Artificial Intelligence

Introduction

Instructors: Dr. Hamed Abdelhaq

(slides adapted from Anca Dragan/ University of California, Berkeley)
 Prerequisites
o Comfortable programming in general-purpose
programming language – we’ll do assignments in Python
o Some knowledge of algorithmic concepts such as running
times of algorithms
o Ideally, some familiarity with probability (we will go over
this from the beginning but we will cover the basics only
briefly)
o Not scared of mathematics; ideally, some background in
discrete mathematics, able to do simple mathematical
proofs
From Sci-Fi to reality in AI?
 Some highly visible recent AI successes in games

DeepMind
Watson defeats AlphaGo defeats Go
achieves human-
Jeopardy champions champion (2016)
level performance
(2011)
on many Atari
games (2015)
 AI
o General-purpose AI like the robots of science fiction is incredibly hard
o Human brain appears to have lots of special and general
functions, integrated in some amazing way that we really
do not understand
o Special-purpose AI is more doable (nontrivial)
E.g., chess/Go playing programs, logistics planning,
automated translation, speech and image recognition, web
search, data mining, medical diagnosis, keeping a car on
the road, … … … …
 Today
o What is artificial intelligence?

o Where did it come from/What

can AI do?
o What should we and shouldn’t
we worry about?
o What is this course?
 What is AI?
The science of making machines that:

Think like Think rationally

people

Act like people Act rationally

 Rational Decisions
We’ll use the term rational in a very specific, technical way:
▪ Rational: maximally achieving pre-defined goals
▪ Rationality only concerns what decisions are made
(not the thought process behind them)
▪ Goals are expressed in terms of the utility of outcomes
▪ Being rational means maximizing your expected utility

A better title for this course would be:

Computational Rationality
Maximize Your
Expected Utility
Maximize Your
Expected Utility
Maximize Your
Expected Utility
Maximize Your
Expected Utility
 What About the Brain?
▪ Brains (human minds) are very
good at making rational decisions,
but not perfect
▪ Brains aren’t as modular as
software, so hard to reverse
engineer!
▪ Lessons learned from the brain:
memory and simulation are key
to decision making
 Designing Rational Agents

o An agent is an entity that perceives and acts.

o A rational agent selects actions that maximize
its (expected) utility.
o Characteristics of the percepts, environment,
and action space dictate techniques for
selecting rational actions
o This course is about:
o General AI techniques for a variety of
problem types Sensors
Percepts
o Learning to recognize when and how a new
problem can be solved with an existing ?
technique
Actuators
Actions
 Pac-Man as an Agent

Agent Environment
Sensors Percepts
?
Actuators Actions

Pac-Man is a registered trademark of Namco-Bandai Games, used here for educational purposes Demo1: pacman-l1.mp4
 AI
Machine Learning
[learning decisions;
sometimes independent]

Robots Rational Human-AI

[physically Agents
embodied] Interaction
[decisions]

NLP
Computer
Vision
 Textbook
o Russell & Norvig, AI: A Modern Approach,
3rd Ed.

o Warning: our presentation does not

necessarily follow the presentation in the
book.
A (Short) History of AI
 A (Short) History of AI
o 1940-1950: Early days
o 1943: McCulloch & Pitts: Boolean circuit model of brain
o 1950: Turing's “Computing Machinery and Intelligence”
o 1950—70: Excitement: Look, Ma, no hands!
o 1950s: Early AI programs, including Samuel's checkers
program, Newell & Simon's Logic Theorist, Gelernter's
Geometry Engine
o 1956: Dartmouth meeting: “Artificial Intelligence” adopted
o 1965: Robinson's complete algorithm for logical reasoning
o 1970—90: Knowledge-based approaches
o 1969—79: Early development of knowledge-based systems
o 1980—88: Expert systems industry booms
o 1988—93: Expert systems industry busts: “AI Winter”
o 1990—: Statistical approaches
o Resurgence of probability, focus on uncertainty
o General increase in technical depth
o Agents and learning systems… “AI Spring”?

o 2000—: Where are we now?

 What Can AI Do?
Which of the following can be done at present?

o Play a decent game of table tennis? ✓

o Play a decent game of Jeopardy? ✓
o Drive safely along a curving mountain road? ✓
o Drive safely along Rafedia street? 
o Buy a week's worth of groceries on the web? ✓
o Converse successfully with another person for an hour? 
o Perform a surgical operation? 
o ✓
Translate spoken Chinese into spoken English in real time?
o Compose music ✓
 Lessons from AI research
o Clearly-defined tasks that we think require intelligence and education from humans
tend to be doable for AI techniques
o Playing chess, drawing logical inferences from clearly-stated facts, performing probability
calculations in well-defined environments, …
o Although, scalability can be a significant issue
o Complex, messy, ambiguous tasks that come naturally to humans (in some cases
other animals) are much harder…
o … though recent years have seen remarkable progress, especially in machine
learning for narrow domains
o Image recognition, speech recognition, reinforcement learning in computer games, self-
driving cars
o AI systems still lack: broad understanding of the world, common sense,
ability to learn from very few examples, truly out-of-the-box creativity…
 Intelligence sources

o Intelligence might be achieved from

o Computation
o Fast search (e.g. Solving a Rubik’s cube)
o Constraint satisfaction
o Finding a schedule that will allow you to graduate (reasoning backwards from the goal)

o Data
o Baye’s Net
o Decision Theory
o Machine learning

Slides adapted from AI @ Duke

University and Berkeley
 Applications
1. NLP
o Speech Recognition
o Text-to-speech
o Question Answering
o Machine translation etc
2. Computer vision
o Face Detection and Recognition
o Semantic Scene Segmentation
3. Robotics
4. Game Playing

Slides adapted from AI @ Duke

University and Berkeley
 Computer Vision

Karpathy & Fei-Fei, 2015; Donahue et al., 2015; Xu et al, 2015; many more
 Game Agents
o Classic Moment: May, '97: Deep Blue vs. Kasparov
o First match won against world champion
o “Intelligent creative” play

o 1996: Kasparov Beats Deep Blue

“I could feel --- I could smell --- a new kind of intelligence across the table.”
o 1997: Deep Blue Beats Kasparov
“Deep Blue hasn't proven anything.”

Text from Bart Selman, image from IBM’s Deep Blue pages
 Game Agents
o Reinforcement learning
Photo: Google / Getty Images
 Simulated Agents

[Schulman, Moritz, Levine, Jordan, Abbeel, ICLR 2016]

 Game Agents
o Reinforcement learning

Pong Enduro
Robots
Human-AI Interaction
 Nonexhaustive list of AI publications

o General AI conferences: IJCAI, AAAI, ECAI

o Reasoning under uncertainty: UAI
o Machine learning: ICML, NIPS
o Multiagent systems: AAMAS
o Vision: ICCV, CVPR
o Some journals: Artificial Intelligence, Journal of
AI Research, Machine Learning, Journal of ML
Research, Journal of Autonomous Agents and
Multi Agent Systems
o AI Magazine
Slides adapted from AI @ Duke
University and Berkeley