I am a Research Affiliate at the MIT Computer Science and Artificial Intelligence Laboratory. I also consult for an Internet startup (Permabit) that I helped start.
My first love is the physics of information and computation, and the informational modeling of physics. This was the subject of my PhD thesis (MIT 1987, 12MB PDF) and has been the focus of most of my subsequent research. I was privileged to work with Edward Fredkin, Tom Toffoli and Charles Bennett at the MIT Information Mechanics Group between 1980 and 1995, first as a PhD student, and then as a Research Scientist.
At MIT I was also the Principal Investigator and leader of the CAM-8 CA machine project, which designed and built a fine-grained, spatially-organized mesh-architecture multiprocessor. The point of this project was to advance the state of Cellular Automata modeling of physics, by making available hardware that can take advantage of the inherent efficiency of simple, uniform, spatially organized computation. The CAM-8 project was intended to provide a tool for investigating the possibilities of the kind of fine-grained parallelism that is available in nature. It was successful in this, but the project ended at the prototype stage, before CAM-8 machines had been built on a scale sufficiently large to enable investigations into the "previously inaccessible band of the computational spectrum" that was our true target. My later SPACERAM design generalized CAM-8's architecture into an almost-ideally-efficient building-block for spatial SIMD computations, but this has not yet been built.
Machine building, and then Permabit, distracted me for several years from pursuing fundamental questions at the interface between physics and computation, but I am now back to that pursuit. My current research is closely related past work on the relation between classical energy and the maximum speed of quantum dynamics. The minimum possible kinetic energy of a mechanical system is proportional to the rate of distinct state change, and so a realistic classical system should only have a finite rate of state change. Thus traditional infinite-state classical mechanics is not energetically realistic, but finite-state classical mechanics (e.g., lattice gas dynamics) is. In fact, just as classical computations can be recast as the simplest examples of quantum computation, finite-state classical mechanical systems can be recast as the simplest examples of quantum mechanics. From these simple examples, fundamental concepts and quantities of quantum mechanics can be seen as generalizations of classical concepts and quantities.
Quantum Computation
(from 1986, in New Techniques and Ideas in Quantum Measurement
Theory, edited by Daniel Greenberger).
Parallel quantum
computation (from 1990, in Complexity, Entropy, and the Physics
of Information, edited by Wojciech Zurek).
A Bridge of
Bits (from 1993, in Proceedings of the Workshop on Physics and
Computation, edited by Doug Matzke).
The
maximum speed of dynamical evolution (from 1998, in Physica D,
page 188).
Crystalline Computation
(from 1999, in Feynman and Computation, edited by Anthony Hey).
Universal cellular
automata based on the collisions of soft spheres (from a
conference in 1999, appears in Collision Based Computation,
edited by Andrew Adamatzky, page 107, (2002), and in New
Constructions in Cellular Automata, edited by David Griffeath and
Cristopher Moore. page 231, (2003)).
Looking at Nature as a Computer
(from a workshop in 2001, appears in International Journal of
Theoretical Physics 42:2, page 309, (2003)).
Mechanical Systems that are
both Classical and Quantum (based on a talk given at the
Unconventional Computation Workshop, Santa Fe, March 22 2007). Roger
Critchlow turned one of the examples in this paper into a cute quantum/classical clock
demonstration in which the exact continuous motion of the clock's
hands is displayed
as a continuously evolving superposition of integer-time states.
First moment distinguishability bounds (2011, extends work on maximum speed of dynamics).
Quantum emulation of classical dynamics (2011, provides an isomorphism between classical finite state dynamics and quantum finite-energy dynamics).
Cellular-automata supercomputers for
fluid dynamics modeling (from 1986, in Physical Review Letters,
page 1694).
CAM-8: a computer
architecture based on cellular automata
(from 1993, in Pattern Formation and Lattice-Gas Automata,
edited by A. Lawniczak and R. Kapral).
An FPGA architecture
for DRAM-based systolic computations (from 1997, in Proceedings of
the IEEE Workshop on FPGAs for Custom Computing Machines, edited by
Arnold et. al., page 2).
An Embedded DRAM Architecture for
Large-Scale Spatial-Lattice Computations (from 2000, in The 27th
Annual International Symposium on Computer Architecture, page 149).
Mechanism for
efficient data access and communication in parallel computations on an
emulated spatial lattice (United States Patent 6,205,533
applied for in 1999, issued in 2001).
Emulating
Physics: Cellular Automata that exhibit finite-state, locality,
invertibility and conservation laws
(a talk given at the Computing Beyond Silicon Summer
School, CalTech, June 24, 2002).
Physical
Worlds: Cellular Automata with computation universality at small and large scales
(a talk given at the Computing Beyond Silicon Summer
School, CalTech, June 25, 2002).
Spatial
Computers: Architectures and algorithms for large-scale spatial computations
(a talk given at the Computing Beyond Silicon Summer
School, CalTech, June 26, 2002).
Nature as
Computer / Computer as Physics: Physical concepts enter Computer
Science and computer concepts enter Physics
(a talk given at the Computing Beyond Silicon Summer
School, CalTech, June 27, 2002).
The models depicted in these movies are discussed above in the paper "Crystalline Computation" and in the lecture "Emulating Physics." All simulations were performed on CAM-8.
Diffusion and
sound waves in a reversible lattice gas (10MB): the four direction
TM lattice gas is started with a 50% density of particles, except for
an empty region (black) in the center. Half of the particles are
colored blue and half yellow, so that both diffusion and waves are
visible at the same time. The lattice is 512x512.
Lattice gas fluid
flow (5MB): a simulation of a six direction lattice gas fluid
flowing past a half cylinder, exhibiting vortex shedding. Visualized
by also simulating a "smoke" fluid within the CA. System is 2Kx1K.
Long range
forces in a lattice gas(3.5MB): a simulation of a six direction
lattice gas fluid with long-range forces. Force particles act at
three discrete distances to produce clumps that form an elastic
crystal. The model is discussed in A lattice
gas with long range interactions coupled to a heat bath (Yepez,
1993).
A reversible model of
crystal growth (8MB): when a grey gas particle diffuses next to a
green crystal particle, it joins the crystal and emits a red heat
particle. The reverse also happens. The model is discussed in A thermodynamically
reversible generalization of diffusion limited aggregation (D'Souza and
Margolus, 1998).
 from Superman 144, April 1961 |