Untangling Feynman's puzzle: Examining the future of computing

Sadasivan Shankar
(MSM Participant)
Institution/ Affiliation
Sadasivan Shankar - SLAC, Stanford and Harvard University
Presentation Details (date, conference, etc.)

Wednesday, March 9, 2022

3:40pm - 4:40pm PT

4:40pm - 5:40pm MT

6:40pm - 7:40pm ET

University of Colorado Boulder, Physics Colloquium

Host: Sean Shaheen  

Abstract: Have you wondered as to how nature computes reality? For example, a single grammole of water can have of the order of 1023 molecules. How does each of the molecules compute its own trajectory so that the motion leads to fluid flow? As we go from atoms to materials, we could think of moving from a periodic table of hundred or so usable elements to simple molecules like carbon dioxide, methane to complex macromolecules like polymers or more complex materials (Alvisatos et.al, 1999). Extending this type of thinking to biology, complexity increases in many different ways. A human brain has about eighty billion neurons with about 5000 to 10,000 to synapses per neuron resulting in possibly hundreds of trillion connections. With the advent of machine learning and deep neural networks, it appears that we can simulate the complexity of applications by artificial intelligence-based methods. However, even for training these networks, you need copious amount of data.

Meanwhile, the computing revolution known by the moniker Moore’s law, rolling past its fifty-year march as one of the most significant advancements of human civilization has been enabled by a confluence of breakthroughs in science and engineering (Moore, 1965). We have had at least about twenty-five doublings of computing ability, with the floating-point instructions per second with 20 doublings, which has enabled simulations of realistic problems. We are on the verge of exa-scale computing, which can compute about 1018 floating point operations per second. No matter we look at this, the computing capabilities and energy needed to simulate natural systems seem exceedingly high even for a small phase space of these systems, given the state of our current knowledge.

To address these questions, building on older concepts from Turing and von Neumann, and using a new framework, we will help bridge physics, chemistry, and biology with computing. Using these concepts we will get closer to addressing Feynman’s puzzle on computing the physics of a “stinky tiny piece of space-time”.