Leo and Gary, (01)
I'd like to comment on both of your notes under this thread. (02)
Leo
> about Hector Levesque’s (U. Toronto) view on the flaws of current AI:
>
>
>http://www.newyorker.com/online/blogs/elements/2013/08/why-cant-my-computer-understand-me.html (03)
I agree. But Alan Perlis summarized those points more succinctly:
> A year spent in artificial intelligence is enough to make one believe in God. (04)
Leo
> Hector Levesque’s IJCAI 2013 paper:
> http://www.cs.toronto.edu/%7Ehector/Papers/ijcai-13-paper.pdf. (05)
HL
> 4.4 A radical approach
> So to account for what people are actually able to do, we need to consider
> what it would take to have a system that knows a lot about its world and
> can apply that knowledge as needed, the way people can.
>
> One possibility is this... (06)
My main disagreement is with the emphasis on "one". In my 1984 book,
the final chapter was "The Limits of Conceptualization". In my KR book
(2000), Chapter 6 was "Knowledge Soup". There is no such thing as
"one" ideal logic, ontology, methodology, classification scheme, or
set of principles or primitives for deriving them. (07)
I won't deny the possibility that some infinite being (call it whatever
you like -- God, Nature, Logos, Tao, etc.) might have such a scheme.
But no scheme that is humanly knowable can ever be complete. (08)
Gary raises some issues that illustrate this principle. (09)
Gary
> William Frank went on to note that:
>
> Combinations of classifiers that are part of orthogonal classification
> schemes need to be accommodated in a different manner (most effectively,
> in my experience, through composition, but most commonly through
> "multiple inheritance" (but me, I do not know what "inheritance" means,
> except in biology and in class-oriented programming languages --
> ((though I used to know, before I thought about it much)). (010)
In that point, William was discussing automated methods for classifying
things described by some kinds of features, attributes, properties,
facets, or monadic relations. Arbitrary conjunction of such features
generate a partial ordering. If you fill in the gaps of the partial
ordering, you get a lattice. (011)
But if you look at most of those classification schemes, the so-called
features or attributes are often described by English phrases that have
a complex internal structure. That indicates some *hidden structure*
that is independent of and not represented by the classification. (012)
Gary
> Hacking makes his most critical comments on what he describes as its
> mis-application of classification schemes from Biology to mental illness.
> He stated it this way (note - NOS, stands for Not Otherwise Specified)... (013)
I agree that NOS is a warning sign. It's the tip of an iceberg that has
a huge amount of hidden structure. I like the following point: (014)
Ian Hacking
> Sauvages’s dream of classifying mental illness on the model of botany was
> just as misguided as the plan to classify the chemical elements on the model
> of botany. There is an amazingly deep organisation of the elements – the
> periodic table – but it is quite unlike the organisation of plants, which
> arises ultimately from descent. Linnaean tables of elements (there were
>plenty)
> did not represent nature.” (015)
Hacking is comparing three systems whose hidden structure is very
different: mental illness, botany, and chemical elements. For each
of them, that structure determines their observable properties. But
different structures would lead to different methods of organization: (016)
1. Biological species are an extremely important, but *unusual*
special case. The method of reproduction determines a tree,
but very few non-biological classes are as strictly tree-like.
Even bacteria, for example, can exchange genes with other bacteria.
With modern methods of genetic engineering, the strict trees for
other biological species can be broken down. (017)
2. The so-called periodic table is the result of the "hidden structure"
of atoms. But as physicists discover more about that structure,
the simple table is becoming much more complex. Even Mendeleev's
original "table" was more complex than a table. (018)
See http://www.aip.org/history/curie/periodic.htm . (019)
3. The DSM certainly can be and has been severely criticized for many
reasons. The underlying causes and interrelationships among mental
symptoms and their grouping into "illnesses" are mostly unknown.
The structure of a printed book and its table of contents is a tree.
But nobody knows the appropriate attributes or "primitives" that
could generate that tree or a more appropriate classification. (020)
These observations about biological species, chemical elements, and
mental illnesses can be repeated for *every* branch of human knowledge,
experience, or activity -- and for the terminology about them. (021)
These are among the many reasons why I emphasize the weaknesses
of any proposal for a single ideal, universal, all-purpose system
for classifying everything that exists -- or all the ways of talking
about everything that people want to talk about. (022)
It's important to distinguish systems for classifying things,
organizing things, finding things, analyzing things, and reasoning
about things. Those are related, but different purposes. No single
logic, ontology, or methodology could be ideal for all of them. (023)
John (024)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J (025)
|