Mais conteúdo relacionado Semelhante a Putting IBM Watson to Work.. Saxena Semelhante a Putting IBM Watson to Work.. Saxena (20) Putting IBM Watson to Work.. Saxena1. IBM
Watson
Making
a
Market,
Making
a
Difference
Manoj
Saxena
General
Manager,
IBM
Watson
Solu:ons
© 2012 IBM Corporation
2. On
February
14,
2011,
IBM
Watson
made
history
Result
of
IBM
Research
“Grand
Challenge”
© 2012 IBM Corporation
2
3. Brief
history
of
IBM
Watson
IBM
Jeopardy!
Watson
Watson
Watson
Research
Project
Grand
Challenge
for
for
Financial
Industry
(2006
–
)
(Feb
2011)
Healthcare
Services
Solu1ons
(Aug
2011
–)
(Mar
2012
–
)
(2012
–
)
Cross-‐industry
Expansion
Applica1ons
Commercializa1on
Demonstra1on
R&D
New
Division
© 2012 IBM Corporation
3
4. The
world
is
geTng
smarter
+ +
Instrumented
Interconnected
Intelligent
Over
10
billion
2
billion
people
Stockholm
CPUs
were
were
on
the
leverages
GPS
data
produced
in
Web
by
2011
...
to
predict
traffic
–
2008,
up
1000%
with
a
trillion
reducing
connected
conges:ons
and
in
8
years.
objects.
emissions.
4
© 2012 IBM Corporation
5. Businesses
are
“dying
of
thirst
in
an
ocean
of
data”
90%
80%
20%
of
the
world’s
data
of
the
world’s
is
the
amount
of
was
created
in
the
data
today
is
available
data
last
two
years
unstructured
tradi:onal
systems
leverages
1
in
2
83%
2.2X
business
leaders
of
CIOs
cited
BI
and
more
likely
that
top
don’t
have
access
to
analy:cs
as
part
of
their
performers
use
data
they
need
visionary
plan
business
analy:cs
© 2012 IBM Corporation
6. So
why
is
it
so
hard
for
computers
to
understand
humans?
Person Organization
“If leadership is an art
L. Gerstner IBM then surely Jack Welch
Welch ran has proved himself a
this? J. Welch GE master painter during
W. Gates Microsoft his tenure at GE.”
Noses that run and feet that smell?
How can a house can burn up as it burns down?
Does CPD represent a complex comorbidity of lung cancer?
What mix of zero-coupon, non-callable, A+ munis fit my risk portfolio?
© 2012 IBM Corporation
7. IBM
Watson
brings
together
transforma:onal
technologies
to
drive
op:mized
outcomes
2
Generates
and
evaluates
hypothesis
for
1
Understands
beZer
outcomes
natural
language
99%
and
human
60%
10%
speech
3
Adapts
and
Learns
from
user
responses
…built
on
a
massively
parallel
architecture
op4mized
for
POWER7
© 2012 IBM Corporation
8. How
Watson
Works:
DeepQA
Architecture
Learned Models
help combine and
weigh the Evidence
Evidence
Sources Balance
Answer Models Models
& Combine
Sources Deep
Inquiry Answer Evidence Models Models
Evidence
Scoring Retrieval
Scoring
Primary Candidate 100,000’s Scores from
many Deep Analysis Models Models
Search Answer 1000’s of
Generation Pieces of Evidence Algorithms
100’s Possible
Answers
100’s
sources
Inquiry/Topic Inquiry Hypothesis Hypothesis and Evidence Final Confidence
Multiple Synthesis
Generation Scoring Merging & Ranking
Interpretations Decomposition
Analysis
of a question
Hypothesis Hypothesis and Evidence
Generation Scoring Responses with
Confidence
© 2012 IBM Corporation
9. Moving
beyond
Jeopardy!
is
a
non-‐trivial
challenge
Watson
at
Play
Watson
at
Work
1
User
10s
of
thousands
concurrent
users
Max.
input
was
two
sentences
Pages
of
input
(e.g.
medical
record)
5+
days
to
retrain
Dynamic
content
inges:on
Evidence
not
present
Suppor:ng
evidence
integral
Text-‐only
input
Text,
tables
and
images
as
input
Q&A
model
Both
Q&A
+
Conversa:on
model
Basic
security
High
security
(e.g.
HIPAA)
© 2012 IBM Corporation
10. Healthcare
industry
is
beset
with
some
of
the
most
complex
informa:on
challenges
we
collec:vely
face
Medical
informa:on
is
doubling
every
5
years,
much
of
which
is
unstructured
81%
of
physicians
report
spending
5
hours
or
less
per
month
reading
medical
journals
“Medicine
has
become
too
complex.
Only
about
20%
of
the
knowledge
clinicians
use
today
is
evidence-‐base.” Steven
Shapiro,
Chief
Medical
&
Scien1fic
Officer,
UPMC
© 2012 IBM Corporation
11. PuTng
the
pieces
together
at
point
of
impact
can
be
life
changing
difficulty swallowing
Symptoms
Family
Medications
Findings
Symptoms
fever
Patient dry mouth Diagnosis
Models
Confidence
thirst
History
A Her medications werepositive forher
58-year-old woman presented to of
A 58-year-old woman levothyroxine,
urine dipstick was complains anorexia
History
primary care esterase and dry mouth,and
leukocyte physician pravastatin,
hydroxychloroquine,after several days
dizziness, anorexia, nitrites. The
of dizziness, given aand frequent fo
increased anorexia, dry mouth,
alendronate.
patient thirst, prescription
frequent urination
dizziness
no abdominal pain
Renal Failure
UTI
increased thirst, notable forhadtract
Herurination. Sheand frequent urination.
Herhistory historyfor aalso cutaneous
ciprofloxacin had urinary a and
family was included oral fever. no back pain
no cough
SheShe reported3 a fever in mother,
lupus, hyperlipidemia, andpatient
had also cancerpain osteoporosis,that
bladder hadno in her her abdomen,
infection. days later, reported no diarrhea
food woulddisease in twoor diarrhea.
reported weakness and sisters, was
Graves'and nostuck” when she
back, “get cough, dizziness. Diabetes
frequent urinary tract infections, a left Oral cancer
swallowing. Shefor inbenign cyst, and
hemochromatosis a one no pain in her
Her supine reported sister, and
oophorectomy blood pressure was
History
Bladder cancer
Family
abdomen,mm Hg, and pulse wascough,
primary hypothyroidism,and no 88. a
120/80 back, or flank diagnosed
idiopathic thrombocytopenic Hemochromatosis Influenza
shortness of breath, diarrhea, or dysuria
purpura inearlier Purpura
year one sister Graves’ Disease
(Thyroid Autoimmune) Hypokalemia
cutaneous lupus
Medications History
Patient
osteoporosis
hyperlipidemia Esophagitis
frequent UTI
hypothyroidism
• Extract Symptoms from record
Alendronate Most Confident Diagnosis: Esophagitis
Most Confident Diagnosis: Influenza
UTI
Diabetes
• Use paraphrasings mined from text to handle
pravastatin • • Extract Medications and variants
• • Extract Patient History
Identify Family History
Extract negative Symptoms
alternate phrasings
levothyroxine • • Use database mined relationsgeneralize medical
• Reason with Taxonomies to to explain away
Use Medical of drug side-effects
• Perform broad search for possible diagnoses
hydroxychloroquine • • Together, multiple is consistent w/ bestthe models
symptoms (thirst granularity may UTI) explain
conditions to the diagnoses used bybased on
Score Confidence in each diagnosis
symptomsso far
evidence
urine dipstick:
Findings
• Extract Findings: Confirms that UTI was present
leukocyte esterase
supine 120/80 mm HG
heart rate: 88 bpm
urine culture: E. Coli
11
© 2012 IBM Corporation
12. Cancer
is
an
insidious
disease
and
the
second
highest
cause
of
death
1
in
3
3X
individuals
will
die
rate
cancer
cost
climbs
vs.
from
cancer
std.
health
costs
or
15-‐18%
/
yr.
X
+ + IBM
20%
$263.8B
Workingin
Together to Beat Cancer
of
cancer
cases
receive
the
overall
costs
of
cancer
wrong
diagnosis
ini:ally
the
US
in
2010
with
some
as
high
as
44%
$$$$$$$$$$$$
✔ ✔ $$$$$$$$$$$$
✔ ✔
✔ ✔ ✔
$$$$$$$$$$$$
$$$$$$$$$$$$ + + IBM
Working Together to Beat Cancer
Source: American Cancer Society, National Health Institute
12 © 2012 IBM Corporation
13. IBM Watson goes to work in healthcare
1.
Accelerate
Time
to
2.
Improve
Decisions
Clinical
Insights
and
Outcomes
Support
researchers
and
clinicians
Assist
physicians
and
care
providers
in
discovery
of
new
cancer
with
evidence
based
diagnosis
and
therapies
treatment
Care
Provider
Genomic
Researcher
Analy:cs
Expert
Pa:ent
Therapy
Designer
Oncologist
MEDICAL
RESEARCH
MEDICAL
PRACTICE
&
PAYMENTS
13 Ultimate Goal: Become the Most Essential Company © 2012 IBM Corporation
14. Demonstra:on
of
Watson
Cancer
Care
Solu:on
IBM Watson
Oncology Advisor
IBM Confidential: References to potential future products are subject to the Important Disclaimer provided earlier in the presentation
14 © 2012 IBM Corporation
15. Watson
enables
three
classes
of
cogni:ve
solu:ons
Ask
•
Leverage
vast
amounts
of
data
•
Ask
ques:ons
for
greater
insights
•
Natural
language
inquiries
•
e.g.
-‐
Next
genera:on
Chat
Discover
• Find
the
ra:onale
for
given
answers
• Prompt
for
inputs
to
yield
improved
responses
• Inspire
considera:ons
of
new
ideas
• e.g.
-‐
Next
genera:on
Search
Discovery
Decide
• Ingest
and
analyze
domain
sources,
info
models
• Generate
evidence
based
decisions
with
confidence
• Learn
with
new
outcomes
and
ac:ons
• e.g.
-‐
Next
genera:on
Apps
Probabilis:c
Apps
15 © 2012 IBM Corporation
16. Imagine if…
… call center agents could
find better answers to
customer questions 50%
faster.
That’s exactly what a major
provider of financial
management software did.
“Contact centers of the future will
improve precision and personalization,
transforming centers from a cost
orientation to a strategic assets.”
- Leading Telco Supplier
ASK
16 © 2012 IBM Corporation
17. Imagine if…
. . . new insights from medical
research find their way to patient
treatment programs in months
instead of years?
That’s exactly what a global
leader in cancer care is doing
today.
“Watson will be an invaluable resource
for our physicians and will dramatically
enhance the quality and effectiveness of
medical care.”
- Dr Sam Nussbaum,
Chief Medical Officer, WellPoint
17
DISCOVER
© 2012 IBM Corporation
18. Imagine if…
. . . the 1.5M people
diagnosed with cancer in the
US last year had a better
prognosis?
That’s exactly what a major
health plan provider is
working to accomplish.
“Watson can aggregate information
and give probabilities that will
enable (experts) to zero in on the
most likely diagnosis.”
- Dr. Steven Nissen,
Cleveland Clinic
DECIDE
18 © 2012 IBM Corporation
19. How
it
Works:
Watson
Technology
&
Infrastructure
Watson
for
Watson
For
Watson
for
Client
Watson
for
Healthcare
Financial
Svcs.
Engagement
Industry
Advisor
Solu1ons
Advisor
Solu1ons
Advisor
Solu1ons
Ins1tu1onal
Re1rement
Knowledge
Call
Center
U1liza1on
Ins1tu1on
Help
Desk
Oncology
Technical
Diabetes
Cardiac
Banking
ASK Services DISCOVER Services DECISION Services
NLP & Machine Big Data Analytics Cloud Mobile Workload Optimized
Learning Systems
Source Model Train Learn
19 © 2012 IBM Corporation
20. How
it
Works:
Cloud
Delivery
and
Outcome
Based
Pricing
Dynamic
Capacity
Hybrid
Delivery
Automate
and
control
Extend
&
integrate
service
provisioning
on-‐premise
solu:on
with
cloud
offering
Flexible
Consump1on
Time
to
Value
Support
alterna:ve
Enable
incremental
delivery
and
value
automa:on
and
pricing
models
business
agility
20 © 2012 IBM Corporation
21. Getting Started
with
Watson
GeTng
Started with Watson
1.
Discovery
2.
Pilots
3.
Scale
Out
Workshops
(6-‐12
Months)
(Ongoing)
(4-‐6
Weeks)
Step
1:
Op:on
A:
Step
3:
• Iden:fy
Candidate
Deploy
ini:al
pilot
• Add
Users
Use
Cases
• Build
• Add
Geographies
• Assess
Content
• Teach
• Add
Content
Availability
• Run
• Expand
Processes
• Model
Business
Op:on
B:
Value
Apply
Ready
for
Watson
Progression
Paths
• Smarter
Analy:cs
• Big
Data
• Industry
Solu:ons
21 © 2012 IBM Corporation
22. Watson
is
ushering
in
a
new
era
of
compu:ng
.
.
.
System
Intelligence
Cogni1ve
Programma:c
Search
Discovery
Determinis:c
Probabilis:c
Enterprise
data
Big
Data
1
Tabula:on
Machine
language
Natural
language
Punch
cards
Time
card
readers
Simple
outputs
Intelligent
op:ons
1900 1950 2011
.
.
.enabling
new
opportuni:es
and
outcomes
1
22 IBM Confidential © 2012 International Business Machines Corporation