industrial ai - web.stanford.edu

/SLAC

Industrial AIee392b Seminar Overview

Dimitry Gorinevsky

Consulting Professor in Electrical Engineering

www.stanford.edu/~gorin

March 31, 2021

/SLACStanford

Class Topic: Industrial AI Big Picture

• What do companies mean when they hire AI engineers?• Why are companies compelled to do AI?• What is the outcome they expect?• What are the real issues there?

March 2021 © D. Gorinevsky: www.stanford.edu/~gorin 2

/SLACStanford

DIGITAL TRANSFORMATION

March 2021 © D. Gorinevsky: www.stanford.edu/~gorin 3

/SLACStanford

Digital Transformation

• Digital Transformation• Software is eating the world• Digital technology transforms all areas of economy

– Internet of People (IoP)– Internet of Things (IoT)– Business Processes

March 2021 © D. Gorinevsky: www.stanford.edu/~gorin 4

/SLACStanford

Analytics Applications

• Much of software business is in data management platforms

• Added value comes from business process changes driven by applications

March 2021 © D. Gorinevsky: www.stanford.edu/~gorin 5

www.forbes.com/sites/cognitiveworld/2019/08/29/software-ate-the-world-now-ai-is-eating-software/?sh=7ffaefab5810

/SLACStanford

Analytics in Industry

• Business Intelligence (BI)– Spreadsheets

• Data Science– Exploratory analysis services

• AI – Automated applications

March 2021 © D. Gorinevsky: www.stanford.edu/~gorin 6In

tegr

atio

n Scaled to Big Data

Descriptive analytics

Predictive analytics

Prescriptive analytics

/SLACStanford

Industrial AI in Digital Transformation

March 2021 © D. Gorinevsky: www.stanford.edu/~gorin 7

Data from Things

Analytical Applications

Industrial AI

Data from People

/SLACStanford

Digital Transformation and Industrial Revolution

• Industry 4.0 – 4th Industrial Revolution – Manufacturing: Data from Machines and Processes

March 2021 © D. Gorinevsky: www.stanford.edu/~gorin 8

/SLACStanford

Digital Transformation with Data From Things

• Smart Cities, Smart Buildings, Smart Homes• Industrial IoT (includes Industry 4.0, Smart Cities, and more)

– Smart Grid – Supply Chains – Transportation (Cars, Planes, Trains)

• More… – Insurance

• Data Centers and IT Systems (AI Ops)

March 2021 © D. Gorinevsky: www.stanford.edu/~gorin 9

/SLACStanford

INDUSTRIAL AI APPLICATIONS

March 2021 © D. Gorinevsky: www.stanford.edu/~gorin 10

/SLACStanford

What is Special about Industrial AI?

• Mission Critical Analytics – Explainable AI – Digital Twins models – Verification and validation processes, engineering rigor

• E.g., Tesla, Waymo

• Automated Data-driven Apps– Scalable, reduced deployment effort– Broaden access to the benefits of advanced analytics

March 2021 © D. Gorinevsky: www.stanford.edu/~gorin 11

/SLACStanford

Mission Critical Analytics

March 2021 © D. Gorinevsky: www.stanford.edu/~gorin 12

WebAdvertising

0.3%

High-freqTrading

51%

Industrial AIDecision Support

95%

Machine Learning/AIData Science

Control SystemsRocket Science

99.9999%

CivilAviation

PowerSystems

99.97% generality, fast development | reliability, accuracy

CS ENG

IT OTIoP IoT

AI Ops

/SLACStanford

Computing Platform: IT/OT Convergence

March 2021 © D. Gorinevsky: www.stanford.edu/~gorin 13

↑↓

• IT: Enterprise computing. Cloud. Information Technology

Operational Technology

Edge

Device

Cloud IT

• OT: Industrial systems. Secure, closed networks.

Edge computing. Fog.

OT

/SLACStanford

INDUSTRIAL AI AT STANFORD

March 2021 © D. Gorinevsky: www.stanford.edu/~gorin 14

/SLACStanford

Analytics Domains

March 2021 © D. Gorinevsky: www.stanford.edu/~gorin 15

IT

Embedded

↑

↑

↑

Engineering MethodsModeling & Simulation

Rigorous Analysis

Operations Research

Industrial AIMission Critical, Explainable

Decision and Control

Signal Processing

Data-Driven MethodsStatistics and

Machine Learning

Disciplines Methodology

IIoT Persistent Data

OT (Operational Technology)

/SLACStanford

Industrial AI Initiative

Rigorous Algorithms IndustrialMission Critical ApplicationsData Driven Modeling AI At Scale Deployment

March 2021 © D. Gorinevsky: www.stanford.edu/~gorin 16

Industrial AIData from ‘Things’

Mission Critical Applications

At Scale Deployment

StanfordIndustrial

AIInitiative

StanfordRigorous Analysis and Provable Algorithms

Data-Driven Modeling and Optimization

https://iai.stanford.edu

/SLACStanford

IMPORTANT USE CASES

March 2021 © D. Gorinevsky: www.stanford.edu/~gorin 17

/SLACStanford

Digital Transformation of Operations and Support

• Engineering and manufacturing – 10-15% of the lifecycle cost

• Operations and supply chains– 65-80% of the lifecycle cost

March 2021 © D. Gorinevsky: www.stanford.edu/~gorin 18

DoD CBM+

/SLACStanford

AIOps

• AI for IT Operations • Computing center as an industrial plant• Control plane provides ‘OT’• Tech Industry

– Much more dynamic than traditional industries

March 2021 © D. Gorinevsky: www.stanford.edu/~gorin 19

/SLACStanford

SYLLABUS

March 2021 © D. Gorinevsky: www.stanford.edu/~gorin 20

/SLACStanford

Planned Lectures

March 31, Introductory LectureApril 7, Industrial Enterprise, HitachiApril 14, AIOps, AlibabaApril 21, VC, GreylockApril 28, Consulting, Accenture (panel)May 5, VC, The Robotics HubMay 12, AIOps, GoogleMay 19, AIOps, MicrosoftMay 26, AI Technology, AspenJune 2, Consulting, Deloitte (panel)

March 2021 © D. Gorinevsky: www.stanford.edu/~gorin 21

/SLACStanford

GreylockVC

I-AI Lecture Map

March 2021 © D. Gorinevsky: www.stanford.edu/~gorin 22

Industrial AI

AIOps

Technology Business Enterprise

Google

Alibaba

HitachiRobotics

HubVC

Microsoft

Deloitte

Aspen Tech

Accenture