying computational mechanics to structure in
disordered materials, distributed coordination in collectives of
intelligent agents, prebiotic evolution, quantum computation, and
the dynamics of learning itself.

Synergistic Activities:

Scientific Director of a major NSFfunded science museum exhibit series
on pattern formation and complex systems—Turbulent Landscapes: The Forces
that Shape Our World at the San Francisco Exploratorium, JulyDecember 1996. The show toured the
nation's science museums since that time and, most recently, was on
display at the British Museum in London.

Director of SFI's Intelsponsored Program
on Network Dynamics.

Director of the SFI's Dynamics of
Learning Group and Computation, Dynamics, and Learning Program.

With composer David Dunn, constructing the Theater of Pattern
Formation, a largescale multichannel videoaudio exploration of structure and
emergence in the spatial and acoustic domains. A work in progress, it has been performed
in a number of venues including the 2006 California Institute of the Arts
Center for
Experiments in Art, Information and Technology festival and Burning Man.
Target venues are sensoryimmersive interative environments and theaters.

Cofounder, Vice President, and Scientific Director of the
Art and Science Laboratory, Santa Fe,
New Mexico, a nonprofit research center that supports
collaborations between artists and scientists working in the computing arts.

Referee and Editorial Board member for journals in theoretical physics,
mathematical biology, computer science, nonlinear mathematics, engineering,
and complex systems.

Member of the Information Technology and
Creativity committee of the National Academies' Computer
Science and Technology Board.
