Hi Joe,
You wrote
People do not understand what is going on in a black box AI system and therefore do not trust the result.
I think you my have overlooked explanations, or more formally, proofs. There are many examples of proof extraction from inference processes in the AI literature. For example, in the system online at the site below, proofs are presented in hypertexted English at the end-author / end-user level, even if the actual inference was handed off to automatically generated SQL for big data efficiency.
BTW, at first glance, the Combs method appears to apply to non-fuzzy logics as well as to fuzzy ones.
Cheers, -- Adrian
Internet Business Logic
A Wiki and SOA Endpoint for Executable Open Vocabulary English Q/A over SQL and RDF
Online at www.reengineeringllc.com
Shared use is free, and there are no advertisements
Adrian Walker
Reengineering
On Sat, Feb 4, 2012 at 1:06 AM, joseph simpson <jjs0sbw@xxxxxxxxx> wrote:
John:
There may be two specific areas of artificial intelligence (AI) application that could be improved.
The first area is generally referred to as combinatorial "rule explosion" as detailed by William Combs in the Combs Method.
http://en.wikipedia.org/wiki/Combs_method
The second area is generally referred to as "trust."
People do not understand what is going on in a black box AI system and therefore do not trust the result.
I believe that there is a way to address both of these issues by applying a controlled abstraction technique that is composed of "structure nodes" that are constructed from restricted set of items. These nodes are then evaluated using a context dependent relation.
The act of aggregation and structure development provides an approach where the human expert can check the model for validity.
Early conceptual design of a system of this type is ongoing with the structure nodes being named ConceptCubes (sm) (CC).
The size and structure of the CC is designed to allow a human expert to reason about the items and relationship without overloading the cognitive capability of the human experts.
Designing AI systems that are sensitive to the cognitive limitations of humans is the best way to encourage the mainstream application of AI, in my opinion.
I agree that there must be improvements in the deployment and use of AI systems and addressing human cognitive limits is a great place to start the AI system modification process.
Have fun,
Joe
On Fri, Feb 3, 2012 at 9:29 PM, John F. Sowa <sowa@xxxxxxxxxxx> wrote:
Amanda,
I certainly agree that last year's communiqué was a respectable,
well balanced statement to which I contributed a few words and
some of the editing. But in my complaint, I was expressing my
frustration that the good, hard work by many good people just
dropped silently into the depths of the WWW without creating
a noticeable ripple.
JFS
>> Last year's summit did not get into those issues. It consisted of talks
>> that said "Ontologies are wonderful. You should be using them." A year
>> later, we see that the major web companies have abandoned any pretense
>> of using ontologies, and people who need to share Big Data ignore them.
AV
> This is not an accurate or fair characterization of last year's
> summit. While the focus of the last year's summit was indeed on making
> the case, it was not assumed or widely believed that any such blanket
> case should be made.
Does anyone know whether it had any useful effects?
AV
> I am very much supportive of, and motivated to invest my time further by,
> the very degree to which this forum has avoided the mistakes of which
> you here accuse it.
I wasn't accusing anybody of making any mistakes. I said that the page
for the quality track that you put together was good. But what I found
frustrating is that the kinds of "ontologies" people are actually using
are nothing but terminologies.
AV
> In the enormous effort that went into the development of the
> Application Cases...
> In the value metrics synthesis... in which an analysis is offered of
> where and how ontologies are currently capable of providing value.
Those are good summaries, and I have no criticisms of them. But I
was responding to Nicola's point:
NG
>> I believe that the questions below ("Why haven't semantic technologies
>> been adopted as an integral part of the IT mainstream?" and similar),
>> although very relevant, are out of scope for the present Summit. Indeed,
>> these issues have been discussed rather extensively in the latest Summit,
> Last year's summit did not get into those issues.
JFS
> Last year's summit did not get into those issues.
That is not a criticism of last year's work. It just makes the point
that it did not address the question why AI technology in general
and ontology in particular has not made a dent on mainstream IT.
I was an "early adopter" of the word 'ontology' in my 1984 book
(finished in 1983). I worked on AI projects at IBM before and
after that time, and I have been participating in conferences,
workshops, projects, emails, etc., about ontology for years.
Like many people who have been working in the AI field for years,
I would like to see something useful come out of it.
John
Joe Simpson
Sent From My DROID!!
_________________________________________________________________
Msg Archives: http://ontolog.cim3.net/forum/ontology-summit/
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontology-summit/
Unsubscribe: mailto:ontology-summit-leave@xxxxxxxxxxxxxxxx
Community Files: http://ontolog.cim3.net/file/work/OntologySummit2012/
Community Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OntologySummit2012
Community Portal: http://ontolog.cim3.net/wiki/
_________________________________________________________________
Msg Archives: http://ontolog.cim3.net/forum/ontology-summit/
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontology-summit/
Unsubscribe: mailto:ontology-summit-leave@xxxxxxxxxxxxxxxx
Community Files: http://ontolog.cim3.net/file/work/OntologySummit2012/
Community Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OntologySummit2012
Community Portal: http://ontolog.cim3.net/wiki/ (01)
|