EE-MSE Special Seminar: Building the next room temperature ubiquitous computing platform

Building the next room temperature ubiquitous computing platform Bridge to Artificial General Intelligence
Friday, November 15, 2019 - 3:00pm
Packard 101
Sasikanth Manipatruni (Kepler Computing)
Abstract / Description: 

Computing is at a momentous point as devices approach the 10 nm dimensions even as new breakthrough AI/QC architectures evolve. In this talk, I will outline the framework that combines the energy/dimension scaling (Moore's law), computer error rates (Shannon computing) and modern AI architectures. I also outline how the next room temperature computing device can be made from quantum materials exploiting the correlated electron phenomena.

Beyond CMOS and Quantum Materials: I describe a quantum and memory materials-centric approach with to enable the compute device layer for beyond the CMOS era. I will outline a number of pathways for computing devices that utilize quantum materials. In particular, the path to computing at aJ/bit with 10-20X relaxation of the interconnect requirements will be described. I will generalize the search for next switch with a comprehensive list of quantum materials classes.

AI and memory: In the second half, I will describe the connection between AI computing and memory, identifying the key bottlenecks for reaching 1000 TOPS/W (a 100X leap over today's GPUs/ASICs). Quantum materials, especially next-generation ferro-electrics and magnetics, can mitigate this memory wall.

AGI and computing hardware: Finally, I describe potential ways to build AGI class hardware overcoming interconnect, memory and compute bottlenecks.



Sasikanth Manipatruni is the Chief Technology Officer at Kepler Computing. Prior to this he is the founding research director of Intel-FEINMAN center (Functional Electronics Integration and Manufacturing), to build the next room temperature transistor with quantum materials. He received PhD from Cornell working with Prof. Michal Lipson in silicon photonics where he demonstrated ultra-fast silicon electro-optic switches, opto-mechanical non-reciprocity and synchronization of opto- mechanical systems. At Intel, he developed materials & devices for beyond CMOS memory/logic and built 1st industrially adopted spintronic/quantum SPICE tool. He was awarded the US-National Academy of Engineering recognition for young engineers 2019, IEEE/ACM under 40 innovator award at DAC'17, Mahboob Khan outstanding liaison award '16, CSPIN outstanding industry liaison award '16, and serves on several industry panels for US-wide research selections. His work is cited ~4000 times & holds ~ 200 granted/applied patents in spin/photonics/MEMS/CAD/AI/QC. He coaches school students for USA-PHO physics Olympiads.