7. From Early AI to Cognitive Systems
AARON - The First Artificial
Intelligence Creative Artist
(Harold Cohen, UCSD)
1973–present
The Aaron system
composes and physically
paints novel art work. It is a
rule-based expert system
using a declarative
language. 7
AIBM Researcher Gerald
Tesauro (1994) developed a
self-teaching backgammon
program called TD-Gammon.
Starting from a random initial
strategy, and learning its
strategy almost entirely from
self-play, TD-Gammon
achieved a human world-
champion level of
performance.
On May 11, 1997, IBM’s
Deep Blue (manned by co-
creator Murray Campbell
above) beat the world
chess champion Garry
Kasparov after a six-game
match: two wins for IBM,
one for the champion and
three draws.
Watson competed against
Jeopardy’s two all-time
greatest champions. This
match appeared on
television in February of
2011. Watson won the
match, outscoring both
opponents combined.
10. Big Data: This is just the beginning
You are here
Sensors &
Devices
Social
Media
VoIP
Enterprise
Data
9000
8000
7000
6000
5000
4000
3000
11. Knowledge & Interaction
11
Where was Einstein born?
One day, from among his city views of Ulm, Otto
chose a watercolor to send to Albert Einstein as a
remembrance of Einstein´s birthplace.
12. Knowledge & Interaction
12
Welch ran this…
If leadership is an art then surely Jack Welch has
proved himself a master painter during his tenure at GE.
13. A new era of computing…
Cognitive Systems learn and interact naturally
with people to amplify what either humans or
machines could do on their own. They help
us solve problems by penetrating the
complexity of Big Data.
Programmable Systems
» Structured data (local)
» Deterministic Applications
» Search Oriented
» Small Data
» Machine Language
» Systems of records
Cognitive Systems
» Structured & unstructured (global)
» Probabilistic Applications
» Discovery Oriented
» Big Data
» Natural Language
» Systems of engagement
Tabulating Systems Era
Programmable
System Era
Cognitive Systems Era
14. Knowledge & Interaction
14
Where was Einstein born?
One day, from among his city views of Ulm, Otto
chose a watercolor to send to Albert Einstein as a
remembrance of Einstein´s birthplace.
Person Born in
A. Einstein Ulm
15. Knowledge & Interaction
15
Welch ran this…
If leadership is an art then surely Jack Welch has
proved himself a master painter during his tenure at GE.
Person Organization
J. Welsh GE
16. Recent advances in AI are enabling a new
generation of scenarios
16
Earlier AI systems stalled due to…
» Reliance on a large number of manually
designed rules for specific purposes
» Lack of sufficient computational power
» Trouble scaling to complexities of real
applications
Recent trends are driving change…
» Probability and statistics a fundamental
formalism for AI – probabilistic reasoning,
graphical models, and HMMs
» More powerful and sophisticated machine
learning algorithms - DBN
» The availability of huge computing power
and vast amounts of data
» Individuals overwhelmed by information
overload in private and professional lives
17. Computers and Brain: Different &
Complementary
~5 GHz, sequential, linear, clocked
Separates memory,
computation, communication
100 W/cm2
~1 year / rapidly evolving architecture
10 Hz, parallel, high fanout, event-driven
Integrates memory,
computation, communication
10 mW/cm2
~106 years / pace of architecture change
18. 18
The Watson Developer Cloud — A set of
purpose-built, REST APIs with AI built in
http://www.ibmwatson.com/developercloud
19. Cognitive Systems
Building Cognitive Systems requires
» A deep understanding of users’ problems “in the wild”
» Soft sciences: cognitive, psychology, neuroscience
» Engineering: machine learning technologies, sensory components, analytics,
interaction
» Application Development & Deployment: Easy to build and deliver
How can we build a Cognitive System
» We know something about each discipline
» We are faced with sort of an “inverse problem”
Two observations:
» A small number of perspectives is not enough
» Deeper relationship between these perspectives needed 19
20. An Engineering Approach
» Don’t ignore Symbolic knowledge – it plays a vital role
» Use an “atomic” structure that can learn at the highest
possible level – Deep Neural Networks
» Be task oriented and use humans and automated
techniques to build the largest training data
» Use Machine Learning to optimize the components of the
system
» Perform accuracy tests regularly and iterate on all above
steps
» System of Systems still a major challenge
20
Neuro-
science
Nano-
technology
Super-
computing
Cognitive
Computing
Cognitive
Science
Machine
Learning
Quantum
Computing