Greg Detre
Wednesday, February 19, 2003
wolfram has
the same problem as types of computation � the theory�s not rich because almost
all types of problem fall into the same overly-large category � wolfram�s book
is an enormous failure because it doesn�t add anything new to the original
4-part classification scheme
SoM
� causes + clauses
computer
science is valuable because it�s a way of describing processes, and came up
with loads of useful ideas like stacks and cache memory
although there was a pocket of mathematics called recursive function
theory in the early 20th century previous to compsci
there is a
class of machines just slightly weaker than turing machines but there�s no
known use for them
Newell
+ Simon � the variety of learning in a general problem solver
nobody paid any attention to it, and they should have
Proust as
one of the best sources of ideas
�The Mezzanine� by Nicholson Baker??? � subject of thought vs number of
times thought occurred per year � novel about what he thought about for 15
minutes
tEM
�resources� are the new name for �agents�, because that has come to mean independent self-contained functional parts
top-down rather than bottom-up like SoM
Susumo Ono, �Duplication and diversity� � Nobel prize-winning genetics of spinal segments
4 closest relatives � chimps, baboons, orangutangs, gorillas
one
conjecture of the main difference is that we have a stack that�s three deep,
and theirs is only two deep
credit assignment
presumably you could get a bigger brain just by turning off the growth hormone later � but bigger brains use more energy, make you less mobile
more imporantly, the brains would just have been filled with junk and so the credit assignment would be that much harder
last year, theories about the cerebellum
were blown apart, because it seems to be implicated in higher-level cognition
�don�t study neuroscience to learn anything, but rather because these people are lost and maybe I can help them� J
the important thing about language
is the ability to make a parallel structure serial
we already have representations, even if not a lexicon, so it�s not simply the ability to label things that helps elevate man above animal
does that mean that if we had parallel languages, they�d lose their power???
maybe big brains were waiting for language, in order to speed up the learning process enough to make it work in time
perhaps language uses prepositions as a means of mapping similar
concepts between realms (see Lakoff)
the Jungian archetypes don�t need to be innate, because they�re all based on the sort of basic universal rules that you need in order for cultures not to fall apart
maybe some
of the old foundational mechanisms from our animal past are just holding us
back
we don�t know if a dog can see objects
you might
be able to isolate a sub-domain of common sense
e.g. blocksworld, or non-social
he doesn�t think that there�s a critical mass of CS knowledge necessary,
so much as a critical mass architecture (e.g. that knows what is and is not
worthwhile learning - context)
Colossus �
D F Jones
he�s saying that before you get too precious about how much computational power there is in the brain, bear in mind that you can�t remember much beyond a 7-digit phone number J
he reckons
that possibly once you�ve made an architecture, it takes a different type of
engineer to make it reliable � but perhaps that approach is a mistake
ungrounded
� little brain with a simple world, small number of
states � neurologists should look for this
he thinks
that we don�t do 3D spatial reasoning, only 2D, except by tricks
are there psych experiments to undermine this???
blind people can develop an equally rich spatial conception (with
different auditory dimensions)
what happens to deaf children who don�t learn a sign language?
perhaps
�liking� isn�t something positive, so much as one part of your brain
over-riding other impulses
if
everybody agrees, it must be a contagious mental disease (e.g. grounding) �
though rarely, it�s actually true
didn�t
think that Spielberg�s AI had a single good idea
"...Psychologically we must keep all the theories in our heads, and every theoretical physicist who is any good knows six or seven different theoretical representations for exactly the same physics. He knows that they are all equivalent, and that nobody is ever going to be able to decide which one is right at that level, but he keeps them in his head, hoping that they will give him different ideas for guessing."[7]
see questions doc