>> I'm afraid you'll end up with a virtual
poet with this approach !
It will result in a narrow AI, or a vastly inflated expert system, not a true AGI. People have already done this mistake earlier, why repeat it?
-Sanjay >>
It will result in a narrow AI, or a vastly inflated expert system, not a true AGI. People have already done this mistake earlier, why repeat it?
-Sanjay >>
I
wasn't planning on teaching it poetry Sanjay. That would probably be
a dead giveaway on the Turing test. So unless Virtual Poet is a technical term
that I have not been exposed to, I'm not sure where you're coming from with
that.
My eventual goal is definitely strong AI but in the short term I would be
satisfied with an application which had strong commercial value in the areas of
help desk agents, natural language program specification and knowledge
management. The proceeds from such a venture would go a long ways towards
financing the later development phases.
An expert system is usually an idiot savant and is limited to a narrowly
defined problem domain. My program aims to be conversant in a wide
conversational domain sufficient to pass a limited Turing test. (No visual
recognition required)
I
agree that people have spent time developing bots before and they have always
been thwarted by the sheer number of different ways that a human being can
express a single thought as well as the sheer number of items that humans talk
about.
Which is why my research has led to a pattern language that can
compress all of the synonymous thoughts into a single
pattern.
While we can't constrain the number of items the bot knows about and
still have it appear/be intelligent. We can study what things that people
talk about most frequently and what most people know so that the bot are
conversant on the same subjects as an average person.
I
for example would probably not be very
conversant of topics in nuclear energy beyond maybe 20 or 30 facts, concerns,
and opinions.
It is also difficult to quantify the reason most bot projects have been
abandoned. Lack of time, lack of money, lack of interest. It is a
very big project from a knowledge acquisition standpoint. Most
research grants have to occur within a limited time period where you either
publish or perish. A project this large would not fit into such a
timeframe. In the past when such program have been attempted present
day computing resources were not available. I myself had to shelve a few
designs for six years until the computing power became affordable to me. I
have one back shelf design now that offer me incremental improvement but
requires an eightfold improvement in computing speed over what I have
now.
What is the largest bot project you know of and at what point was it
abandoned and why? How may patterns did it have? What percent of it's
responses were correct in extended conversations. Where do I find these
research results?
I
don't know the answer to this and I have done a good deal of
reading. Most people don't spend a lot of time documenting their
failures. If they're lucky they are recognized for what they have learned
and are absorbed into new position on a more conventional project with a shorter
payback which makes it commercially viable.
Actually reading supports that the largest medical expert systems were
equal to or superior to their human counterparts but were abandoned largely
because of legal liability concerns about the danger of misdiagnosis. I
guess no one could figure out how to make money selling malpractice insurance to
an expert system that didn't drink and make errors in accepted medical
protocols. And if an intern could use a program to diagnose an illness
with a program as accurately as a medical specialist then there wouldn't be a
need to pay those specialists as much.
As far as repeating mistakes goes, mythology tells us of Icarus who
flew to close to the sun and fell to his death, Leonardo Davinci designed the
first airplanes, and it wasn't until the Wright Brothers that things literally
got off the ground.
Name me a a single great inventor who did not fly in the face of
convention and did not need to endure the discouragement of mediocre
minds.
From: sanjay padmane [mailto:[EMAIL PROTECTED]
Sent: Sunday, May 07, 2006 12:31 PM
To: [email protected]
Subject: Re: [agi] Logic and Knowledge Representation
On 5/7/06, Gary
Miller <[EMAIL PROTECTED]>
wrote:
My opinion on the most probable route to a true AI Entity is:1. Build a better fuzzy pattern representation language with an inference mechanism for extracting inducible information from user inputs. Fuzziness allows thelanguage to understand utterances with misspellings words run together etc...2. Build a bot based on said language3. Build a large knowledgebase which captures a large enough percentage of real world knowledge to allow the bot learn from natural language data sources i.e. the web.4. Build a pattern generator which allows the bot learn the information it has read and build new patterns itself to represent the knowledge.5. Build a reasoning module based on Bayesian logic to allow simple reasoning to be conducted.6. Build a conflict resolution module to allow the Bot to resolve/correct conflicting information or ask for help with clarification to build correct mental model.7. Build a goal and planning module which allow the Bot to operate more autonomously to aid in the goals which we give it i.e.. achieve singularity.
I'm afraid you'll end up with a virtual poet with this approach !
It will result in a narrow AI, or a vastly inflated expert system, not a true AGI. People have already done this mistake earlier, why repeat it?
-Sanjay
To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/[EMAIL PROTECTED]
To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/[EMAIL PROTECTED]
