Presentation - TIM101, Winter 16, Section 01

Download Report

Transcript Presentation - TIM101, Winter 16, Section 01

ARTIFICIAL INTELLIGENCE
2016
Ira Pohl
TM Feb 25, 2016
Talk
•
•
•
•
•
•
What is AI?
A brief history
Use in Industry
Use in eLearning
UCSC work
Future
What is AI?
• AI – a science/engineering of intelligence
– In analogy to aeronautical engineering/flying
• AI – model of “human/cognitive” system
– Is done as a theory of human intelligence
Acting humanly: Turing Test
• Turing (1950) "Computing machinery and
intelligence":
• Operational test for intelligent behavior:
the Imitation Game
•
• Predicted that by 2000, a machine might
have a 30% chance of fooling a lay person
for 5 minutes
• Anticipated all major arguments against AI
• Loebner Prize
Thinking humanly: cognitive
modeling
• 1960s "cognitive revolution": informationprocessing psychology Newell and Simon GPS
•
• Requires scientific theories of internal activities
of the brain
•
• -- How to validate? Requires
1) Predicting and testing behavior of human subjects
(top-down)
or 2) Direct identification from neurological data
(bottom-up)
Thinking rationally: "laws of
thought"
•
•
•
•
•
Several Greek schools developed
various forms of logic: notation and rules
of derivation for thoughts; may or may
not have proceeded to the idea of
mechanization
Direct line through mathematics and
philosophy to modern AI
-Leibnitz,Boole
Kleene, Church, Turing – McCarthy,
Robinson
AI prehistory
• Philosophy
• Mathematics
• Economics/OR
• Neuroscience
• Psychology
• Computer Science
• Linguistics
Logic, methods of reasoning, mind as physical
system foundations of learning, language,
rationality
Formal representation and proof algorithms,
computation, (un)decidability, (in)tractability,
probability,bayes
utility, decision theory
physical substrate for mental activity
phenomena of perception and motor control,
experimental techniques
building fast computers, algorithms
knowledge representation, grammar
Abridged history of AI
•
•
•
1943
1950
1950s
•
•
•
•
•
•
•
•
•
•
1956
1965
1980
1995
1997
2003
2011
2011
2014
2015
McCulloch & Pitts: Boolean circuit model of brain
Turing's "Computing Machinery and Intelligence"
Early AI programs, including Samuel's checkers
program, Newell & Simon's Logic Theorist,
Dartmouth meeting: "Artificial Intelligence" adopted
Robinson's complete algorithm for logical reasoning
AI industry –Symbolics & Knowledge Based Systems
The emergence of intelligent agents
Kasparov loses to Deep Blue
iRobot – Roomba, Pacbot510
Google Car –self driving 300,000 miles
MOOCs – autograded classes, edX software
Hawking “spell the end of the Human race” –BBC
Google AlphaGo beats European champion-using learning
Autonomous Car
Riba Description
• RIBA — short for “Robot for Interactive Body
Assistance” — was developed by researchers at
Japan’s Institute of Physical and Chemical
Research (RIKEN) and Tokai Rubber Industries,
Ltd. (TRI). Designed primarily to assist nurses by
lifting patients in and out of their beds and
wheelchairs (as well as on and off the toilet), the
180-kilogram (400-lb) robot can safely pick up
and carry people weighing as much as 61
kilograms (135 lbs).
RIBA -2008 Nurse Robot
Robot arm –in use in solar panel
factory - NYT 2012
Skilled Work without the Worker
NYT_ Aug 18, 2012 by John Markoff
• Take the cavernous solar-panel factory run by
Flextronics in Milpitas, south of San Francisco. A large
banner proudly proclaims “Bringing Jobs &
Manufacturing Back to California!”
• Yet in the state-of-the-art plant, where the assembly line
runs 24 hours a day, seven days a week, there are
robots everywhere and few human workers. All of the
heavy lifting and almost all of the precise work is done by
robots that string together solar cells and seal them
under glass. The human workers do things like trimming
excess material, threading wires and screwing a handful
of fasteners into a simple frame for each panel.
GO- Learning/MC
• Go Program is now (Google Go program
European Champ) about to surpass
human ability
• How: learning, big data, MC computation;
• Distinctly not “human” cognition
• Implication: Machine Intelligence is
surpassing human ability in ways that
humans cannot “understand”
State of the art
• Deep Blue defeated the reigning world chess champion Garry
Kasparov in 1997 -current best 3100+ vs 2800 for humans
• Proved a mathematical conjecture (Robbins conjecture) unsolved for
decades 1996 by EQP
• No hands across America (driving autonomously 98% of the time
from Pittsburgh to San Diego) in 1995 -2005 132mile unknown
terrain-completely autonomous
• During the 1991 Gulf War, US forces deployed an AI logistics
planning and scheduling program that involved up to 50,000
vehicles, cargo, and people
• Roomba 2002 > Now pet hair version, mopping, eaves, windows
• RIBA 2007
• IBM WATSON- 2011 – wins Jeopardy against all time champs
• Google robot Car 2012 300,000 miles Now > 1,000,000
UCSC AI-and related
•
•
•
•
•
•
Pohl- Heuristic Search, Should Robots have Nuclear ARMs
Getoor- Big Data Inference
Mateas-Games Intelligent dialog for characters,Façade Interactive drama
Helmbold, Warmuth- effective learning algorithms
McDowell,Doshay- Go program(machine champ in 2006)
Elkaim- Autonomous Sailboat > now a degree program
Achievements
• LISP, Time Sharing
• Games – early space war games
• Intellectual Games – mastery in Chess,
checkers, Hex, othello, backgammon, scrabbleand now GO and Poker
• MACSYMA –Mathematica, MATLAB
• DENDRAL(chemistry, medicine … experts)
• Robotics
• Speech and Handwriting recognition (SIRI)
Large trends
• Big data and learning- google can
translate effectively between natural
languages;
• Big Computation- monte carlo simulations
with simple metrics can outperform
humans with deep knowledge
MOOCs and AI
• One last story
• Educating Humans as an AI problem
• Why? Difficult and Expensive
Events
• In the 1960s, Patrick Suppes and Richard
C. Atkinson (the future president of
the University of California) conducted
experiments in using computers to teach
math and reading to schoolchildren in
the Palo Alto area.
World University for free!
• Norvig –Thune AI class in 2011 has 160000
attendees – autograding
• Leads to coursera, udacity, edX
• Together with Kahnacademy website- these
resources are reshaping education
• Here at UCSC Baumgarten-Kenez class
Holocaust – 2013 Coursera: 18,000 students
• Pohl – C++ for C Programmers -2013: 60,000
The larger Universe or maybe the
Global University
• Coursera: Started at Stanford
• Principals : Daphne Koller, Andrew Ng
• Members: 16 AAU Universities
– Princeton, Cal Tech, Duke, Univ of Penn
– UW, Michigan, UVA
– Edinburgh, U of Toronto
– UCSF
• Classes: 124+ over all subjects 2013
2014
• Coursera 106 institutions worldwide
• 500 plus class curriculum
• Over 5 million unique names
• Udacity: Georgia Tech $6600 Masters in
CS(partner) ATT
• 2016 UCSC Holocaust Class, C++ for C
Programmers - on demand/cohorts
Yet to come
• Will humans need to work?
• Will their be a “singularity”
• If work and intelligence do not define us;
what does?
Baxter - $3.00 per hour
• Rodney Brooks : general purpose
trainable indutrial robot: cost $22,000.
Packbot 510