Unsustainability of AI/ML and Computing Trajectories: Reframing Computing using Lessons from Nature and Principles from Physics

Energy IRC Seminar (in person event)

Speaker: Professor Sadasivan (Sadas) Shankar, Stanford Material Science & Engineering and SLAC National Laboratory, California, USA

All welcome to attend. Coffee available before and during event.

Abstract

Based on the exponentially increasing energy demands for computing and its applications and the ubiquitous use of digitalization, the trajectory of computing appears unsustainable both from energy and materials perspectives. This problem is further compounded by the slowing of Moore’s law and the increasing use of Artificial Intelligence Methods in all aspects of the modern economy in the 21st century. The energy needed per simulation can be several tens of orders of magnitude or higher than the thermodynamic energy limit at room temperatures. This disparity is due to several factors. Some of these factors include slowing down of the traditional geometrical scaling semiconductor devices (Moore’s law), widespread use of smart devices which are connected to internet, wider adoption of use of data centers for processing information at lower cost to consumers, and limitations of traditional digital computer architectures in processing large amounts of data. In addition, demands on data and accuracy can be computationally intensive requiring large numbers of higher precision operations. In combination, these effects, further accelerated by the digitization of the economy and ubiquitous availability of data analytic tools, computing devices, and increasing use of sensors, have all together led to an increase in energy requirements across all layers of computing systems. 

If so, what can be done to address this insatiable appetite for computing, which in turn needs increasing amounts of energy to operate?  

The current era of computing is mainly driven by mostly general-purpose computing architectures (e.g., von Neumann architectures) and deep neural networks (DNN) for wider applications for Machine Intelligence in driverless cars (Level 3 & 4), natural language processing (ChatGPT), and scientific applications (High Performance Computing). However, these are limiting in many cases given the requirements of large sets of data for every application and as illustrated in previous work were specific guiding principles for use of AI/ML methods in chemistry and materials. We will address these inefficiencies and whether even more advanced architectures like nature-inspired and quantum computing can be solutions to address all applications. We are currently developing a logical framework for classifying all information processing systems, which is expected to help in formulating a theoretical basis for designing specialized computing as needed by the application, a potential pathway to a Cambrian era in computing.

Biography

Sadasivan (Sadas) Shankar is Research Technology Manager at SLAC National Laboratory and Adjunct Professor in Stanford Materials Science and Engineering. He was the first Margaret and Will Hearst Visiting Lecturer in Harvard University and the first Distinguished Scientist in Residence at the Harvard Institute of Applied Computational Sciences. He has co-instructed classes related to materials, computing, and sustainability and was awarded Harvard University Teaching Excellence Award. He is involved in research in materials, chemistry, and specialized AI methods for complex problems in physical and natural sciences, new frameworks for studying computing, and a new course on Translation: From Invention to Innovation. He is a co-founder and the Chief Scientist in Material Alchemy, a “last mile” translational and independent venture for sustainable design of materials.

He was a Senior Fellow in UCLA-IPAM during a program on Machine Learning and Many-body Physics, invited speaker in The Camille and Henry Dreyfus Foundation on application of Machine Learning for chemistry and materials, Carnegie Science Foundation panelist for Brain and Computing, National Academies speaker on Revolutions in Manufacturing through Mathematics, invited to White House event for Materials Genome, Visiting Lecturer in Kavli Institute of Theoretical Physics in UCSB, and the first Intel Distinguished Lecturer in Caltech and MIT. He has given several colloquia and lectures in universities all over the world, and also worked in the semiconductor industry in the areas of materials, reliability, processing, manufacturing, and is a co-inventor in over twenty patent filings. His work was also featured in the journal Science, and as a TED talk.