The Thinking Computer
An Introduction to Artiﬁcial Intelligence
• Deﬁne artiﬁcial intelligence
• Deﬁne symbolic representation and give examples of simple
representations or symbols
• List the approaches to AI and explain how they work in terms of search
(what is being searched, what are we looking for?)
• Motivate AI research with a few examples of beneﬁts that you think are
• List some applications of AI today
• Be able to provide a list of limitations that AI researchers currently face
• Discuss the risks and the effect Hollywood has on our perceptions
What is AI?
• A branch of science concerned with helping
machines solve problems in more human-like ways
• That is, we translate “human intelligence” into
computer programs, etc.
• Notice I said “science” above... AI includes
• Computer Science, Math, Psychology, Linguistics,
Biology, Cognitive Science, Education, Kinesiology,
Anthropology, Economics... phew!
What is AI?
• There are many branches to AI research, some
more human-oriented than others
• Many set out to solve domain-speciﬁc problems...
small problems that might beneﬁt from a human-
• Others want to understand how we think by
modeling it in computers
• Still others wish to determine once and for all if AI
is even possible, and what its limitations are...
• In order for us to teach a computer to think like
us, we ﬁrst have to, well, teach it to think!
• Note that just what intelligent thought is is a
point of some contention... you’ll get to discuss
this with your labmates this week
• We can say something about thought, though... it
• That is, to solve a problem we ﬁrst have to
• For us, representation just involves thinking about a
problem, and somehow, magically, it’s represented in
• For a computer, though, we need to be able to give
it the information it needs to solve a problem.
• This information is called the representation of
• Examples include a small virtual world, a word
problem in math, a list of paths between cities,
• There are two main ways to represent information:
symbolically and “non”-symbolically (implicit)
• The latter is the stuff of Cognitive Systems
• We’ll just concern ourselves with symbolic
• A symbolic representation means that each
concept of a problem is represented as a symbol
• This symbol means two: “2”
• This symbol means cat: “cat”
• This symbol means that cats are mammals:
• This symbol means that cats have four legs: “legs(cat,
• How might you represent a room with four
• If the computer “knows” that cats have four legs,
then it can use this information in solving
• E.g. Put socks on Socks
• If you have a cat named Socks, and you want
the computer to put socks on it, it will know
that it needs 4 socks.
Remember I said a lot of AI was search?
• Search and Optimization
• Constraint Satisfaction
• Constrain the “world” to represent the problem,
then search the possibilities to ﬁnd a solution
• Function Optimization
• Finding the best set of parameters for a function (i.e.
providing the best tools for the job) E.g. Hill-climbing
• Finding the best sequence of actions
• Machine Learning
• Neural networks
• Decision Tree learning
• Data mining
• Bayesian networks
• Inductive programming...
• Pattern Recognition
• Natural Language Processing / Computational
• Artiﬁcial Life
• Computers are very, very fast at certain
applications and they don’t get tired!
• If we can utilize this speed in intelligent
applications, we can use computers as our
• In situations where it is too dangerous for a
• For monotonous or lengthy tasks
• Tasks where extra care is required
• We can also understand ourselves better!
• Medical diagnosis
• Intelligent data storage and retrieval
• Personal digital secretaries
• Large-scale data analysis
• Assembly lines
• Behaviour analysis (e.g. banks, hospitals, etc)
• Rescue operations
• That was just a brain dump.. there are lots more!
• Story time!
• We don’t know what intelligence is!
• Right now the focus is on domain-speciﬁc, lucrative
applications in industry, etc
• But these are not focusing on the greater science of AI in
• Sometimes called “narrow intelligence”
• Mixing up our goals
• Strong versus weak AI / AI versus AHI (artiﬁcial human
• Overstating our progress
• What are the ethical concerns?
• Are we biased by Hollywood? (Skynet, anyone?) What are
the real concerns?
• Here are some:
• Reduction in jobs for humans
• Sensitive activities being performed by unfeeling machines
• Machines programmed by humans-- who sure as heck
• Losing touch with humanity... or worse?