Transcript Artificial Intelligence

Artificial Intelligence
By Michelle Witcofsky
And Evan Flanagan
Definition
The science that automates intelligent
behaviors. It is a system that thinks and
acts like humans, as in rationally and
intelligently. It is the study of mental
faculties through the use of computational
methods. The use of computers is to do
symbolic reasoning, pattern recognition,
learning, and some forms of inference.
<source>
Intelligence and Machines
• Agents: devices that respond to stimuli from their
environment
• Most agents have sensors that receive data
• Also have actuators that affect their environment
• Goal of A.I.: To build agents that act intelligently
• Two types of knowledge:
Procedural knowledge – learning “how”
Declarative knowledge – learning “what”
Performance v. Simulation
• 2 different approaches to research
• Performance oriented
develop A.I. to enhance performance
• Simulation oriented
develop A.I. to mimic how humans
respond in certain situations
• Turing Test
to determine difference between human or
machine response
Understanding Images
• Weak A.I.: belief that machines can be
programmed to exhibit intelligent behavior
(accepted by a wide audience)
• Strong A.I.: belief that machines can be
programmed to possess intelligence, and
even consciousness; widely debated
because they are internal human
characteristics that cannot be identified
directly
Reasoning
• Production Systems
• (1) Collection of states – where states are
situations that might occur in runtime
• (2) Collection of productions – rules or
moves of operation that are used to
produce results during runtime
• (3) Control system – logic that solves the
problem of moving from start to finish, or
goal state
Artificial Neural Networks
• Basic properties
• Processing units – like neurons in
organisms that transmit brain messages
• Output of 1 or 0, combinations of outputs
to create more complicated messages
• Idea of inhibiting or exciting output
mechanisms
Associative Memory
• Retrieval of information associated with
the information being used in a situation
• Constructing machines with this has been
a research goal for many years; could lead
to highly developed A.I.
• One of the main principles of the idea of
Artificial Intelligence
Genetic Algorithms
• Key concept in Artificial Neural Networking
• Research area that seeks to apply
understanding of natural evolution to the
task of solving a problem
• The trick is making a computer’s problemsolving algorithm mirror that of how we do
it as humans
Evolutionary Programming
• Main part of creating genetic algorithms
• Developing programs by allowing them to evolve
on their own, i.e., not just explicitly typing
everything necessary for the program to run
• Computer/compiler makes inferences about
what should be done in a given situation that is
not explained in the code
-field is still in its infancy
-only obtained in very basic examples
Other Areas of Research
• Language Processing:
– Requires several levels of analysis
– Syntactic Analysis: breaks sentences down
into parts of speech, subject/clause, etc
– Semantic Analysis: identifies tone and
meaning the role of each word brings to a
sentence
– Contextual Analysis: Brought into
understanding process, an understanding of
what is actually meant by the text.
Other Areas of Research
• Information Retrieval/Extraction: The goal
of Language Processing is to allow a
machine to interpret a block of text by
isolating the commands it contains, thus
understanding what the user wants it to do
– Example: We write “emacs prog.c &” to open
a file in Linux, but what if the computer could
understand us saying “Open prog.c in the
emacs editing program.” The machine’s ability
to extract a command from that sentence is a
specific goal in Artificial Intelligence
Other Areas of Research
• Robotics
– Aimed directly at
creating machines that
behave intelligently
– Main focus of popculture view on
Artificial Intelligence
– Most recognizable
symbol of A.I. research
Other Areas of Research
• Database Systems
– Knowing when to apply certain shared
information
– Like a catalog of useful data, not unlike the
function of the human brain, and also the
intelligence to know which data is applicable
in a given situation without being told in code
– “closed-world database” – contains all true
facts relating to a certain topic
Other Areas of Research
• Expert Systems
– Software packages designed to assist in
certain situations when an expert is required
– Perfect for systems that work for specific
areas – can hold all usable information for a
specific function or job
– Often organized as a collection of rules or
facts that are relevant in the given situation
Considering the Consequences
• Great potential to benefit
– Helping mankind solve hard problems we
struggle with
• However, also great potential to be
harmful
– Challenging human intellect
– Altering the image of humanity
– Perhaps even surpassing humanity in
intelligence
– Focus of motion pictures like “The Matrix”
trilogy and “I, Robot”