Noam Chomsky on Where Artificial Intelligence Went Wrong
<p>Dec 06, 2012 06:15 PM  videos also available:
 
age 83 !!!  article by  Yarden Katz  in Dept of Brain and Cog Sci at MIT.
 
porblem of decoding ourselves:  understnading the inner working of our minds and brains...
 
1956:  John McCarthy coined the term  AI:  
 
chomsky worked on uncovering the mental reps and rules that underlie our percep and cog abilities.
had to overthorw behviorism  bf skinner;
 
1967  Chomsky paper:  critical review of Skinner's book  Verbal Behavior...
 
chomsky stressed the complexity of internal reps...
sophisticated comput system... cannot be broken down into assocs.
 
Behviorist princips cud not explain richness of language,
our endless creative use of it
or how quickly children acquire it...
 
the language faculty was part of the orgs genetic endowment... much like vision, the immune system,
or the circ system....
 
David Marr:   influential book: VISION
3 distinct levels:
first level:  inputs and outputs  :  inputs  pixel matrix... output (say  ID of objects)
Katz's words... not mine:  (visual world is much much richer than just the ID of objects -
look at all you can say about the visual world);
 
second level:  Algo level: by what steps is the input converted to the output;
third level  implementation level:  how dow the bio hardware implements the algo.
 
what is the internal stuc of the system to perform the task.
goal is to dig into the black box...
much like a computer sci wud explain how a piece of software works...
 
Cog sci is a story of the unquivocal triumph of Chomsky over Skinner behaviorism...
the cognit revolution!
 
But behaviorism is not dead for animal behavior studying rodents, eg.
 
May 2011:  150th Anniv of the MIT  symp  Brains Minds and Machines.
Chomsky was NOT enthused.
Chomsky thot AI was reminiscent of ... associationism...
heavy use of statist techniques
unlikely to yield explanations.
For Chom. the NEW AI focused on using stat learning to better mine and predict data...
 
this Critique sparked reply from PETER NORVIG;  defended stat models;
argues that AI's new methods  are similar to other sciences.
 
Chom:  we wud not have taught the computer much abt thephrase "physicist Sir Isaac Newton"
really means... even if we can build a search engine that works.
 
Sydney Brenner.. called genetics:  low input, high thruput, NO OUTPUT science.
When describing  CONNECTOMICS,  Brnner called it a FORM OF INSANITY!!!
 
Isaiah Berlin's famus essay:  Hedge Hog vs Fox.
 
hedgehog.. meticulous and speicalized worker
Fox:  flashier ideas-driven thinker
 
Chomsky:  both  wide and deep;
 
Q: why not more progress in understanding brain:  = YK
if u ask neurosci they say, its cuzof billions of cells and we can't record from all of them.
 
NC= Noam Ch:  correct.  Physics is easy... system is simple...
when a molecule gets big they give it to the chemists;
if bigger, then to the biologists,
if bigger, then to the psychologists, if bigger still then to the literary critics...
 
NC cont:  but neurosci on the wrong track:  neurosci becan enthralled by associationism
(argues RAndy Gallistel and King...new book)...  so.. loking for things like assoc.
 
YK:  Like Hebbian plasticity?
 
NC:  yes... like streng synapses.
Gallistel has argued for years... u shud be asking q like David Marr...
he's inter in insects.
 
Ants go path integration
Bee naviagtion complex computations... position of sun
humans too... it's comput systems.
what are the units of computation.
 
NC:  eg  what is VISION system doing  (a la Marr)
 
YK:  key idea of MARR:  what is the right level of abstraction.
 
 
 
 
 
 
 
 
 
 
 
 
 </p>