Video: https://youtu.be/ZS8rGSzb_9I
The context of this talk is this statement from the host institution's provost: "We are trying to mobilize our campus activities around AI.” I connect academic initiatives in Interdisciplinary AI with industry needs.
--- Original abstract -----
Every company now is an AI company: Now, Near Future, or Distant Future?
Amit Sheth, AI Institute, University of South Carolina
“Every company now is an AI company. The industrial companies are changing, the supply chain…every single sector, it’s not only tech.” said Steven Pagliuca, CEO of Bain Capital at the 2019 World Economic Forum. With this statement as the context, I will provide an overview of AI landscape -- what AI capabilities are for real, what is being oversold, what is nonexistent, what is unlikely in our lifetime. I will also provide an anecdote-supported review through a broad variety of current and eminent applications of AI that rely on some of the well-developed and emerging AI capabilities. The objective is to help those considering AI applications start thinking of new business opportunities, new products and services, and new revenue/business models in the context of rapid penetration of AI technologies everywhere. I will seek to answer: Is AI just hype or something already happening? If it has not happened in your industry, is it impending? Do bad impacts of AI outweigh the good?
Finely Chair talk: Every company is an AI company - and why Universities should train in interdisciplinary AI
1.
2. Every Company Is an AI Company:
Now, Near Future, or Distant Future?
Finely Chair Webinar - Nov. 11, 2021
Dr. Amit Sheth
Professor, Founding Director of AI Institute
University of South Carolina
amit@sc.edu
http://aiisc.ai
#AIISC
3. “We are trying to mobilize our campus activities around
AI”
- Chancellor Angle
● State of AI
● AI Applications; AI in Industry
● Campus-wide AI initiative at the UofSC -
why, what and how of the AI Institute #AIISC
5. “Information is cheap.
Understanding is expansive.”
Karl Fast, Professor of UX Design
Kent State University
AI is about converting data into
knowledge, insights and actions.
6. “Every company now is an AI company. The
industrial companies are changing, the
supply chain… Every single sector, it’s not
only tech. ”
Steven Pagliuca
CEO of Bain Capital, WEF2019
7. IBM CEO Krishna:
“Every company will be
an AI company. ”
https://www.zdnet.com/article/ibm-ceo-krishna-every-company-will-be-an-ai-
company/
8.
9. Rapid Change in Leadership: Industrials to Tech and AI
9
What you should know about AI
from Amy Webb’s ‘The Big Nine’;
The US G-MAFIA: Google,
Microsoft, Apple, Facebook, IBM
and Amazon & the China-BAT:
Baidu, Alibaba and Tencent ... own
most of the technology and patents
and can attract the best talent and
partnerships with universities that
teach AI/machine learning.
10. Rapid Growth in AI Investments
https://www.brookings.edu/techstream/what-investment-trends-reveal-about-the-global-ai-landscape/
13. “While US is ahead in AI research, China is
significantly ahead in AI development
and monetization.”
Kai-Fu Lee
CEO of Sinovation Ventures, Author of “AI Superpowers”
Former President of Google-China
16. AI SUBAREAS
KEY AI
SUBAREAS
Conversational
AI
Machine &
Deep
Learning
Natural
Language
Processing
(NLP)
Computer
Vision
Robotics
Knowledge
Graph
(Ontology)
Dr Harik's neXt LIVE with Dr. Amit Sheth on AI in Manufacturing
17. Revolutionary Role of AI - But Not in Isolation
When we talk about AI, it is not just computing or algorithms, or deep learning (it
is of course important)…. it is the ability to draw insights from broad variety of
data and other digital tools:
● Internet of Things/Sensors
● Biotechnology
● Behavioral Science/psychology - understanding of humans
● Digital Payment
Management needs to appreciate the need to put together multidisciplinary
teams!
18. AIISC
• First university-wide AI Inst in US SE, with the objective to be among
the top in AI in US SE and in AI applications in the nation.
• Core research on AI topics such as knowledge infused learning,
neuro-symbolic and brain-inspired (semantic-cognitive-perceptual)
computing, collaborative & conversational agents
• Translational research with nearly all colleges at UofSC
• More at: http://j.mp/AII0720 , http://aiisc.ai
Amit Sheth – Vision of Data Science @ Vaibhav, 8 Oct 2020
20. Scope of the university-wide AI Institute
Education: Started an AI certification
for our MS-CS degree. Engaging
high school and undergraduate
student (also diversity and inclusion).
Working on MS and PhD in
Interdisciplinary AI.
Spin of Companies/founder using AI
technology developed at Univ:
Taalee/Semagix; Cognovi Labs. Also
cofounded: ezDI.
21. Translational Research at AIISC with...
Pharmacy#
Public Health**
Neuro & Cog Sc**#
Manufacturing****
Education**#
Personalized Medicine****#
Science (e.g., Astrophysics) *##
Engg (E.g., radiation, civil infrastructure) *#
Nursing *##
Others: Law, Journalism, Finance *#
21
* = funded project, # = pending project [as of Nov 2021]
22. Automated Planning,
Smart Manufacturing &
Factory of Future
The global AI in manufacturing market size was USD 1.82 Billion in 2019 and is projected to
reach USD 9.89 Billion by 2027, exhibiting a CAGR of 24.2% during the forecast period.
[Fortune Business Insight]
23.
24. Current State of Material Planning
▰ BMW works with a highly complex supply chain, comprising thousands of material numbers
and hundreds of suppliers
▰ BMW’s material planners must juggle complex KPIs and an ever-shifting procurement
landscape to keep the line running at maximum efficiency
▰ BMW would like to reduce downtime due to missing or late parts, optimize its ordering
strategy, and shift material planning responsibilities to more critical needs
▰ Our proposed project uses AI and automation to aid BMW material planners and to improve
material planning processes and outcomes for BMW
25.
26.
27. Future Factories
▰ Smart Production and Logistic
System
▰ Smart Data and Cloud
Computing Infrastructure
▰ AI -based Innovative
Manufacturing
▰ Industry 4.0 Standards
28. Complex Manufacturing Event Understanding
▰ Perform processing and analytical tasks on the real-time
collected data to aid in real time decision making by extracting
actionable knowledge from raw inputs.
▰ Comprehensive domain knowledge such as data capture
capabilities and product specifications can be infused with real-
time inferred data for predictive monitoring measurements.
29. Future Factories Digital Twin
▰ Enables communication with the
industrial assets at the factory.
○ Open Platform Communications
(OPC)
▰ Annotating the data at the level of
devices near the source is sufficient to
address interoperability issues.
▰ Tiny Semantics
30. Autonomous Vehicles
AI in Automotive Market size exceeded USD 1 billion in 2019 and is estimated to grow at
over 35% CAGR between 2020 and 2026 (according to the Global Market Insights).
32. Knowledge Graph Embeddings for Automotive Data
Application: Computing Scene Similarity
Approaches:
- Similarity based on the topology of KG
- Similarity based on the textual descriptions of the scenes in the dataset
- Similarity computed using the Knowledge Graphs Embeddings learned from the Driving Scenes KG
R. Wickramarachchi, C. Henson, and A. Sheth, “An evaluation of knowledge graph embeddings for autonomous driving data: Experience and practice” AAAI
2020 Spring Symposium on Combining Machine Learning and Knowledge Engineering in Practice (AAAI-MAKE 2020), 2020.
33. Knowledge-based Entity Prediction (KEP) in Driving Scenes
Definition: “KEP is the task of predicting the inclusion of potentially unrecognized entities in a scene,
given the current and background knowledge of the scene represented as a knowledge graph”
What’s the probability of
seeing a child nearby?
R. Wickramarachchi, C. Henson, and A. Sheth, “Knowledge-Infused Learning for Entity Prediction in Driving Scenes.” Frontiers in Big Data 4:759110,(2021)
doi: 10.3389/fdata.2021.759110
34. Causal Knowledge Graph
Gary Marcus and Ernest Davis: “we need to stop building computer systems that
merely get better and better at detecting statistical patterns in data sets—often using an
approach known as ‘Deep Learning’—and start building computer systems that from
the moment of their assembly innately grasp three basic concepts: time, space, and
causality.”1
1Bishop, J. Mark. "Artificial Intelligence Is Stupid and Causal Reasoning Will Not Fix It." Frontiers in Psychology 11 (2021): 2603
35. Understanding and Representation of Causal and Counterfactual
phenomenon in the Artificial Intelligence Systems
➢ Causality is a relationship “A” causes “B”
➢ Causality is at the core of everything we see,
do, and imagine.
➢ Human mind has an ability to conceive
alternative, nonexistent worlds known as
counterfactual scenarios
➢ Correlation is not Causation
○ Younger drivers have high probability
of being in an accident
○ Does not imply younger drivers cause
accidents
Representation of causality in artificial intelligence (AI) systems leading to better
explainability and understanding of AI systems by humans
36. Causal Questions in the Driving Scene
Understanding causal relationship between entities in the driving scene
How would a stop line marking in the driving scene effect
the pedestrian behavior (i.e., standing, walking, etc.)?
WHAT IF a pedestrian is jaywalking; how would it effect
the vehicle’s behavior (i.e., stop or keep moving)?
WHAT IF the vehicle fails to identify the stop line marking;
how would it effect the vehicle’s behavior with respective
to pedestrian?
37. Causal Knowledge Graph
Climbing the ladder of causation from association to
counterfactual for improved scene understanding with
Causal Knowledge Graph
39. Health Care,
Public Health &
Life Sciences
According to the report published by Allied Market Research, the global AI in Healthcare Market
generated $8.23 billion in 2020, and is estimated to reach $194.4 billion by 2030, growing at a
CAGR of 38.1% from 2021 to 2030.
41. “In 1970’s, a woman diagnosed with breast cancer had roughly a 40%
chance of surviving the next 10 years. Today, the probability has almost
doubled, thanks to new drugs, cutting-edge screening methods, and
effective surgery”
- Thomas Clozel, TechCrunch (2021)
AI is playing an important role in early detection of breast cancer.
AI shines in the realm of low level tasks such as classification and detection.
Startups using AI to detect breast cancer
(thereby tackling the shortage of radiographers, especially due to the
pandemic)
42. Personalized Digital Health
Patient-generated Health Data (PGHD)
is becoming the most
important data in healthcare.
Source: https://patientengagementhit.com/news/what-are-the-pros-and-cons-of-patient-generated-health-data
https://www.healthaffairs.org/doi/full/10.1377/hlthaff.2015.1362?siteid=healthaff&keytype=ref&ijkey=6C1y7.jaIT7qU&#aff-
1
43. 1. Self Monitoring
Constant and remote
monitoring of
disease specific
health indicators for
any given patient
2. Self Appraisal
Interpretation of
the data collected
with respect to
disease context for
the patient to
evaluate themselves
3. Self
Management
Identify the deviation
from normal and assist
patients to get back to
prescribed care plan
4. Intervention
Change in the care
plan - with the
converted smart data
by APH, provide
decision support for
treatment
adjustments
5. Disease
Progression and
Tracking
Longitudinal data
collection and
analysis to enhance
patients health over
the time
Sheth, et.al. How will the Internet of Things enable Augmented Personalized Health?
Future Health Management Strategy: Augmented Personalized Health
44. 44
Using Chatbots to Go Beyond Traditional
Patient-Doctor Consultation
Socio-
economic
Demo-
graphic
Family &
social
Psychological
Environment
Genetic
Susceptibility
Source: Why do people consult the doctor?
- Stephen M Campbell and Martin O Roland
Decision
Making
Can voice assistant (chatbot) technology
substantially improve monitoring of
patient’s conditions and needs?
Simple Tasks
● Appointment scheduling
● Information retrieval
● Scripted-automation
Complex & Demanding Tasks
● Multimodal input and output
● Natural communication
● Augmented Personalized Health
(serving different levels of health needs)
Contextualization
Personalization
Abstraction
Different modality of data
Images
Text Speech Videos IoTs
45. 45
Mobile Apps and Virtual Health Assistants
Asthma
Nutrition
(Type 1 Diabetes)
Mental Health
Active
mApps/
virtual health
assistant
kHealth Framework: a knowledge-enabled semantic platform
that captures the data and analyzes it to produce actionable
information.
1. NOURICH: Conversational Nutrition
Management (image processing,
nutrition knowledge,....)
1. Personalized Asthma
Management: Contextualized &
Personalized Conversations involving
Multimodal data (IoT & Devices,
Signal Processing)
1. kAgent Mental Health: Self
management of mild mental health
condition (anxiety, depression,...):
knowledge infused reinforcement
learning for enhanced conversation
management with domain/clinical
knowledge and personalization
The global market for healthcare virtual assistants should
grow from $1.1 billion in 2021 to $6.0 billion by 2026, at a
compound annual growth rate (CAGR) of 39.5% for the
period of 2021-2026.
47. NOURICH: Nutrition Management Chatbot
▰ Many diseases can be controlled by proper diet management - diabetes, obesity,
hypertension and so on.
▰ Monitoring an individual's diet and cumulative calorie intake and recommending meals can
help them in making informed decisions about their meals.
A personalized nutrition management chatbot incorporated with AI techniques can aid and assist
the users in this process.
48. AI Techniques and Applications
Techniques
▰ Image Recognition: Semi-supervised learning and meta learning to utilize unlabelled data.
▰ Volume Estimation: Image segmentation to identify food items and estimate volume.
▰ Nutritional Information: Using large nutrition knowledge base to estimate nutrition.
▰ Food Recommendation: Personalized food recommendation using user-specific knowledge graph (if
recommended by clinician) that stores user’s health condition, food preferences and so on.
Applications
▰ Type-1 Diabetes: Patients need to know daily amount of carbohydrate intake.
▰ Hypertension: Patients need to avoid high sodium foods and follow healthy food habits.
49.
50. If the video does not play, check out NOURICH video at:
http://wiki.aiisc.ai/index.php/KHealth_Chatbots
51. AI in Pharmaceuticals
DRUG
DISCOVERY
SELECTION OF
PATIENTS FOR
CLINICAL TRIALS
AUTOMATION OF
PHARMACEUTICAL
REPORTING
● Modelling of different
types of cancer cells to
work out what conditions
allowed the disease to
develop
● Use the information to try
and create new
treatments
● AI Matches drugs to
larger databases of
patients quicker than
human annotation
● Using data from clinical trials
to generate sections of the CSR
report
● Using AI to automate pharma
reports
○ - Pharmacovigilance
● Frees up medical writers’ time
● Allowing them focus on more
high value analysis and adding
technical insight to reports.
Automate report writing
Source: https://pixabay.com/de/illustrations/medizin-pharma-pille-flasche-2801025/, https://www.resources.yseop.com/CSR-use-case
52. AI in Pharmaceuticals: Adverse Drug Reactions
Drug Use/Abuse:
Loperamide Discovery
▰ In a Web forum dataset, it was observed that users reported taking the anti-
diarrhea treatment drug Loperamide (sold over the counter in Imodium) to self-
medicate from withdrawal symptoms. The opioid addictions treatment drugs
Buprenorphine and Methadone are commonly prescribed for treatment of
withdrawal symptoms. Until now, it was unknown that Loperamide, can be (and
is being) used for the same purpose. Which is more, it was observed that users
reported the possibility of mild psychoactive (opiated) effects from megadosing
- which is the practice of taking severely excessive amounts of a drug.
▰ Three toxicology studies followed citing our work.
▰ FDA Warning in 2016.
▰ More at: http://wiki.aiisc.ai/index.php/PREDOSE
Source: R. Daniulaityte, R. Carlson, R. Falck, D. Cameron, S. Perera, L. Chen, A. P. Sheth. "I Just Wanted to Tell You That Loperamide WILL
WORK": A Web-Based Study of Extra-Medical Use of Loperamide. Journal of Drug and Alcohol Dependence. 130(1-3): 241-244, 2013.
54. Public Health - COVID-19 Big Data (USA)
How does real-world events and policy decisions (school closing, nonessential business
closing, number of cases, availability of clinical services), varying by time, geography (e.g.,
state), and demography (GenZ, Millennials, ..) impact public and social health, such as
▰ Mental health including depression
▰ Addiction (alcohol, opioid, marijuana, etc)
▰ Domestic Violence
COVID-related Big data: >8000 Million tweets (~450M with location), ~700K news articles
"A feeling of hopelessness. Seems I am in a dark age. #coronavirus #COVID19"
55. Results: Relative State Rankings Reveal Patterns
e.g., IN, NH, OH, OR, WA, WY are worsening.
56. Results: Cluster - A Non-Linear SQI Ranking
WI, RI, NV,
NJ, CT, LA,
OK.
SQI worse SQI better SQI better
SQI worse
Frequency
Depression: 91,480
Addiction: 103549
Anxiety: 88293
Total: 283322
Frequency
Depression: 62825
Addiction: 81400
Anxiety: 54184
Total: 198409
Frequency
Depression: 58223
Addiction: 76232
Anxiety: 41484
Total: 175949
Frequency
Depression: 78061
Addiction: 87463
Anxiety: 63865
Total: 229389
March14-20 March21-27 March28-April 3
April 4-10
59. Disaster Coordination
DisasterRecord substantially reduces the burden of analysis, interpretation, and decision making during
major disasters. It analyzes geographical data and integrates satellite imagery for better decision making.
▰ Humanitarian organization: analyze the situation at a community level for deploying and mobilizing
necessary help.
▰ First response coordinator: monitor a specific type of emergency needs.
▰ Affected individuals: need to know about the nearest available help.
▰ Persons wishing to provide support: identify current needs in the geographic proximity for the type of
help they can provide.
60. Also online at: http://wiki.aiisc.ai/index.php/DisasterRecord
61. AI in Education with Embibe (India):
Personalized Learning Platform for Everyone through
world’s best Artificial Intelligence Platform in Education
Improve outcome through
behaviour nudges, Machine Learning
61
The global AI in education market is projected to reach USD 3.68 billion by 2023, at a CAGR
of 47% during the forecast period 2018 till 2023.
62. Four Key
Components
Multi-dimensional graph of concepts that
captures the flow of learning through life.
Educational Knowledge Base
Intelligent content authoring and curation
Educational data lake
Intelligent intervention layer
Machine learning and education
domain knowledge combined to
deliver robust learning outcomes
for students and efficiency in
operations for institutions
Massive usage data lake created
and leveraged to power
intelligent intervention & content
authoring
Content creation & curation platform
designed to serve content need while
ensuring diagnosis and remedy happens
at personalised level
AI PLATFORM FOR EDUCATION
STUDENT PRODUCTS
TEACHER PRODUCTS
PARENTS STUDENT
64. Impact on Education using AI
▰ User Intelligence
◆ Learning outcome oriented learning
◆ Personalized learning paths
▰ Content Intelligence
◆ Practically infinite content availability
◇ Automated content creation, curation and tagging
▰ Mentor Intelligence
◆ Automated optimal lesson plans
◆ Social Emotional Learning (SEL)
65. So far, we talked about AI’s success,
BUT
AI is quite overhyped.
AI still has a long way to go.
65
66. What’s Next for AI
M. Jordan. “Artificial Intelligence - The Revolution Hasn’t Happened Yet”, MIT Press, Jul 2019.
67. “The average AI system isn’t smarter than a fifth-
grader”
“We need to build AI that captures how humans think”
Gary Marcus
Professor of Psychology, NYU
Source: https://technical.ly/brooklyn/2017/04/10/nyu-gary-marcus-artificial-intelligence-contrarian/
J. Harris. “The Future of AI: Gary Marcus talks with Lex Fridman”, Medium, Oct 2019.
68. Focus of Most AI Systems so far
Classification Recommendation
Prediction Language Processing and Text Generation
What else do we need for higher levels of
machine intelligence?
69. Narrow, well-defined tasks
(Reflects lower-levels of human-like
intelligence)
Human-like, broad spectrum behavior for
“looking after humans, companion to humans”
(Reflects higher-levels of human-like intelligence:
broad, complex, multi-faceted)