[Top] [All Lists]

Re: [ontolog-forum] electric sheep

To: "Barker, Sean (UK)" <Sean.Barker@xxxxxxxxxxxxxx>
Cc: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Pat Hayes <phayes@xxxxxxx>
Date: Fri, 31 Aug 2007 12:17:52 -0500
Message-id: <p06230929c2fdf70a8064@[]>
>To continue on from (was ckae)
>       I go into my local greengrocer, and ask the robot assistant for
>"three green apples, please". it goes off to the drawer marked apples,
>checks the colour against a colour chart, and counts out three (I assume
>Decrement accumulator Jump on Zero is built into the machine code). From
>the point of pragmatics, so far so good. In this description, I have no
>need to ask about semantics, or the concepts the robot is using.    (01)

You really need to find out more about actual robotics and AI 
generally. These points you are trying to make belong in a discussion 
from 40 years ago. There is absolutely no way that such a robot could 
be made to do any of this without having a huge internal system of 
knowledge represented in some processable form, probably in fact in 
an ontology.    (02)

>       If I look at the question of "how did it understand me?", I
>could propose a simple syntactic solution - the robot expects sentences
>of the form <quantity> [<qualifier>] <product>.    (03)

Not a hope in hell that this would work. I have colleagues here who 
are in fact building systems which converse at just about this kind 
of conversational level with naive humans (see for example 
for an overview, and the references there for more details). How such 
systems understand is a very complicated question to answer, 
involving perception of the social context, the common task, grammar, 
phonology and discourse structure.    (04)

>       I could do something a little more complex, and add a dictionary
>which, among other things, includes the information that a word is one
>of {quantifier | qualifier | product* | noise}. (Is this an ontology?)    (05)

No.    (06)

>This probably allows the robot to be more flexible, for example, to deal
>with requests such as "A bottle of your best claret, my good man". I
>could adduce more complex approaches, and perhaps replace the robot by a
>self programming soft machine, which will happy argue semantics with the
>next man.    (07)

Self programming soft machine?? I'm sorry, but you really do not seem 
to have any idea what you are talking about. There is no point in 
continuing this discussion thread. Learn something about AI and NLP 
before giving us all such bland, shallow advice about how to design 
ontologies, please.    (08)

>       At what level of complexity do I need to start concerning myself
>with Semantics rather that just Pragmatics?    (09)

At about anything much past a single chip: and if that chip is itself 
a RISC processor, for example, then inside the chip itself. Programs 
have semantics as well, you see.    (010)

>At what point would one say
>the robot "understands concepts", rather than behaves according to
>particular pragmatics?    (011)

"Understands" is a very loaded word. I try to avoid it. BUt if we 
change this to, at what point would one say that the operation of the 
robot can only be understood or explained by considering in part the 
semantics of the formalisms it uses internally, then the answer is, 
at a very early point indeed. One cannot even do recognition of voice 
phonemes without getting involved with ontological/Krep issues, let 
alone understand such English as "my good man" (which is well beyond 
the current state of the AI art: nobody knows how to process irony, 
humor and indirect allusions.)    (012)

>       I should add that as we develop increasing complex autonomous
>systems, we need to create architectures that provide proper separation
>of concerns, so this is primarily a question about engineering, rather
>than philosophy.    (013)

Yes, we are aware of this. Thank you for your advice, however.    (014)

Pat    (015)

IHMC            (850)434 8903 or (650)494 3973   home
40 South Alcaniz St.    (850)202 4416   office
Pensacola                       (850)202 4440   fax
FL 32502                        (850)291 0667    cell
phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes    (016)

Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Subscribe/Config: 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 Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (017)

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