Transcript English - Washington and Lee University

Glossodiversity and Artificial
Intelligence
Endangered Language
Preservation and the Future
of Smart Machines
Simon D. Levy
Computer Science Department
Washington & Lee University
Lexington, VA 24450
http://www.cs.wlu.edu/~levy
Biodiversity and
Pharmacology
Biodiversity and
Pharmacology
Biodiversity and
Pharmacology
Biodiversity and
Pharmacology
• Greatest plant biodiversity is in
rainforests: 170,000 of the world's
250,000 known plant species.
• “We are trying to do biology
knowing perhaps only a tenth, or
one hundredth, of our species”
– – Terry Gosliner, National Geographic
Biodiversity and
Pharmacology
• Among top 10 most
ecologically diverse
countries on earth
• Third highest
deforestation rate
(1.1%) in South
America
(http://www.mongabay.com)
Biodiversity and
Pharmacology
Language: A Window on the
Mind
• Affects how we think about
the world
• Amazing variety of ways of
saying the “same thing”
• Counter-intuitive constraints
not derivable (?) from more
general principles
Fallacies & Pitfalls
• “Eskimo has over 100
words for snow.”
•“Primitive” languages
•Turing equivalence vs.
Sapir-Whorf Hypothesis
Example: Transitive vs.
Ergative Languages
• English:
He saw her
She saw him
• Basque (http://en.wikipedia.org/wiki/Ergative-absolutive_language):
Gizona etorri da.
"The man has arrived.“
Gizonak mutila ikusi du.
"The man saw the boy."
Example: “Gender”
• English:
(1) Masculine
(2) Feminine
• Dyirbal (Dixon 1979):
(1) Animate objects, men
(2) Women, water, fire, violence
(3) Edible fruit and vegetables
(4) Miscellaneous
Example: Counting
Quantity
!Kung
Warlmanpa
r|e'e
jinta
tsã
jirrama
1
2
n!eni
3
4
Example: Counting
Quantity
!Kung
Warlmanpa
Spanish
English
French
r|e'e
jinta
uno / pimero
one / first
un / premier
tsã
jirrama
dos / segundo
two / second
deux /
deuxième
tres / tercero
three / third
trois /
troisième
cuatro / cuarto
four / fourth
quatre /
quatrième
1
2
n!eni
3
4
Languages Are Also
Disappearing
• “Of the more than 6,000
languages currently being spoken,
fewer than half are likely to
survive the [21st] century”
– Douglas Whalen, Endangered Language Fund
• Appears to correlate with
biodiversity
(Lisa Manne, Ecology & Evolutionary Research 2003)
What Does This Have to Do
with AI?
• AI is programs.
• Different programming languages
express the same thing in
different ways.
• These differences constrain
our ability to think about
and solve problems.
What Does This Have to Do
with AI?
public static int factorial(int n) {
// Java version
if (n == 0)
return 1;
else
return n * factorial(n-1);
}
What Does This Have to Do
with AI?
factorial(0, 1).
factorial(N, F) :N1 is N-1,
factorial(N1, F1),
F is N * F1.
% Prolog version
Classical AI View
• Artificial flying machines don’t
work like natural flying machines
(birds, bats), so why should
artificial minds work like brains?
So Where are the Artificial
Minds?
• To make machines that think, we
need [pace Minsky] to understand
how the human brain/mind works.
The Bigger Picture
• We can’t (just yet) directly study
how the brain works.
The Bigger Picture
We can use phenomena like
language to broaden our view of
how the mind works (what would
an “ergative programming
language” look like?)
The Bigger Picture
The languages most likely to give us
these insights are the ones that
are most endangered.
Let’s (Not) Get Carried
Away…
Links
• Endangered Language Fund:
http://sapir.ling.yale.edu/~elf
• International Clearing House for
Endangered Languages:
http://www.tooyoo.l.utokyo.ac.jp/ichel/ichel.html