From: Ben Goertzel (ben@goertzel.org)
Date: Fri Feb 11 2005 - 14:59:39 MST
Matt,
You mention Lojban ... I am curious about this language; I have read about
it but haven't learned it...
Do you find it's pragmatically possible to express a wide variety of things
in Lojban? -- including ambiguous emotional, intuitive, creative and
speculative things, i.e. things which are far removed from the formal-logic
foundation of Lojban semantics?
I have often wondered if it would be a good idea to use Lojban to help teach
an AI system.
For our Novamente AI system (www.agiri.org,
http://www.realai.net/AAAI04.pdf), we have an interactive user interface
that allows the system to understand English sentences correctly after some
interaction with the user who corrects the system's misinterpretations, but
this is a cumbersome process. Of course, the system's comprehension
improves gradually, but we're not sure how fast this will be....
I suspect that if we made a Lojban version of our interactive English
system, the system would pick up the language a lot easier.
I understand that Lojban has a completely formal syntax, but syntax
processing isn't all of comprehension -- I don't have a good sense of how
easy a time Novamente would have with Lojban semantics.... I imagine it
would have a pretty easy time since Lojban is loosely based on predicate
logic and Novamente's probabilistic term logic knowledge representation also
shares a lot of aspects with predicate logic.
I stress that this is just a speculative discussion -- being as profoundly
resource-starved as the Novamente project currently is, we're not about to
undertake a speculative side-project like this, but it's interesting to
think about...
Novamente's current treatment of syntax is based on a modified version of
the Link Grammar
http://www.link.cs.cmu.edu/link/
If a similar grammar were created for Lojban then it would be relatively
easy to make a Lojban interface for Novamente. I wonder if you have a sense
for how hard a job it would be to make a grammar of this form for Lojban.
It seems to me like it should be fairly simple -- though NOT trivial --
given the existing yacc parsing rules for Lojban.
-- Ben
> -----Original Message-----
> From: owner-sl4@sl4.org [mailto:owner-sl4@sl4.org]On Behalf Of Matt
> Arnold
> Sent: Friday, February 11, 2005 3:41 PM
> To: sl4@sl4.org
> Subject: JOIN: Matt Arnold
>
>
> Hi, I'm new to the list. A resident of the suburbs of Detroit,
> Michigan, I have been a transhumanist since almost back to the turn of
> the twenty-first century. I am more or less the same age as Eliezer
> Yudkowski. To find my tastes in reading, turn to the bibliography in
> David Pulver's GURPS Transhuman Space main sourcebook. I help to run
> the science fiction conventions in the area, ConFusion and Penguicon.
> (This last is a combination free/open-source software and SF
> convention.) I'm trying to install Linux and learn how to run a web
> server, but I'm not yet a tech geek and possess more of an artistic
> than an engineering mindset. I am a Universist and a member of the
> Fellowship of Reason and the Church of Virus. I speak the artificial
> language Lojban passably well and continue to improve. If any of you
> are on the Lojban mailing list, I'm the one who signs "la epcat". My
> website is http://www.geocities.com/nemorathwald/.
> I finally signed up to the list to get some advice about preparing for
> a singularity or technocalypse, so that I can improve the rules to a
> little game I'm creating which crudely represents terraforming and
> pantropy and is played on a Hoberman Sphere. More about that later.
> -Matt Arnold
>
>
This archive was generated by hypermail 2.1.5 : Tue Feb 21 2006 - 04:22:53 MST