Kazakhstan digital media_at_svl 20191018 v5

Future of AI
Jim Spohrer
Director, IBM Cognitive OpenTech
Presentations on line at: http://slideshare.net/spohrer

10/18/2019 (c) IBM MAP COG .| 2

Questions
• What is the timeline for solving AI and IA?
• TBD: When can a CEO buy AI capability <X> for price <Y>?
• Who are the leaders driving AI progress?
• What will the biggest benefits from AI be?
• What are the biggest risks associated with AI, and are they real?
• What other technologies may have a bigger impact than AI?
• What are the implications for stakeholders?
• How should we prepare to get the benefits and avoid the risks?
10/18/2019 (c) IBM 2017, Cognitive Opentech Group 3

Timeline: Every 20 years,
compute costs are down by 1000x
• Cost of Digital Workers
• Moore’s Law can be thought of as
lowering costs by a factor of a…
• Thousand times lower
in 20 years
• Million times lower
in 40 years
• Billion times lower
in 60 years
• Smarter Tools (Terascale)
• Terascale (2017) = $3K
• Terascale (2020) = ~$1K
• Narrow Worker (Petascale)
• Recognition (Fast)
• Petascale (2040) = ~$1K
• Broad Worker (Exascale)
• Reasoning (Slow)
• Exascale (2060) = ~$1K
410/18/2019 (c) IBM 2017, Cognitive Opentech Group
2080204020001960
$1K
$1M
$1B
$1T
206020201980
+/- 10 years
$1
Person Average
Annual Salary
(Living Income)
Super Computer
Cost
Mainframe Cost
Smartphone Cost
T
P
E
T P E
AI Progress on Open Leaderboards
Benchmark Roadmap to solve AI/IA

Timeline: GDP/Employee
(Source)
Lower compute costs translate into increasing productivity and GDP/employees for nations
Increasing productivity and GDP/employees should translate into wealthier citizens

Timeline: Leaderboards FrameworkAI Progress on Open Leaderboards - Benchmark Roadmap
Perceive World Develop Cognition Build Relationships Fill Roles
Pattern
recognition
Video
understanding
Memory Reasoning Social
interactions
Fluent
conversation
Assistant &
Collaborator
Coach &
Mediator
Speech Actions Declarative Deduction Scripts Speech Acts Tasks Institutions
Chime Thumos SQuAD SAT ROC Story ConvAI
Images Context Episodic Induction Plans Intentions Summarization Values
ImageNet VQA DSTC RALI General-AI
Translation Narration Dynamic Abductive Goals Cultures Debate Negotiation
WMT DeepVideo Alexa Prize ICCMA AT
Learning from Labeled Training Data and Searching (Optimization)
Learning by Watching and Reading (Education)
Learning by Doing and being Responsible (Exploration)
2018 2021 2024 2027 2030 2033 2036 2039
Which experts would be really surprised if it takes less time… and which experts really surprised if it takes longer?
Approx.
Year
Human
Level ->
+3

10/18/2019
© IBM 2015, IBM Upward University Programs Worldwide
accelerating regional development
8

Who is winning
https://www.technologyreview.com/s/608112/who-is-winning-the-ai-race/

Robots by Country
• Industrial robots per 10,000 people by country
10/18/2019 IBM #OpenTechAI 10
34

Sweden
10/18/2019 (c) IBM MAP COG .| 11

Economic Growth Rates 2035: AI Projected Impact
10/18/2019 (c) IBM MAP COG .| 12

AI Benefits
• Access to expertise
• “Insanely great” labor productivity for trusted service providers
• Digital workers for healthcare, education, finance, etc.
• Better choices
• ”Insanely great” collaborations with others on what matters most
• AI for IA = Augmented Intelligence and higher value co-creation interactions

AI Risks
• Job Loss
• Shorter term bigger risk
= de-skilling
• Super-intelligence
• Shorter term bigger risk
= bad actors

Other Technologies: Bigger impact? Yes.
• Augmented Reality (AR)/
Virtual Reality (VR)
• Game worlds
grow-up
• Blockchain/
Security Systems
• Trust and security
immutable
• Advanced Materials/
Energy Systems
• Manufacturing as cheap,
local recycling service
(utility fog, artificial leaf, etc.)

“The best way to predict the future is to inspire the
next generation of students to build it better”
Digital Natives Transportation Water Manufacturing
Energy Construction ICT Retail
Finance Healthcare Education Government

Step Comment
GitHub Get an account and read the guide
MAX CODAIT’s Model Asset Exchange
Learn 3 R's - Read, Redo, Report Read (Medium/arXiv), Redo (GitHub), Report (Jupyter Notebook)
PapersWithCode Stay on top of recent advances; Do 3 R’s.
Kaggle Compete in a Kaggle competition
Leaderboards Compete to advance AI progress
Linux Foundation AI Help end-to-end open source industry AI & Data infrastructure
Mozilla Common Voice Donate your speech; Label and verify data; Recruit others.
Figure Eight Generate a set of labeled data (also Mechanical Turk)
Design New Challenges Build for Call for Code/Code and Response; Build your AI Helper;
Build test-taker, that can switch to tutor-mode; Etc.
Open Source Guide Establish open source culture in your organization
10/18/2019 IBM Code #OpenTechAI 17

10/18/2019 (c) IBM MAP COG .| 18
Microsoft acquiring GitHub $7.5B
2018 John Marks on Open Source
Models will run the world
Why SW is eating the world

10/18/2019 19
1955 1975 1995 2015 2035 2055
Better Building Blocks

10/18/2019 (c) IBM MAP COG .| 20

10/18/2019
21
Cognitive Mediators
for all people in all roles

Occupations = Many Tasks
10/18/2019
22

Watson Discovery Advisor
10/18/2019
23
Simonite, T. 2014. Software Mines Science Papers to Make New Discoveries. MIT. November 25, 2014.
URL: http://m.technologyreview.com/news/520461/software-mines-science-papers-to-make-new-discoveries/

10 million minutes of experience
10/18/2019 Understanding Cognitive Systems 24

2 million minutes of experience

Hardware < Software < Data < Experience < Transformation
Value migrates to transformation – becoming our future selves; people, businesses, nations = service system entities
Pine & Gilmore (1999)
Transformation
Roy et al (2006)
Data
Osati (2014)
Experience
Life Log

Courses
• 2015
• “How to build a cognitive system for Q&A task.”
• 9 months to 40% question answering accuracy
• 1-2 years for 90% accuracy, which questions to reject
• 2025
• “How to use a cognitive system to be a better
professional X.”
• Tools to build a student level Q&A from textbook in 1
week
• 2035
• “How to use your cognitive mediator to build a
startup.”
• Tools to build faculty level Q&A for textbook in one day
• Cognitive mediator knows a person better than they
know themselves
• 2055
• “How to manage your workforce of digital workers.”
• Most people have 100 digital workers.
10/18/2019 27
Take free online cognitive classes today at cognitiveclass.ai

10/18/2019
28
I have…
Have you noticed how the building blocks just
keep getting better?

Learning to program:
My first program
10/18/2019
29
Early Computer Science Class:
Watson Center at Columbia 1945
Jim Spohrer’s
First Program 1972

10/18/2019
© IBM UPWard 2016
30
Fast Forward 2016:
Consider this…

Microsoft CaptionBot June 19, 2016
10/18/2019
© IBM UPWard 2016
31

Microsoft CaptionBot June 20, 2016
10/18/2019
© IBM UPWard 2016
32

IBM Image Tagging
10/18/2019
© IBM UPWard 2016
33

Today: November 10, 2017
10/18/2019
© IBM DBG COG 2017
34
IBM

Types: Progression of models
Models = instruction set of future
Task & World Model/
Planning & Decisions
Self Model/
Capacity & Limits
User Model/
Episodic Memory
Institutions Model/
Trust & Social Acts
Tool + - - -
Assistant ++ + - -
Collaborator +++ ++ + -
Coach ++++ +++ ++ +
Mediator +++++ ++++ +++ ++
Cognitive
Tool
Cognitive
Assistant
Cognitive
Collaborator
Cognitive
Coach
Cognitive
Mediator

Trust: Two Communities
Service
Science
OpenTech
AI
Trust:
Value Co-Creation,
Transdisciplinary
Trust:
Ethical, Safe, Explainable,
Open Communities
Special Issue
AI Magazine?
Handbook of
OpenTech AI?

Resilience:
Rapidly Rebuilding From Scratch
• Dartnell L (2012) The Knowledge: How to
Rebuild Civilization in the Aftermath of a
Cataclysm. Westminster London: Penguin
Books.

Narrow AI
Emerging
Broad AI
Disruptive and
Pervasive
General AI
Revolutionary
▼ We are here 2050 and beyond 38IBM Research AI © 2018 IBM Corporation
The evolution of AI
Borrowed from David Cox, IBM-MIT Lead

Narrow AI
Single task, single domain
Superhuman accuracy and
speed for certain tasks
Broad AI
Multi-task, multi-domain
Multi-modal
Distributed AI
Explainable
General AI
Cross-domain
learning and reasoning
Broad autonomy
39IBM Research AI © 2018 IBM Corporation
The evolution of AI

“Teddy Bear”
Meret Oppenheim, Le Déjeuner en fourrure

Gatys et al. 2015 Brock et al. 2018

Lake, Ullman, Tenenbaum & Gershman, 2016

Chen et al. 2018
Pin-yu Chen
IBM

The path to Broad AI
Explainability Security
+
Ethics
restaurant
cook
Follow
recipe
person
sweet
cheesecake
dessert
satisfy
hunger
oven
bake
survive
swallow
eat
cake
Learn more from small data
ReasonsLearns to transfer
+
Infrastructure
Physics of AI

IBM-MIT $240M
over 10 year AI mission

Cognitive OpenTech & CODAIT / Future of AI / March 23, 2018 / © 2019 IBM Corporation 47
Linus Torvalds & Todd Moore

Join: Registrations Open 2019-03-25 - https://callforcode.org/
48IBM Open Technologies / Future of AI / April 19, 2019 / © 2019 IBM Corporation
This multi-year global initiative rallies developers to create practical, effective, and high-quality applications based on cloud, data, and artificial
intelligence that can have an immediate and lasting impact on humanitarian issues. Call for Code brings startup, academic, and enterprise developers
together and inspires them to solve the most pressing societal issues of our time - for example, faster and more resilient recovery from natural disasters.

IBM MAX
• Model Asset Exchange
10/18/2019 (c) IBM MAP COG .| 50

AI Fairness 360
10/18/2019 (c) IBM MAP COG .| 51

Timeline: Short History
10/18/2019
© IBM Cognitive Opentech Group (COG)
53
Dota 2
“Deep Learning” for
“AI Pattern Recognition”
depends on massive
amounts of “labeled data”
and computing power
available since ~2012;
Labeled data is simply
input and output pairs,
such as a sound and word,
or image and word, or
English sentence and French
sentence, or road scene
and car control settings –
labeled data means having
both input and output data
in massive quantities.
For example, 100K images
of skin, half with skin
cancer and half without to
learn to recognize presence
of skin cancer.

Today’s talk
• Introduction
• AI at IBM: Past, Present, Future (Summary)
• Types of Systems
• AI at the peak of the hype cycle
• What’s really going on?
• Your data is becoming your AI… IA transformation
• Part 1: Solving AI: Leaderboards
• Roadmap and implications
• Open technologies, innovation
• Part 2: Solving IA: Better Building Blocks
• Solving problems faster, creates new problems
• Identity, social contracts, trust, resilience

AI at IBM: Past (Nathan Rochester)
10/18/2019 (c) IBM MAP COG .| 56

10/18/2019
© IBM UPWard 2016
57
AI (Artificial Intelligence) is popular again… you see it mentioned on billboards in SF
However, pattern recognition does not equal AI
Deep learning works if you have lots of data and compute power
We finally have lots of data and compute power – hurray!!!
So finally, deep learning for pattern recognition is working pretty well
However, AI is more than deep learning for pattern recognition…
AI requires commonsense reasoning – that will take another 5-10 years of research
How do we know this? Look at the AI leaderboards – we will get to that…

10/18/2019 (c) IBM MAP COG .| 58

Smartphones pass entrance exams? When?
… when will
your smartphone
be able to take and
pass any online
course? And then
be your coach, so
you can pass too?

IBM-MIT $240M
over 10 year AI mission

IBM Quantum
10/18/2019 (c) IBM MAP COG .| 61

Quantum
Risk
Assessment
10/18/2019 (c) IBM MAP COG .| 62
URL: https://www.nature.com/articles/s41534-019-0130-6

Icons of AI Progress
• 1956: Dartmouth Conference
organized by:
• John McCarthy (Dartmouth, later
Stanford)
• Marvin Minsky (MIT)
• and two senior scientists:
• Claude Shannon (Bell Labs)
• Nathan Rochester (IBM)
• 1997: Deep Blue (IBM) - Chess
• 2011: Watson Jeopardy! (IBM)
• 2016: AlphaGo (Google DeepMinds)

Timeline: Short History
10/18/2019
© IBM Cognitive Opentech Group (COG)
64
Dota 2
“Deep Learning” for
“AI Pattern Recognition”
depends on massive
amounts of “labeled data”
and computing power
available since ~2012;
Labeled data is simply
input and output pairs,
such as a sound and word,
or image and word, or
English sentence and French
sentence, or road scene
and car control settings –
labeled data means having
both input and output data
in massive quantities.
For example, 100K images
of skin, half with skin
cancer and half without to
learn to recognize presence
of skin cancer.

Timeline: Every 20 years,
compute costs are down by 1000x
• Cost of Digital Workers
• Moore’s Law can be thought of as
lowering costs by a factor of a…
• Thousand times lower
in 20 years
• Million times lower
in 40 years
• Billion times lower
in 60 years
• Smarter Tools (Terascale)
• Terascale (2017) = $3K
• Terascale (2020) = ~$1K
• Narrow Worker (Petascale)
• Recognition (Fast)
• Petascale (2040) = ~$1K
• Broad Worker (Exascale)
• Reasoning (Slow)
• Exascale (2060) = ~$1K
6510/18/2019 (c) IBM 2017, Cognitive Opentech Group
2080204020001960
$1K
$1M
$1B
$1T
206020201980
+/- 10 years
$1
Person Average
Annual Salary
(Living Income)
Super Computer
Cost
Mainframe Cost
Smartphone Cost
T
P
E
T P E

Stakeholders = service system entities
• Individuals
• Families
• Businesses and
other Organizations
• Industry Groups and
Professional Associations
• Regional
Governments:
• Cities
• States
• Nations
“there is nothing as practical as a good abstraction” -> service science studies service system entities

Artificial Leaf
• Daniel Nocera, a professor of energy
science at Harvard who pioneered the
use of artificial photosynthesis, says that
he and his colleague Pamela Silver have
devised a system that completes the
process of making liquid fuel from
sunlight, carbon dioxide, and water. And
they’ve done it at an efficiency of 10
percent, using pure carbon dioxide—in
other words, one-tenth of the energy in
sunlight is captured and turned into fuel.
That is much higher than natural
photosynthesis, which converts about 1
percent of solar energy into the
carbohydrates used by plants, and it
could be a milestone in the shift away
from fossil fuels. The new system is
described in a new paper in Science.

Food from Air
• Although the technology is in its infancy,
researchers hope the "protein reactor"
could become a household item.
• Juha-Pekka Pitkänen, a scientist at VTT,
said: "In practice, all the raw materials
are available from the air. In the future,
the technology can be transported to,
for instance, deserts and other areas
facing famine.
• "One possible alternative is a home
reactor, a type of domestic appliance
that the consumer can use to produce
the needed protein."
• According to the researchers, the
process of creating food from electricity
can be nearly 10 times as energy
efficient as photosynthesis, the process
used by plants.

Exoskeletons for Elderly
• A walker is a “very cost-effective”
solution for people with limited
mobility, but “it completely
disempowers, removes dignity,
removes freedom, and causes a
whole host of other psychological
problems,” SRI Ventures president
Manish Kothari says. “Superflex’s
goal is to remove all of those areas
that cause psychological-type
encumbrances and, ultimately,
redignify the individual."

Computing: Then, Now, Projected
10/18/2019 70
2035
2055

Be Prepared
• Understand open AI code + data +
models + stacks + communities
• Leaderboards
• Ethical conduct
• Learn 3 R’s of IBM’s Cognitive
Opentech Group (COG)
• Read arXiv
• Redo with Github
• Report with Jupyter notebooks on DSX
and/or leaderboards
• Improve your team’s skills of rapidly
rebuilding from scratch
• Build your open code eminence
• Understand open innovation
• Communities + Leaderboards
1972 used
Punch cards
2016 used
IBM Watson
Open APIs to win…

10/18/2019 72
1955 1975 1995 2015 2035 2055
Better Building Blocks

Cupertino Teens
• IBM Watson on Bluemix
AI for NLP
entity identification

10/18/2019
75
Cognitive Mediators
for all people in all roles

Occupations = Many Tasks
10/18/2019
76

Watson Discovery Advisor
10/18/2019
77
Simonite, T. 2014. Software Mines Science Papers to Make New Discoveries. MIT. November 25, 2014.
URL: http://m.technologyreview.com/news/520461/software-mines-science-papers-to-make-new-discoveries/

Types: Progression of models
Models = instruction set of future
Task & World Model/
Planning & Decisions
Self Model/
Capacity & Limits
User Model/
Episodic Memory
Institutions Model/
Trust & Social Acts
Tool + - - -
Assistant ++ + - -
Collaborator +++ ++ + -
Coach ++++ +++ ++ +
Mediator +++++ ++++ +++ ++
Cognitive
Tool
Cognitive
Assistant
Cognitive
Collaborator
Cognitive
Coach
Cognitive
Mediator

Prepare for AI Future
• Do you have a GitHub account? Get it.
• Yes: proceed
• No: sign up
• Do you program? Either OK, partnering is best.
• Yes: Learn and do 3 R’s (read, redo, report)
• Github master: Code, Content (Data), Community (IBM Code can help)
• No: Learn to read and execute code with partner (T2T)
• Do you have favorite AI leaderboards?
• Yes: Learn and do 3 R’s (read, redo, report advances)
• Kaggle master: Combine top decorrelated solution, new solution
• No: Find a mentor with favorites, do together
• Are you AI prepared? Do you know/do data, models, solutions?
• Yes: Find favorite leaderboards you can do 3 R’s for today
• Figure-Eight master: Labeled data that matters most
• No: Wait until one model: one model that can do them all
• Then rapidly rebuild in least time, energy (“zorch”), data, code
10/18/2019
© IBM Cognitive Opentech Group 2018
79
1. Where do we get labeled data?
We create it: Figure Eight,
Mechanical Turk, etc.
2. External/internal challenge?
10M minutes from birth to adult
2M minutes from novice to expert
Not just external states, but
internal states are data as well…
The challenge of data for AI models
3. AI models as ”data” instruction set
Computer’s have instruction sets
Arithmetic, Logic, etc.
Models are becoming instructions
Models are data/experience

Call for papers: Advances in Cognitive Systems, Aug 2-5,
MIT, Cambridge, MA
http://www.cogsys.org/conference/2019/ … submission
deadline: May 10th. @neurobongo
https://twitter.com/JimSpohrer/status/1118180598583
947269

Cognitive OpenTech
Director, Jim Spohrer CODAIT (Center for
Opensource Data and AI
Technologies)
- Vijay Bommireddipalli
AI OpenTech &
Systems Performance
- Tom Truong
Watson Developer
Advocates
- Josh Cheng
Susan Malaika
(STSM)
Augustina Ragwitz
Open AI Advocato,
Computational Anthropology IBM Silicon Alley, NYC, NYIBM Silicon Valley, San Jose, CAIBM Watson West,
San Francisco, CA
IBM Almaden,
San Jose, CA

Kazakhstan digital media_at_svl 20191018 v5

Recommended

Recommended

More Related Content

What's hot

What's hot (20)

Similar to Kazakhstan digital media_at_svl 20191018 v5

Similar to Kazakhstan digital media_at_svl 20191018 v5 (18)

More from ISSIP

More from ISSIP (20)

Recently uploaded

Recently uploaded (20)

Kazakhstan digital media_at_svl 20191018 v5

Editor's Notes