Santa Fe Institute Collaboration Platform

Thermodynamics of Computation

Michael Frank

From Thermodynamics of Computation
Revision as of 19:11, November 12, 2018 by Mpfrank (talk | contribs) (Inserted one more ref)

Biography: My first exposure to the thermodynamics of computation was in 1985 when I read a Scientific American article (1) about the subject during high school. Then in 1988, I took the world's first class on Nanotechnology, taught by the visionary K. Eric Drexler at Stanford University, in which Drexler discussed an architecture (2-4) for reversible nanomechanical computation which he was designing at MIT. I graduated from Stanford in 1991 with distinction with a B.Sci. in Symbolic Systems (focusing on CS and AI), and went on to do my graduate work in the EECS department at MIT. After completing a Masters thesis on decision-theoretic methods in AI (5), I decided that we really needed faster computers and so, with support from an NSF Graduate Fellowship, I began studying nanoscale computation in earnest. In 1995 I designed the world's first method for universal computation using DNA chemistry (6), this method was required to be reversible for fundamental thermochemical reasons. I then joined the DARPA-funded MIT Reversible Computing Project, led by the legendary hacker (7) Tom Knight, and cellular automata expert (8) Norm Margolus, who was an alumnus of the Information Mechanics group (9) led by Ed Fredkin, former director of the MIT Lab for Computer Science and inventor of the Billiard-Ball Model of computation (10), which was the first ballistic model of reversible computation. Knight and his student Saed Younis had invented the first complete, fully adiabatic and reversible CMOS-based logic technology (11), and in the subsequent project, I and other students built on their work to create the first adiabatic and fully-reversible universal processor chips (12,13). During this period I also showed that the aggregate performance of parallel 3D adiabatic reversible machines scales better physically (by polynomial factors), within power dissipation constraints, than any possible irreversible computing architecture (14-15). My 1999 MIT doctoral dissertation (16) derived several other related scaling results, and explored a number of other aspects of reversible computing, from physics through application algorithms. After graduating, I continued my research in faculty positions at the University of Florida (17) and Florida State University (18), and currently as a senior-level engineering scientist at Sandia National Laboratories (19). While at UF and FSU, I occasionally taught a research survey course on The Physical Limits of Computation (20). In 2005 I organized the first workshop on reversible computing (21) and I currently serve on the program committee for the ongoing Reversible Computation conference series (22). Recently, I have reviewed and clarified the foundations of Landauer's principle (23-24) and reversible computing theory (25-27), and I am also presently working on a couple of funded R&D projects at Sandia which are aiming to demonstrate new engineering implementations of reversible computing (28). Numerous of my other papers and talks in this field are available through my research webpages (17-19).

Field(s) of Research: Computer Science Engineering to Address Energy Costs, Computer Science Theory, Logically Reversible Computing

Related links