Artificial Intelligence

Download Report

Transcript Artificial Intelligence

What is intelligence?
 What do you think makes a person “intelligent”?
 Some ideas:
being able to do lots
of math in your head
being able to memorize
lots of names and dates
grandiloquence (a fancy
word for using fancy words)
having a photographic
memory
knowing everything
in the encyclopedia
knowing how to say words
in different languages
Can you think of any others?
Intelligence is complicated
 These things make you smart, but not necessarily
intelligent
 Intelligence is more than just “book smarts”
 A computer can solve massive equations in less than a
second but it still isn’t intelligent like you are
 Humans have many different types of intelligence;
computers only really have one.
Gardener’s Theory of Multiple
Intelligences








Bodily-kinesthetic
Interpersonal
Verbal-linguistic
Logical-mathematical
Naturalistic (*)
Intrapersonal
Spatial
Musical
 http://www.literacyworks.org/mi/assessment/index.html
Computers and Intelligence
 How well do you think a computer would do in your
math class?
 What about your English class?
 How did you learn to be good at these things?
 Practice, practice, practice…
Can a computer “think”?
 “The question of whether Machines Can Think... is
about as relevant as the question of whether
Submarines Can Swim.”
 Edsger W. Dijkstra
 Is thinking the same thing as learning?
 Let’s find out…
More on “thinking”
 "Can machines fly?" This is true, since airplanes fly.
 "Can machines swim?" This is false, because
submarines don't swim.
 "Can machines think?" Is this question like the first, or
like the second?
 Drew McDermott: "Saying Deep Blue doesn't really
think about chess is like saying an airplane doesn't
really fly because it doesn't flap its wings.“
 John Searle: A thinking machine can only be a
simulation
How do we know if something is
intelligent?
 We know how to judge intelligence in humans, so how
can we test for intelligence in machines?
 We compare them to humans!
 Alan Turing thought up a test
to see if a machine is as
intelligent as a human.
 It goes like this:
The Turing Test
Turing Test
 Put a person at a computer. Hook
this computer up to 2 other
computers in a chat program:
 One has a person on the other
end, the other has a computer
 If the 1st person can’t tell which
is the machine and which is
human, it’s true Artificial
Intelligence
However…
 Not one single computer has ever passed the Turing
test!
 Here’s an example of one trying, and failing
miserably…
So much for intelligence…
Hi, how are you?
I’m fine.
What are you going to do with your
day?
What are you going to do with
yours?
I'm learning about artificial
intelligence.
I have none.
You have no intelligence?
You have no chocolate.
I didn’t say anything about chocolate.
Did you know the ocean is
orange?
I’m pretty sure it’s blue.
What is me?
 A chat with Cleverbot
More unintelligent robots
 http://www.youtube.com/watch?v=WnzlbyTZsQY
 Siri?
 Autocorrect?
What next?
 So we’ve established that so far, no computer has ever
managed to pass as a human
 Why do you think this is?
 What do you think its learning algorithms look like?
Uncanny Valley
 Do we need or want robots that are human-like?
Social Robots
 Kismet, Leo, and social robots:
 http://www.ted.com/talks/cynthia_breazeal_the_rise_
of_personal_robots.html
Discussion
 The moving camera robot seems to be like having a
hologram, but physical. What do you think of this over
taking a plane to go visit family overseas?
 Is it creepy or cool to have a robo-friend? Do you think
these will replace real friends?
“Strong” vs. “weak” AI
 “Strong” AI means intelligence in the way we humans
think of it: being able to act and think like us
 “Weak” AI, also “applied” or “narrow” AI, is made to
respond to specific situations intelligently, though not
necessarily with human intelligence
Is strong AI possible?
 Computers fail the Turing Test because it tests for
general intelligent action or “strong AI”
 Strong AI encompasses everything a human can do,
which is a lot.
 Some robots, however, can be programmed to learn, so
even if they can’t do something now, they can write
their own programs for themselves.
 Some philosophers don’t even think strong AI is
possible.
Synthetic Intelligence
 Another term for AI that emphasizes the idea that
computer intelligence doesn’t need to imitate human
intelligence
 We’ve come up with many weak AI and calculating
programs that are much more efficient than humans
 Calculators

Such as programs you learned to write in Python for
converting temperatures and the Pythagorean theorem.
Applications of AI
 Speech recognition
 Handwriting recognition
 Natural language translation
(harder than it seems; that’s
why you should never use
FreeTranslation.com to do your
Spanish homework)
 Video game virtual opponents
 Diagnostics (WebMD, anyone?)
Philosophy of AI
 You’ve all seen sci-fi movies with evil robots or
machines that can think
 Is it even possible to program something to think for
itself?
The Future of AI
 We may only have weak AI now, but people are still
doing research to make strong AI in the future
 Perfecting learning algorithms
 There’s research in using neural networks to build
better AI
 Predictive modeling, adaptive control (like neurons in
the brain)
What are people doing right now?
 Robots that fly, cooperate, and map rooms:
 http://www.ted.com/talks/vijay_kumar_robots_that_f
ly_and_cooperate.html
 Vijay Kumar and his team build flying, small, agile
robots that swarm, sense each other, and form
specialized teams -- for construction, surveying
disasters and far more.
Games!
 Intelligent Piece of Paper Game
 20 Questions
 http://www.20q.net/
 Akinator (20 Questions for people/characters)
 http://en.akinator.com/