Edge-AI: Rise of the Neural Accelerators
Peter Corcoran, Professor at National University of Ireland
We are at the cusp of a new era in AI. Recently deep learning has revolutionised the field of Computer Vision, disrupting more traditional algorithmic approaches to object recognition, scene analysis and image understanding. But to date most of the applications of AI have relied on cloud-based services doing the heavy-lifting. Today we are at the cusp of a new era where AI data analysis is ready to move from the cloud and transition into the end device itself. In this talk we will review some of the Neural accelerators that are available today and look 1-2 years into the future to explore some of the potential impacts of edge-AI. We will also look at some of the new challenges and opportunities as AI migrates into individual devices: data-processing in the cloud won't go away, but more interestingly new computational & storage demands will emerge to support training of edge-AI. Among these will be the rise of 'Fake Data' and the VR tools and associated infrastructure required to scale the training of advanced Edge-AI solutions.
Peter Corcoran is Professor and Vice-Dean of Research & Graduate Studies in the College of Engineering & Informatics at NUI Galway. Prof. Corcoran is an IEEE Fellow with more than 450 technical publications, 85+ peer reviewed journal papers, 120+ International peer reviewed conference papers; co-inventor on 300+ granted US patents, 80+ granted European, with another 30+ patents currently pending. He has numerous accomplishments including former Editor-in-Chief of IEEE Consumer Electronics Magazine until Dec 2016, completed 20+ funded research projects and managed 3M in research funding over the last 10 years. His research interests include (i) smart-imaging/advanced digital imaging solutions; (ii) biometrics for handheld & IoT devices; (iii) AI, deep learning, data augmentation & data generation (GANs); (iv) fog & mobile edge computing; (v) internet tech & IoT (vi) Embedded Edge-AI.