ontolog-forum
[Top] [All Lists]

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

To: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Philip Jackson <philipcjacksonjr@xxxxxxxxxxx>
Date: Fri, 30 May 2014 19:17:22 -0400
Message-id: <SNT147-W81A865AC01AD82B416D672C1270@xxxxxxx>
Leo, thanks very much for your comments.
 
The mentalese I propose is not a language for formal logic, but a language based on the syntax of a natural language, for representing concepts normally expressed in natural language.
 
I will look forward to your comments after you've read the thesis. The slides provide an introduction and overview of the thesis.
 
Thanks,
Phil
 
Thesis: http://www.philjackson.prohosting.com/PCJacksonPhDThesis20140422.pdf
Slides: http://www.philjackson.prohosting.com/TowardHumanLevelAI20140422.pdf
 

From: lobrst@xxxxxxxxx
To: ontolog-forum@xxxxxxxxxxxxxxxx
Date: Fri, 30 May 2014 22:25:46 +0000
Subject: Re: [ontolog-forum] Toward Human-Level AI

I’m not in general in favor of any flavor of “mentalese”, including that of logic – at least under the usual notion of “mentalese”, which is a kind of a logic-like language but isn’t, i.e., a language of thought: http://plato.stanford.edu/entries/language-thought/. We argued these issues in the 1980s and largely threw them out, perhaps alas, i.e., in the heyday of philosophy of AI (I just haven’t seen much lately).

 

Also, Philip, I admit that I’ve not yet read your dissertation, so the above opinions are a bit in abeyance.

 

Thanks,

Leo

 

From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx [mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of Philip Jackson
Sent: Friday, May 30, 2014 9:09 AM
To: [ontolog-forum]
Subject: Re: [ontolog-forum] Toward Human-Level AI

 

P.S. John and I are on the same page, but the page is still being written, and we might write parts of it with some differences. So herewith some further comments, to continue the thread. Also changing the subject line since the topic is now broader than intentionality.

 

One topic is the meaning of the word "formal". If this is limited to mathematical logic, then I'd agree that "the hope for a unified formal foundation for AI, NLP, or any truly intelligent system is a dead end". While formal logic plays an important role in representing and processing some concepts, human-level AI needs to work in contexts that contain ambiguities and contradictions, often without attempting to resolve them. It needs the ability of natural language to represent and process concepts with the ability of "stating and relating every paradigm from the most precise to the most vague or fuzzy". Hence Hypothesis II of the thesis proposes a mentalese (called Tala) based very largely on the syntax of a natural language. The use of this mentalese is one "unifying principle" for the TalaMind architecture, at the linguistic level. It might also be considered a "broad generalization" for relating precise and kludgy methods.  Hypotheses I and II also serve as unifying principles.

 

The Tala mentalese may be considered a formal language, i.e. a set of strings of symbols adhering to a syntax, though natural language might also be considered formal in this sense.  Per 3.4.1, the syntax is non-prescriptive, open, and flexible, not limited to perfectly grammatical English. The syntax is extensible via grammatical constructions. The semantics is also extensible and open-ended, embodied in an environment.

 

I'd agree about the usefulness of Common Logic as a unifying principle, especially for formal logical representations, and also agree about the importance of Peirce's semiotics and Minksy's society of mind, both of which are invoked in the thesis.

 

The TalaMind architecture is open to use of formal logic in support of human-level AI, and open to use of other languages and representations when better than natural language for representing some concepts. For example, complex spatial and temporal information can be difficult to state precisely in natural language, and easier to support with diagrammatic notations (e.g. architectural diagrams, electrical diagrams, music notations, etc.) represented as symbolic data structures.

 

As to whether human-level (artificial) intelligence will be "a clear, bright light" or just "a kludge", time will tell. I tend to think it will be a light, composed of many wavelengths. To the extent that it is self-developing, with an architecture having several unifying principles, it may not be a kludge. It may be more like a forest, containing trees of many species that from a distance appear unified in the environment. Yet within the forest there may be some randomness in how the trees develop.


From: philipcjacksonjr@xxxxxxxxxxx
To: ontolog-forum@xxxxxxxxxxxxxxxx
Date: Thu, 29 May 2014 19:23:06 -0400
Subject: Re: [ontolog-forum] Intentionality Best Practices

John,
 
I agree, also. We are on much the same page.

 

Phil
 
> Date: Thu, 29 May 2014 12:18:23 -0400
> From: sowa@xxxxxxxxxxx
> To: ontolog-forum@xxxxxxxxxxxxxxxx
> Subject: Re: [ontolog-forum] Intentionality Best Practices
>
> On 5/29/2014 9:24 AM, Philip Jackson wrote:
> > The thesis proposes we should try to develop a framework that can
> > support many different ways that concepts can be represented, processed,
> > and used to represent meanings, with natural language syntax being a
> > primary way.
>
> I agree. I'd also emphasize the "many different ways" (AKA paradigms).
> Logic is important for KR, but I believe that the hope for a unified
> formal foundation for AI, NLP, or any truly intelligent system is
> a dead end.
>
> For a collection of interesting paradigms that might contribute
> something useful, see http://intelligence.org/research/
>
> That web site has articles on many directions in AI, but the formal
> guys are still trying to unify the world around a single paradigm.
> Years ago, when I took Marvin Minsky's AI course, he made a point
> that I still believe is critical:
>
> Paraphrase of MM by JFS (quoted phrases are verbatim)
> > Some AI researchers claim that "intelligence is a clear, bright light".
> > But others believe that "intelligence is a kludge." The latter are
> > on the right track.
>
> I agree with qualifications:
>
> 1. It's important to search for unifying principles that can relate
> multiple similar paradigms. An example is Common Logic.
>
> 2. It's also important to search for broad generalizations that can
> relate both the precise and the kludgy methods. Examples include
> Minsky's Society of Mind and Peirce's semiotics.
>
> 3. Natural languages -- which can express every paradigm from the most
> formal to the most kludgy -- are capable of stating and relating
> every paradigm from the most precise to the most vague or fuzzy.
> As an example, Wittgenstein's theory of language games shows how
> to relate multiple paradigms within a common framework.
>
> Logicians have complained that Wittgenstein's later philosophy is
> not as formal and elegant as his first book. That's certainly true.
> You can't have a unified formal theory that relates both formal and
> informal methods.
>
> John
>
> _________________________________________________________________
> 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>