Transcript Cognitive Agents Tech - Seidenberg School of Computer Science

Team 2





Maria Azua
Dwight Bygrave
Jonathan Leet
Rick Rodin
Evgeni Sadovski
February 16, 2010
What is the Problem?
 Economic pressures are demanding more automation and
efficient systems
 Massive amount of data and escalating regulatory compliance
laws are requiring more intelligent systems that can –
 Understand highly contextual information

Adjust behavior to context
 Can handle ambiguous situations
 Handle imprecise or implicit information
 Cloud computing is commoditizing compute power. New low
cost compute power is enabling “electronic reasoning”
unaffordable a couple of years ago.
Information Overload
Market Forces – The Perfect Storm
Transition to Digital
 New delivery channels (web, mobile)
Consumer Dynamics
 Push  Pull
 Content type convergence –
text, image, audio, video
Business
Model
Innovation
 “Born Digital” business models
 Emerging competencies -- meta data,
interactive experiences, multi-channel
distribution, analytics.
 Content  Experiences
 Self - Personalization
 Social Consumption
 Cross Channel relationship management
- bundling
Expanding Impact of Technology
 Digital Supply Chain -- workflow automation
 Analytics -- Optimzation, Event management and Prediction
 Resource optimization, variability and seasonality
What is Cognition?
 The Oxford dictionary defines cognition as
knowing or perceiving
 Cognition in Artificial intelligence – Extends the
concept as an interdisciplinary study of the general
principles of intelligence through a synthetic
methodology termed learning by understanding[1].
1. Rolf Pheiger, C.S., Understanding Intelligence. 1999, Cambridge, MA: MIT Press. 720.
What is a Cognitive Agent?
Cognitive Agents – They not just learn by trial and error…but they
understand and set goals by inferring relationships using many data
sources with minimum human intervention.
They utilize:
Uses cases
Taxonomy and relationship rules that enable sensitivity of
highly contextual context and situations
Artificial Neural Networks to evaluate outcomes
Electronic reasoning simulation which consist of three key
components: [2]
1.
2.
3.
problem solving (planning);
Comprehension (story associated with the understanding)
Learning (remembering the outcome of use case)
2. Cox, M.T., Perpetual Self-Aware Cognitive Agents, in Intelligent Distributed Computing. 2007, American
Cognitive Cycle
Observe
the situation
Gather information
about user activities
Act in
accordance
to the plan
Formulate scenarios
and define use cases
Apply and test
application
Develop data points
and algorithms
Create a plan
to achieve
the goal
Form a
goal or
appropriate
behavior
Robots using Cognitive Agents
 Robots learn social cues via Cognitive Agents- Robots as a
Social Technology - research by Cynthia Breazeal
 Self Aware Robots - The learn their environment,
understand themselves and even self-replicate – research by
Hob Lipson
 Robotic Comedian - it gathers audience feedback to tune its
act – research from Heather Knight
PRODOGY high level design
INTRO Architecture
Context Awareness
Watson – Taxonomy & Relationships
Search Scenario
Cloud Scenario
Dynamic Cloud Images
Cloud user customize their images 36% of the time
Cloud Scenario
Cloud Adoption is Limited by Trust
Social Networks Could be used to
Augment Cognition
Compliance Scenario
Compliance Scenario
Compliance Scenario
Mobile Scenario
Conclusion
 Enterprises embrace cloud computing design patterns for
solving problems that they otherwise would shy away from
due to infrastructure constraints.
 Applying cognitive agent research and principles to existing
distributed businesses, real-world automation can be enabled.
 Social software should be embraces not only as an enabler of
collaboration but as the source of implicit and explicit
connections. Being able to mine and understand these
connections will result in smarter systems.
 Finally, one area of concern is employee privacy. If taken too
far cognitive agents could potentially appear “big brother” in
nature.
References
1.
Azua, M., The social factor : innovate, ignite, and win through mass collaboration and social networking. 2010, Upper
Saddle River, NJ: IBM Press/Pearson. 247 p.
2.
Malik, O. Wholesale Internet Bandwidth Prices Keep Falling: . 2008 [cited 2010 2010-12-04 15:21:25]; Available from:
http://gigaom.com/2008/10/07/wholesale-internet-bandwidth-prices-keep-falling/.
3.
Pankaj Deep Kaur, I.C., Unfolding the Distributed Computing Paradigms, in 2010 International Conference on Advances in
Computer Engineering. 2010: Bangalore, India. p. 339 - 342.
4.
Rolf Pheiger, C.S., Understanding Intelligence. 1999, Cambridge, MA: MIT Press. 720.
5.
Cox, M.T., Perpetual Self-Aware Cognitive Agents, in Intelligent Distributed Computing. 2007, American Association for
Artificial Intelligence (www.aaai.org).
6.
Caprarescu, B.A., Robustness and scalability: a dual challenge for autonomic architectures, in Proceedings of the Fourth
European Conference on Software Architecture: Companion Volume. 2010, ACM: Copenhagen, Denmark. p. 22-26.
7.
Baral, C., et al., Using answer set programming to model multi-agent scenarios involving agents' knowledge about other's
knowledge, in Proceedings of the 9th International Conference on Autonomous Agents and Multiagent Systems: volume 1 Volume 1. 2010, International Foundation for Autonomous Agents and Multiagent Systems: Toronto, Canada. p.
259-266.
8.
Veloso, M. PRODIGY Project Home Page. 2010 Dec 12, 2010 [cited 2010 Dec 12, ]; Available from:
http://www.cs.cmu.edu/afs/cs.cmu.edu/project/prodigy/Web/prodigy-home.html.
9.
IBM. Watson - A System Designed for Answers [Online Multimedia on IBM.com site] 2011 [cited 2011 Feb 1]; Watson
cognitive agent competes on Jeopardy ]. Available from: URL: http://www.ibm.com/innovation/us/watson/.
10.
Yen, N.Y., T.K. Shih, and L.R. Chao, Adaptive learning resources search mechanism, in Proceedings of the second ACM
international workshop on Multimedia technologies for distance leaning. 2010, ACM: Firenze, Italy. p. 7-12.