ontolog-forum
[Top] [All Lists]

[ontolog-forum] FW: Toward Human-Level AI

To: "'[ontolog-forum] '" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: "Rich Cooper" <rich@xxxxxxxxxxxxxxxxxxxxxx>
Date: Thu, 12 Jun 2014 15:09:03 -0700
Message-id: <002101cf868a$ec6dee70$c549cb50$@englishlogickernel.com>

 

Dear Phillip,

 

I take exception on the basis that there is no syntax yet known which covers all parts of the English language.  No compilers have been demonstrated that fully represent all those nuances of syntax which people throw together, along with many statements that are not even real sentences but which people use reflexively in normal conversation. 

 

I am exploring the idea of avoiding syntax entirely.  By using layered dictionaries and pattern matching, I wrote a program that is able to parse all existing patent documents available from the USPTO patent database.  That is a corpus of about two million patents, I am told, but the actual number isn’t something I’ve counted. 

 

The point is, I am able to divide sentences and phrases into units, then I am able to parse claim sentences – huge multihundred word sentences in many cases – without any exceptions!  I haven’t tried them all yet, but every one I have tried is parsable that way. 

 

Syntax doesn’t need to be parsed – it needs to be pattern matched, IMHO.

 

I tried the Link Grammar Parser with the English dictionary, and it simply can’t handle the processing load, which takes forever even for simple claim sentences.  So I gave up on that thread entirely, and I suggest you might want to consider something similar.

 

 

-Rich

 

Sincerely,

Rich Cooper

EnglishLogicKernel.com

Rich AT EnglishLogicKernel DOT com

9 4 9 \ 5 2 5 - 5 7 1 2

From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx [mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of Philip Jackson
Sent: Wednesday, June 11, 2014 7:01 AM
To: ontolog-forum@xxxxxxxxxxxxxxxx
Subject: Re: [ontolog-forum] Toward Human-Level AI

 

P.S. To clarify and avoid understatements, as well as overstatements.
 
In certain respects, the following statements from the slides for the thesis describe what is true now, and are not just claims about what can be achieved in the foreseeable future: 

Tala can represent unconstrained,
complex English sentences,
involving self-reference, conjecture,
and higher-level concepts – anything
that can be expressed in English,
with underspecification and
semantic annotation.
 
Tala sentences can describe how to
perform processes and be executable
concepts, i.e. Tala is a simple
programming language, with
variables, pattern-matching, etc.  

English syntax supports complex statements involving self-reference, conjecture, and higher-level concepts. Since the syntax of Tala is based on English syntax, Tala also has the ability to syntactically represent such statements. This is true with the initial syntax design presented in Chapter 5, although developing a comprehensive English syntax is a topic for future research.  It is also shown in the simulations of the demonstration system, along with the ability of Tala sentences to be executable concepts, with variables and pattern-matching.
 

Chapter 3 gives theoretical discussions about semantic representation and processing using English syntax within a TalaMind architecture.  The demonstration system gives a very limited illustration of how semantic processing could be implemented on a computer, though fully developing TalaMind architectures to understand natural language is a topic for future research.

 

Phil

 


From: philipcjacksonjr@xxxxxxxxxxx
To: ontolog-forum@xxxxxxxxxxxxxxxx
Date: Mon, 9 Jun 2014 14:51:53 -0400
Subject: [ontolog-forum] FW: Toward Human-Level AI

Ed,
 
Thanks very much for your comments. I would not agree that the Tala language is "remote from anything that has been implemented on a computer or is likely to be in the foreseeable future."
 
Rather, my thesis presents theoretical arguments intended to show it is possible in the foreseeable future, a design for the syntax of Tala, and a demonstration system to illustrate how it could be implemented on a computer, with variables, pattern-matching, etc.
 
That said, I agree there is much more research to be done to fully implement a TalaMind architecture that could achieve human-level AI. Whether the goal is likely to be achieved, and remote from what has been achieved, may be a matter of one's perspective. I think we are closer to achieving the goal, and more likely to achieve the goal, if we take this direction.  Arguably, we cannot achieve the goal by relying only on formal logic approaches: we need the ability for systems to represent and reason directly within a computer language based on the syntax of a natural language.
 
Phil 


Date: Mon, 9 Jun 2014 13:19:14 -0400
From: eslowry@xxxxxxxxxxxx
To: ontolog-forum@xxxxxxxxxxxxxxxx
CC: philipcjacksonjr@xxxxxxxxxxx
Subject: Re: [ontolog-forum] Toward Human-Level AI    


Phil Jackson describes the Tala language as follows:
*********
Tala can represent unconstrained,
complex English sentences,
involving self-reference, conjecture,
and higher-level concepts – anything
that can be expressed in English,
with underspecification and
semantic annotation.
 
Tala sentences can describe how to
perform processes and be executable
concepts, i.e. Tala is a simple
programming language, with
variables, pattern-matching, etc.
*******
A language like that seems remote from anything that has been implemented
on a computer or is likely to be in the foreseeable future. 

I would value translations of the example below into realistic computer executable
language.  It would help illustrate both the need for improved computer language
and the practical possibility producing substantial improvements.

Ed Lowry


On 6/9/2014 11:49 AM, Philip Jackson wrote:

Ed,
 
Of course, in Tala the question can be stated directly corresponding to:
 

How many states have cities with more than a million people?

 
My arguments are that we need Tala as a new artificial language.
 
That said, I agree that the methods Kowalski describes could play an important role in a TalaMind architecture.
 
Phil

Thesis: http://www.philjackson.prohosting.com/PCJacksonPhDThesis20140422.pdf
Slides: http://www.philjackson.prohosting.com/TowardHumanLevelAI20140422.pdf
(links included for readers who may be new to the thread.)
 


Date: Mon, 9 Jun 2014 11:14:55 -0400
From: eslowry@xxxxxxxxxxxx
To: ontolog-forum@xxxxxxxxxxxxxxxx
CC: john@xxxxxxxxxxxxxxxxxxxx
Subject: Re: [ontolog-forum] Toward Human-Level AI

John and John

John Bottoms writes:
> Tell me again why we need a new computer language?
> I would rather vote to kill off some of the languages we already have.

We need a new programming language because the currently used languages
(including Prolog) are deficient on 7 simplicity-related leading edges compared
with a design circulated at IBM in 1973. See "Inexcusable Complexity for 40 years"
on my web site.  Killing off some current languages would ALSO be a good idea.

If you don't think we need a new design that allows for greater simplicity, please
translate
        6 = count every state where populatn of some city of it > 1000000

into your favorite language and tell us why anyone should allow their brain to be
burdened with such irrelevant complexity. I expect those 14 tokens will require over 35
in the best current languages. The above capability was running at DEC in 1982.

Simplicity of _expression_ is important for advanced AI because that requires
working with information about information and the extraneous complexities
get compounded.

It is possible to design building blocks of information which are
well designed to be easily arranged.   Fundamentals are being ignored.

John Sowa writes:

 > Anybody who has
 > never written at least a few exercises with Prolog or other logic
 > programming (LP) language is not qualified to express an opinion
 > on this topic.

Something similar could be said about the many people who are 40 years behind on
those 7 leading edges.

To John Bottoms: I live in the next town and would be glad to discuss these issues further.

Ed Lowry
Bedford Mass
781 276 4098
eslowry@xxxxxxxxxxxx
http://users.rcn.com/eslowry 
 
                **********************************

On 6/8/2014 10:13 PM, John Bottoms wrote:

On 6/8/2014 3:25 PM, John F Sowa wrote:

I came across a book by Bob Kowalski, who makes a strong argument
for model building as a basis for intelligent systems.  The ideas
are further developments of his work on logic-programming systems.
 
Kowalski, Robert (2011) Computational Logic and Human Thinking: How
to be Artificially Intelligent, Cambridge: Cambridge University Press.

John,
There have been a number of people who have put forward that concept. It seems to be a common basis for the implementation myth. AKA: if you only did this then you would have AI. This was one of the leading contributors to the 5th Generation AI failure. The designers felt that if you implemented logic then you would have a smart machine. It is a variant of the Philosopher's Disease: if you implement all the words in a dictionary, then you will have a smart machine.

Certainly there is a need for modeling, and likewise, there are roadblocks to doing that right now. We don't know what type of KR to use, we don't understand how the models should be structured, and we don't know the details of extensibility.

There is no dearth of programming languages. We now have as many computer languages as we do natural languages. That is because they are increasingly easy to construct with tools provided for just that purpose. Also, there are efficiency justifications for creating new languages. In fact, my natural language is not particularly efficient. The estimates are that English is about 50% redundant...for very good reasons. Tell me again why we need a new computer language? I would rather vote to kill off some of the languages we already have.

Why do you think this work is important? Or, why is this approach more important than any other type of solution? Does it move us forward an order of magnitude in any way?

-John Bottoms
 FirstStar Systems
 Concord, MA USA

You can buy it for $44.10 or download an earlier version for free:
http://www.doc.ic.ac.uk/~rak/papers/LPS%20revision.pdf
 
The papers cited below go into more detail about LP methods for
deriving imperative procedures from declarative statements.
 
Logic Programming
http://www.doc.ic.ac.uk/~rak/papers/History.pdf
 
Completeness of a Reactive System Language
http://www.doc.ic.ac.uk/~rak/papers/KELPS%20Completeness.pdf
 
Reactive Computing as Model Generation
http://www.doc.ic.ac.uk/~rak/papers/LPS%20revision.pdf
 
See below for excerpts from each of these sources.
 
John
_________________________________________________________________________
 
 From the book:
 
Although the applications of Computational Logic in AI require the
use of mathematical notation, its human applications do not. As a
consequence, I have written the main part of this book informally,
to reach as wide an audience as possible. Because human thinking
is also the subject of study in many other fields, I have drawn upon
related studies in Cognitive Psychology, Linguistics, Philosophy,
Law, Management Science and English Composition.
 
In fact, the variant of Computational logic presented in this book
builds not only upon developments of logic in AI, but also upon many
other complementary and competing knowledge representation and problem
solving paradigms. In particular, it incorporates procedural representa-
tions of knowledge from AI and Computing, production systems from AI
and Cognitive Science, and decision analysis from Management Science,
Cognitive Psychology and Philosophy.
 
 From the article, History.pdf
 
The driving force behind logic programming is the idea that a single
formalism suffices for both logic and computation, and that logic
subsumes computation.
 
 From the article, KELP Completeness.pdf
 
KELPS is a first-order, sorted language, including a special sort
for time.  In the version of KELPS presented in this paper, we assume
that time is linear and discrete, and that the succession of time
points is represented by the ticks of a logical clock.
 
 From the article, LPS revisions.pdf
 
In this paper we propose a logic-based framework inspired by artificial
intelligence, but scaled down for practical database and programming
applications.  Computation in the framework is viewed as the task of
generating a sequence of state transitions, with the purpose of making
an agent’s goals all true.
 
_________________________________________________________________
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
 

 

 
_________________________________________________________________
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
 

_________________________________________________________________ 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

 

 
_________________________________________________________________
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
 


_________________________________________________________________ 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


_________________________________________________________________
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    (01)

<Prev in Thread] Current Thread [Next in Thread>