Put AI to Work
April 15-18, 2019
New York, NY

Toward self-aware, resilient systems and ethical artificial intelligence

Pradip Bose (IBM T. J. Watson Research Center)
4:05pm4:45pm Wednesday, April 17, 2019
Implementing AI
Location: Rendezvous
Secondary topics:  AI case studies, Deep Learning and Machine Learning tools, Reliability and Safety

Who is this presentation for?

  • Researchers, practitioners, and executives engaged in defining the future landscape of AI system products and associated software algorithms

Level

Intermediate

Prerequisite knowledge

  • A basic understanding of machine learning, deep learning, and software-hardware system architectures

What you'll learn

  • Understand the modes and mechanisms of failures in today's AI systems and solutions
  • Learn how to make future AI systems more robust and resilient under changing in-field data patterns
  • Explore swarm AI and discover how this technology helps solve cognitive resilience challenges
  • See a simulation-based implementation demonstration of the advanced research being pursued in this project

Description

Pradip Bose details a next-generation AI research project focused on creating “self-aware” AI systems that have built-in autonomic detection and mitigation facilities to avoid faulty or undesirable behavior in the field—in particular, cognitive bias and inaccurate decisions that are perceived as being unethical.

The particular challenge addressed in this ongoing research endeavor is that of cognitive resilience and ethical, bias-free choices. Operating under changing in-field data patterns is difficult in current AI inference engines. Precipitous loss of in-field inferential accuracy can lead to potentially catastrophic errors in decisions. Also, inadequate safeguards in factory-trained model development can lead to decisions at the edge that are perceived as having unintended cognitive bias. But how do you ensure that the cognitive decision making at the edge is bias free, ethical, and resilient to in-field data changes or inconsistencies?

The research is centered around a visionary system architecture, where the cloud provides continuous, scalable support for a connected swarm of edge compute devices operating in the field of analytics and inferential decision making at the edge. Use cases include autonomous (self-driving) vehicles, network cybersecurity through a swarm of intelligent monitoring agents, and distributed, autonomous resource and power management in a data center or server setting. In each of these use cases, the project relies on a novel swarm AI mechanism to facilitate continuous in-field learning. Including a simulated “self-awareness” technique within each intelligent edge device allows interesting new modes of collaborative perception that are not possible with today’s smart robotics technology alone.

Pradip shares simulation- and modeling-based results to show the promise of using self-aware swarm AI technology for advanced cooperative compute and infer tasks in a cloud-backed mobile cognition setting. After establishing the benefit of self-aware swarm AI technology, he discusses the ongoing multiyear R&D activity in implementing it for prototype demonstration in a real-life setting for the sponsoring government client.

Photo of Pradip Bose

Pradip Bose

IBM T. J. Watson Research Center

Pradip Bose is a distinguished research staff member and manager of the Efficient and Resilient Systems Department at the IBM T. J. Watson Research Center as well as an adjunct professor at Columbia University. Pradip has been involved in the design and presilicon modeling of virtually all IBM POWER-series microprocessors, since the pioneering POWER1 (RS/6000) machine, which started as the Cheetah (and subsequently America) superscalar RISC project at IBM Research. Previously, he was the lead performance engineer for POWER3, a high-end processor development project, at IBM Austin and served as a visiting associate professor at the Indian Statistical Institute, where he worked on practical applications of knowledge-based (AI) systems. His current research interests are in high-performance computers, artificial intelligence, power- and reliability-aware microprocessor architectures, accelerator architectures, presilicon modeling, and validation. Pradip is the author or coauthor of over 100 publications (including several book chapters). He’s an IEEE fellow and a member of the IBM Academy of Technology and was the editor-in-chief of IEEE Micro from 2003 to 2006 and the chair of ACM SIGMICRO from 2011 to 2017. He’s received 25 Invention Plateau Awards and several Research Accomplishment and Outstanding Innovation Awards from IBM.

Leave a Comment or Question

Help us make this conference the best it can be for you. Have questions you'd like this speaker to address? Suggestions for issues that deserve extra attention? Feedback that you'd like to share with the speaker and other attendees?

Join the conversation here (requires login)