3. What Is Driving the Need for Cognitive Computing?
3
Percentage of
unstructured data
We are here
Sensors
& Devices
Social
Media
VOIP
Enterprise Da
44 zettabytes
2010 2015 2020
4. We are Entering a New Era of Computing
4
Programmable
Systems Era
Cognitive
Systems Era
Tabulating
Systems Era
cog.ni.tive: of or pertaining
to the mental process of perception
memory, judgment, learning, and
reasoning
5. 1997: Deep Blue
IBM Deep Blue defeats World
Chess Champion
1950: Turing Test
Turing introduces way to test
for intelligent behavior
Pioneers and Significant Events Have Shaped Where We Are Today …
5
1950s 1960s 1970s 1980s 1990s 2000s 2010…
1956: “Birth” of AI
John McCarthy coins term
artificial intelligence (AI) at
Dartmouth Conference
1965: First Expert System
Stanford team led by Ed Feigenbaum
creates DENDRAL
1987 - 1993: 2nd AI “Winter”
1990s: AI on www
AI-based extraction programs
prevalent on www
2011: Watson
IBM’s Watson competes and wins
on Jeopardy!
2005: Autonomous Car
Stanford-built autonomous car wins
DARPA Grand Challenge
2014: Key Market Moves
IBM formation of Watson Group and
Google acquisition of Nest Labs
1974 - 1980: 1st AI “Winter”
6. Alarmists? …. or Realists?
6Bill Gates
Stephen Hawking
Elon Musk “The
development
of
full
artificial
intelligence
could
spell
the
end
of
the
human
race
……
It
would
take
off
on
its
own,
and
re-design
itself
at
an
ever
increasing
rate
……
Humans,
who
are
limited
by
slow
biological
evolution,
couldn’t
compete,
and
would
be
superseded.”
“I
think
we
should
be
careful
about
artificial
intelligence
….
If
I
had
to
guess
at
what
our
biggest
existential
threat,
it
is
probably
that
…..
With
artificial
intelligence,
we’re
summoning
the
demon.”
“First
the
machines
will
do
a
lot
of
jobs
for
us
and
not
be
super
intelligent
…..That
should
be
positive
if
we
manage
it
well
…..
A
few
decades
after
that
though
the
intelligence
is
strong
enough
to
be
a
concern.”
9. So, What Is Cognitive Computing?
9
▪ Cognitive computing and cognitive based systems accelerate, enhance and scale
human expertise by:
Learning and building knowledge,
Understanding natural language and
Interacting more naturally with humans than traditional programmable systems
▪ Over time, cognitive systems will simulate more of how the brain actually works
and help us solve the world's most complex problems by penetrating the
complexity of Big Data
käg-nəә-tiv (adjective): of, relating to, or involving conscious mental
activities (such as thinking, understanding, learning, and
remembering)
10. What are Cognitive Systems Good At?
10
▪ Cognitive systems learn by extracting and organizing the signals emitted in the
natural world, and evaluating patterns that convey meaning
▪ Cognitive systems are especially valuable when dealing with large quantities of
unstructured information (such as text, audio, or video) and disparate
information sources that would otherwise overwhelm the time and space
constraints of human assimilation
Exploration
Collect the
information that you
need to explore your
problem area better
Engagement
Dialog with end users
to answer the
questions needed
around products and
services
Discovery
Help find the
questions you’re not
thinking to ask and
connect the dots that
you’re missing that
will lead to new
inspiration
Evaluation
Evaluate a presented
condition against a
set of written policy
assertions
Decision
Assess the choices
that enable you to
make better decisions
11. 11
Core Technologies
Question
& Answer
Natural
Language
Processing
Machine
Learning
Question
Analysis
Feature
Engineering
Ontology
Analysis
Watson for Jeopardy Comprised a Single API Built
on Five Core Technologies
12. Since Then We Have Grown to 28 APIs –
Based on ~50 Core Technologies
12
Watson News
Speech
to Text
Image
Link
Extraction
Tradeoff
Analytics
Concept
Tagging Image Tagging
Natural
Language
Classifier
Retrieve and
Rank
Author
Extraction
Visual
Recognition
Message
Resonance
Language
Detection
Tone
Analyzer
Question
& Answer
Entity
Extraction
Concept
Expansion
Sentiment
Analysis
Personality
Insights Feed Detection
Face Detection Dialog Keyword
ExtractionTaxonomy
Language
Translation
Concept
Insights Text Extraction
Text to Speech Relationship
Extraction
Question
& Answer
Author Extraction
Colloquialism Processing
Concept Expansion
Convolutional Neural Networks
Deep Learning
Dialog
Entity Extraction
Entity Resolution
Feature Engineering
Feature Weighting
Core Technologies
13. Draws on Five Core Technologies
Speech
to Text
Image
Link
Extraction
Tradeoff
Analytics
Concept
Tagging Image Tagging
Natural
Language
Classifier
Retrieve and
Rank
Author
Extraction
Visual
Recognition
Message
Resonance
Language
Detection
Tone
Analyzer
Question
& Answer
Entity
Extraction
Concept
Expansion
Sentiment
Analysis
Personality
Insights Feed Detection
Face Detection Dialog Keyword
ExtractionTaxonomy
Language
Translation
Concept
Insights Text Extraction
Text to Speech Relationship
Extraction
Case
Evaluation
Q&A
Qualification
Video
Augmentation
Policy
Identification
Knowledge
Graph
Criteria
Classification
Risk
Stratification
Factoid
Pipeline
Usage Insights
Easy
Adaptation
Answer
Generation
Decision
Optimization
Knowledge
Studio Service
Fusion QA
Emotion
Analysis
Knowledge
Canvas
Statistical
Dialogue
Decision
Support
Core Technologies
Author Extraction
Colloquialism Processing
Concept Expansion
Convolutional Neural Networks
Deep Learning
Dialog
Entity Extraction
Entity Resolution
Feature Engineering
Feature Weighting
12
Watson News
In 2016, We Will Add an Additional 15 - 20 APIs
14. Watson for Oncology
Provides clinicians with confidence-ranked,
evidence-based personalized treatment options
based on expert training from MSK physicians
Ingests 300+ medical journals, 200+ textbooks, 15M+ pages of text,
thousands of historical cases and thousands of hours of MSK physician and
analyst training (in conjunction with Watson application Knowledge Studio).
Connects treatment recommendations to supporting evidence from MSK-
curated literature and provides physicians ranked, personalized evidence-
based cancer treatment options for consideration.
Entity
Extraction
Concept
Insights
Retrieve and
Ranke
Together, these APIs power the summation of attributes from longitudinal
patient records to extract meaningful information from natural language –
including the unstructured data in clinician's notes.
Document
Conversion
13
15. Relationship
Extraction
Finds relationships between ingredients from a corpus of recipes to
suggest new kinds of pairings that may not be intuitive to chefs. Helps
Watson understand information about ingredient parts and fabrication
(e.g., a lobster has a shell, but a salmon has skin; oranges are peeled but
blueberries are not).
Parses unstructured English language of the recipes’ content into
structured text and then maps recipes to dish types (i.e. to understand
what recipe is a taco, a dessert pie, a savory pie, etc.).
Natural
Language
Classifier
Entity
Extraction
Identifies all the ingredients in a recipe, the purpose of each ingredient
and how it complements other ingredients.
14
Chef Watson
Assisting chefs in choosing the right combination
of ingredients considering flavor, texture, and
chemical composition of millions of ingredients
16. Question
&
Answer
Allows the Digital Virtual Assistant to draw responses from its corpus of
thousands of pages of GEICO training manuals, policies, and employee
expertise.
Allows customers to ask contextual questions (e.g., “where can I find my
vehicle information number” or “VIN number”), in a very natural way. This
API also learns about the customer from client records, and guides them
through the process based on their unique situation.
Dialog
15
Watson-powered "Digital Virtual Assistant"
Helps guide Geico's customers through the experience
of selecting an insurance policy
17. Watson Platform Built on IBM Bluemix
17
▪ Build your application using callable Watson Service APIs
▪ Can be combined with the 100s of other available services on Bluemix
Language
Translation
Speech to TextText to Speech Dialog Tradeoff
Analytics
Personality
Insights
Natural Lang
Classifier
Concept
Insights
Concept
Expansion
Question and
Answer
Relationship
Extraction
Visual
Recognition
Tone AnalyzerRetrieve and
Rank
Document
Conversion
Message
Resonance
AlchemyAPI
▪ Community of 11,500 developers
- 1,600 daily visitors
- 7,600+ non-IBM organizations
- 10,200+ applications bound to
Watson Services
- 20M+ API calls served in the last
30 days
18. U.S. and EU Governments Investing in Cognitive Computing
18
U.S. Government Agencies
Mission: Understand brain and its diseases;
develop brain-like technologies
135 partner institutions in 26 countries
Funding: 1.2 billion euros over 10 years
Mission: Partnership with IBM to further cognitive
computing and big data research
Funding: UK Government: £ 113M
IBM: £ 200M in people, hardware & software
Mission: DARPA SyNAPSE
Build computer with similar form and
function to dog or cat brain
IARPA, DARPA, DoD, NSF AI,
knowledge discovery, neuroscience
Funding: $ 15M per year in NSF funding
> $100M funding for understanding brain
Mission: Advance cognition, human-robot interaction,
mechatronics, navigation, perception
Funding: 80B euros, 2014 - 2020, from government and
EU private industry
20. Preparing for Tomorrow
20
▪ Digital technologies will continue to accelerate
▪ Business-as-usual won’t solve the problem
▪ A major commitment to increasing education and
skill levels as well as fostering business and
organization innovation is required
▪ Need to reinvent our economy and society to keep
up with accelerating technology