Artificial Intelligence - Impacts of Increased Computing Power on Commerce

conyrankAI and Robotics

Jul 17, 2012 (4 years and 11 months ago)

375 views

Eric Horvitz
Microsoft Research
Washington DC
November 2006
Artificial Intelligence
Artificial Intelligence
horvitz@microsoft.com
horvitz@microsoft.com
http://research.microsoft.com/~horvitz/
http://research.microsoft.com/~horvitz/
© Eric Horvitz 2006
Impacts of Increased Computing Power on Commerce
Artificial Intelligence
Artificial Intelligence


Representation & reasoning with knowledge
Representation & reasoning with knowledge


Sensing, learning & adaptation
Sensing, learning & adaptation


Search & optimization
Search & optimization


Theoretical principles & applications
Theoretical principles & applications


Embedded
Embedded
vs
vs
. foreground; real
. foreground; real
-
-
time
time
vs
vs
. offline;
. offline;
autonomous
autonomous
vs
vs
. supportive, interactive.
. supportive, interactive.
Automating tasks that have required
Automating tasks that have required
human intelligence …and beyond.
human intelligence …and beyond.
Evolution of
Evolution of
Subdisciplines
Subdisciplines
Artificial
Artificial
Intelligence
Intelligence
Metareasoning
Metareasoning
& Control
& Control
Qualitative
Qualitative
Reasoning
Reasoning
Real
Real
-
-
Time
Time
Reasoning
Reasoning
Knowledge
Knowledge
Representation
Representation
Diagnosis
Diagnosis
Planning
Planning
Temporal
Temporal
Reasoning
Reasoning
Logic
Logic
Rationality
Rationality
Cognitive
Cognitive
Science
Science
Decision
Decision
Making
Making
Uncertain
Uncertain
Reasoning
Reasoning
Machine
Machine
Learning &
Learning &
Datamining
Datamining
Distributed
Distributed
AI
AI
Search
Search
Constraint
Constraint
Satisfaction
Satisfaction
Evolution of Specialty Areas
Evolution of Specialty Areas
Artificial
Artificial
Intelligence
Intelligence
Artificial
Artificial
Intelligence
Intelligence
AI in Medicine
AI in Medicine
Sensor
Sensor
Networks
Networks
Computational
Computational
Neuroscience
Neuroscience
Search &
Search &
Retrievel
Retrievel
User
User
Modeling
Modeling
Intelligent
Intelligent
User Interfaces
User Interfaces
Speech
Speech
Recognition
Recognition
Vision
Vision
Natural
Natural
Language
Language
Satisfiability
Satisfiability
& Hardness
& Hardness
Ecommerce
Ecommerce
Ubiquitous
Ubiquitous
Computing
Computing
Robotics
Robotics
Game
Game
Playing
Playing
AI in Education
AI in Education
AI and Law
AI and Law
AI at Forefront of
AI at Forefront of
“Computational Revolution”
“Computational Revolution”
Industrial revolution
Industrial revolution
Computational revolution
Computational revolution
Acceleration of Methods,
Acceleration of Methods,
Applications, Infrastructure
Applications, Infrastructure
Learning and reasoning prowess
Learning and reasoning prowess
Computation
Computation
Memory
Memory
Sensors, connectivity, and content
Sensors, connectivity, and content
Expectations: AI in daily life
Expectations: AI in daily life


Communications
Communications


Time management
Time management


Health & safety
Health & safety


Education
Education


Goals, informational needs
Goals, informational needs


Games, recreation, activities
Games, recreation, activities


Products, purchases, marketing
Products, purchases, marketing


Opportunistic planning
Opportunistic planning


Augmentation of cognition
Augmentation of cognition
Expectations: AI in Science
Expectations: AI in Science


Automated discovery
Automated discovery


Design of experiments
Design of experiments


Triaging of resources
Triaging of resources


Interpretation of data
Interpretation of data


Probing complexity
Probing complexity


Biology, chemistry, medicine, climate
Biology, chemistry, medicine, climate
Key insights and technological
Key insights and technological
breakthroughs will be enabled through
breakthroughs will be enabled through
AI methods
AI methods
Expectation: AI and Infrastructure
Expectation: AI and Infrastructure


Transportation
Transportation


Commerce decision making
Commerce decision making


Agriculture
Agriculture


Engineering & architecture
Engineering & architecture


Power & conservation
Power & conservation
AI and the Consumer
AI and the Consumer


Evolving relationship with computation
Evolving relationship with computation


Sensing, reasoning & learning
Sensing, reasoning & learning


Personalized smart applications
Personalized smart applications


Products & services
Products & services


Challenges
Challenges
and opportunities
and opportunities
with data &
with data &
privacy
privacy
Evolving relationship with computation
Evolving relationship with computation
-
-
Memory, search,
Memory, search,
reminding
reminding
-
-
Predictive, proactive services
Predictive, proactive services
-
-
Triage of alerts & messages
Triage of alerts & messages
-
-
Rich visualizations
Rich visualizations
of inferences
of inferences
Evolving relationship with computation
Evolving relationship with computation
-
-
Mobile search & retrieval
Mobile search & retrieval
-
-
Handsfree
Handsfree
decision support
decision support
-
-
Natural interaction methods
Natural interaction methods
-
-
Presence forecasting
Presence forecasting
Evolving relationship with computation
Evolving relationship with computation
-
-
Facilitating & arranging
Facilitating & arranging
communications
communications
-
-
Reasoning about attention
Reasoning about attention
-
-
Shift of high
Shift of high
-
-
quality expertise and services to consumer
quality expertise and services to consumer
Evolving relationship with computation
Evolving relationship with computation
Evolving relationship with computation
Evolving relationship with computation
α
α
β
β
-
-
“Mixed
“Mixed
-
-
initiative” collaborations
initiative” collaborations
Evolving relationship with computation
Evolving relationship with computation
-
-
“Mixed
“Mixed
-
-
initiative” collaborations
initiative” collaborations
Trends in sensing, reasoning & learning
Trends in sensing, reasoning & learning


Large quantities of data via new sensing
Large quantities of data via new sensing
and online processes
and online processes
+
+


Advances in tractable machine learning
Advances in tractable machine learning
_____________________________________
_____________________________________


New applications and services
New applications and services
Privacy challenges & opportunities
Privacy challenges & opportunities


e.g.,
e.g.,
Structure search over variables
Structure search over variables


Generate and test feasible models
Generate and test feasible models


Build predictive models
Build predictive models
Advances in Machine Learning
Advances in Machine Learning
Causality
Playing an increasing role in the sciences
?
Hidden
variables
*
Proliferation of
Proliferation of
“Preference Machines”
“Preference Machines”


Example: Collaborative Filtering
Example: Collaborative Filtering
Pref1
Prefn
Products,
Content 1.. m
Proliferation of Preference Machines
Proliferation of Preference Machines
Proliferation of Preference Machines
Proliferation of Preference Machines
Proliferation of
Proliferation of
“Intention Machines”
“Intention Machines”


Example: Web search
Example: Web search
Query 1
Query n
Content +
Content -
Ad +
Ad -
Obs. 1
Obs. n
Services
Actions
Feedback
Proliferation of
Proliferation of
“Intention Machines”
“Intention Machines”


Beyond search… multiple services
Beyond search… multiple services
start
T
Δt
Traffic Advisor
Airport
25-min. delay at I-405 & I-90.
Suggest I-5 instead.
Seattle Center
Broad St. closed.
Suggest Denny Way instead.
On your way
On your way
to the airport?
to the airport?
Park here!
Park here!
$8/day.
$8/day.
Directions
Directions


Specials
Specials
Geocentric Services
Geocentric Services


Example: Predestination Project
Example: Predestination Project
Sensing, Learning, and Privacy
Sensing, Learning, and Privacy


Protected sensing & personalization
Protected sensing & personalization


Learning preferences about privacy
Learning preferences about privacy


Proactive inference & modeling
Proactive inference & modeling


Enriched parameters & policies for data
Enriched parameters & policies for data
sharing
sharing


Restricted usage
Restricted usage


Partial revelation
Partial revelation
Privacy challenges …and opportunities
Privacy challenges …and opportunities
Sensing, Learning, and Privacy
Sensing, Learning, and Privacy


Protected sensing & personalization
Protected sensing & personalization


Local sensing, learning, and
Local sensing, learning, and
inferencing
inferencing
within privacy of users’ systems
within privacy of users’ systems


Design of services that share nothing or
Design of services that share nothing or
little externally
little externally
Privacy challenges …and opportunities
Privacy challenges …and opportunities
Protected Sensing & Personalization
Protected Sensing & Personalization
Machine learning
Machine learning
Predictive model
Predictive model
Shroud of privacy
Shroud of privacy


Actions
Actions


Content
Content


Preferences
Preferences


Contextual attributes
Contextual attributes


Sensor data
Sensor data
3rd party content
3rd party content


Actions
Actions


Content
Content


Preferences
Preferences


Contextual attributes
Contextual attributes


Sensor data
Sensor data
Machine learning
Machine learning
Predictive model
Predictive model


Context
Context


Real
Real
-
-
time
time
sensor data
sensor data


Predictions
Predictions


Recommendations
Recommendations


Services
Services
Prebuilt
Prebuilt
model
model
(From subjects of study)
(From subjects of study)
Shroud of privacy
Shroud of privacy
Protected Sensing & Personalization
Protected Sensing & Personalization
3rd party content
3rd party content
Example: Geocentric Services
Example: Geocentric Services
Machine learning
Machine learning


Time, Day
Time, Day


Real
Real
-
-
time
time
GPS data
GPS data


Traffic alerts
Traffic alerts


Alternate routing
Alternate routing


Advert. & services
Advert. & services
Shroud of privacy
Shroud of privacy
Destination
Destination
Prediction
Prediction


Routes
Routes


End
End
-
-
points
points


Time of day
Time of day


Day of week
Day of week


GPS data
GPS data
Pretrained
Pretrained
Model
Model
Example: Personalized Web Search
Example: Personalized Web Search


Email
Email


Documents
Documents


Web activity
Web activity


Location
Location
.
.
.
.
.
.
Personal content &
Personal content &
activities store
activities store
Personalized
Personalized
result ranker
result ranker
Results from web search engine
Results from web search engine
Enriched Policies for Sharing
Enriched Policies for Sharing
Data & Inferences
Data & Inferences


E.g.,
E.g.,
Policies set by identity, time, and activity
Policies set by identity, time, and activity
“If I’m scheduled to see you within an hour…”
“If I’m scheduled to see you within an hour…”


You can access my exact location
You can access my exact location


Your calls will be routed to my cell phone
Your calls will be routed to my cell phone


Otherwise no location and no call routing.
Otherwise no location and no call routing.
Learning Preferences about
Learning Preferences about
Sharing What with Whom
Sharing What with Whom

Salesperson (live or web-based)
Your personal website/blog
Potential or confirmed competitor
Company newsletter
People in an upcoming meeting
People you want to impress (e.g. hire, date)
Corporate lawyer
People in a project for whom it is relevant
People who work for me
Other team members
People in extended family
Young child of yours
Trusted colleague/team member
Sibling
My manager
Adult child of yours
Parent/grandparent
Best friend outside of work
Spouse
Average
Standard Deviation
Transgression that is well understood to be
wrong (e.g. accessing pornographical
images on a work computer)
1.071.501.041.161.171.101.321.231.261.201.341.631.571.611.601.791.822.172.55
1.48
0.40
All of my email content
1.071.171.111.201.301.241.641.501.471.431.661.941.731.682.001.571.892.003.101.620.47
Credit card number
1.571.061.001.001.101.101.401.171.401.271.591.591.431.751.472.432.321.834.38
1.62
0.78
Social Security Number
1.301.061.141.081.171.381.801.311.601.472.072.181.802.612.673.143.412.174.451.990.92
A potential transgression -- action not
universally understood as wrong, more in
a gray area (e.g. using your work computer
for church activities)
1.101.611.141.361.331.411.681.671.651.801.972.122.302.462.402.932.682.903.692.010.70
Outside income
1.481.371.191.211.341.461.791.431.631.522.252.251.762.591.833.143.262.974.36
2.04
0.86
Salary
1.411.281.181.161.171.462.161.171.391.482.072.062.142.524.002.623.112.664.392.080.96
Large personal failure (e.g. fired from
previous job)
1.301.501.251.241.481.461.881.621.581.802.212.592.332.792.473.003.293.334.212.170.85
Personal behavior I feel bad about (e.g.
spoke sharply to a colleague)
1.171.561.361.561.701.622.002.002.052.272.282.352.732.682.772.792.893.133.792.250.68
Buddy list (who’s on my list)
1.321.251.551.611.621.701.721.671.932.052.573.212.193.052.333.223.193.103.50
2.25
0.75
N
on-work related websites I've looked at a
t
1.241.671.571.561.801.831.921.931.792.202.642.762.602.702.603.082.743.173.712.290.66
Recent history of status (looking for trends
)
1.571.411.621.852.052.051.902.142.132.272.322.872.592.602.732.802.622.683.38
2.29
0.51
History of my job performance scores
1.171.291.391.681.532.032.241.801.741.832.282.592.402.964.173.003.072.974.102.330.87
Opinions I have about other people
(assume in digital form)
1.431.671.461.561.731.791.561.901.892.032.792.532.503.112.373.433.293.503.972.340.79
My application to another job/school
1.331.611.251.441.431.761.521.631.581.672.932.712.303.501.973.713.793.804.62
2.34
1.06
Access to my computer with personal
assurance that they won't look at anything
1.201.191.291.321.571.691.881.932.532.402.542.713.232.813.373.153.223.143.82
2.37
0.84
Small personal failure (e.g. project
missteps that led to failure)
1.301.501.291.481.732.002.082.072.472.432.592.763.033.073.333.213.363.474.03
2.48
0.82
What email groups I belong to (external to
t
1.501.711.591.922.212.212.082.322.292.282.703.002.623.112.623.273.423.594.112.560.71
Record/summary of database/sharepoint in
t
1.291.291.761.912.732.602.593.003.173.312.592.873.852.903.962.902.952.913.77
2.76
0.76
Preferences (politics, religion, associates,
etc.) (assume in digital form)
1.702.331.791.721.772.381.842.102.322.333.663.652.904.002.574.074.114.274.52
2.84
1.00
Work-related documents I've accessed
1.291.181.792.042.832.762.923.033.423.532.562.944.202.714.232.913.003.083.762.850.84
Your health status
1.501.781.962.122.172.242.362.302.502.573.413.473.233.573.103.933.934.034.55
2.88
0.88
Specific calendar entries
1.521.782.002.173.072.592.543.173.163.173.083.403.483.363.623.383.403.564.182.980.70
Pregnancy status
1.872.112.002.072.272.292.402.332.222.733.734.133.533.793.403.834.004.204.64
3.03
0.92
Work in progress
1.321.671.672.163.213.072.963.763.673.722.893.504.212.924.313.173.083.143.783.060.84
Desktop video conference number
1.291.832.503.173.833.203.504.333.333.833.002.753.713.204.292.003.333.173.83
3.16
0.81
Work-related websites I've looked at
1.411.722.002.523.203.033.163.603.263.703.393.534.133.334.233.623.373.484.003.190.77
Current location
1.631.942.112.253.073.072.763.733.263.373.414.063.773.753.703.643.793.834.31
3.23
0.76
Current status (on line, “busy”) from IM
1.781.822.332.243.183.052.583.453.473.453.774.073.554.053.503.783.863.914.193.270.76
Past finished papers, products
1.552.612.433.243.633.593.643.933.794.133.253.474.433.264.603.313.483.484.00
3.46
0.71
What email groups I belong to (internal to
t
1.451.812.433.323.693.523.763.974.114.333.193.384.633.524.633.583.643.784.083.520.84
Small personal success (e.g. project
chosen to demo)
1.872.502.893.163.233.453.243.473.653.533.724.124.004.074.304.294.254.304.763.620.71
When available (on a shared calendar)
1.672.002.302.614.103.393.294.244.214.073.694.134.384.124.384.004.044.224.59
3.65
0.88
Home phone number
1.551.472.072.252.763.323.173.313.443.594.755.004.454.964.595.004.964.935.003.711.25



Variances, differences, similarities
Variances, differences, similarities
Summary
Summary


Efforts in artificial intelligence leading to
Efforts in artificial intelligence leading to
valuable new services and applications
valuable new services and applications


Innovations in privacy will be enabling
Innovations in privacy will be enabling


AI methods for learning & reasoning
AI methods for learning & reasoning
promises to play a key role in these
promises to play a key role in these
innovations
innovations