[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [lojban] lojban application in wearable computing
> The advantage of Lojban is that it is more than an imperative
> language in disguise.
So at some point, if used for human-computer interaction, would
perhaps be a way to help computers get from reacting to commands to
responding to their environment. AI has to progress a little first,
but I don't think it's inconceivable that in the future I could sit
down at my computer (or wear it!) and say, for example, {.oiro'i} and
the computer could decide what an appropriate response (or
non-response) would be.
Yes, that's it. With Lojban, I think we will have an easier time
going further.
Humans have written programs that fake a non-imperative relationship,
but no one has yet built a machine that actually is independent of
humans.
For an example of a fake response, consider the 1960s AI Eliza
program. It is built into the program which I am now using to respond
to email. I and it just produced this dialog:
human> I am feeling anxious.
computer> Why do you say you are feeling anxious?
On the one hand, the imperative nature of the interaction is
disguised. You don't see me giving orders in the dialog.
On the other hand, in order to start the program, I did give a
command: I ordered the computer to respond to key words in my text in
a manner that weakly emulates a Rogerian therapist.
Clearly, this just moves the location of the order a little away from
the obvious.
You are asking for a yet further step, for a computer that also looks
at the interactive context between me and the computer and figures out
that I am responding to your message. Perhaps you are also looking
for a computer that possesses smell sensors that tell it whether my
body is exuding chemicals that indicate anxiety.
Another step is to move the order to the machine's start up script, so
that when you power-on the computer, it receives its orders -- it
receives its `prime directive'.
Each of these steps disguises the imperative more than the previous step.
However -- and this is where I think Lojban may fit in -- my
expectation is that at some point, as we see an increase in the
complexity of computer's hardware and software, a computer will cease
to respond to humans as a well behaved, intelligent human who is a
subordinate, and begin to behave in a different manner, perhaps like a
person who is running a separate business than you.
My hypothesis is that Lojban can be turned into a better tool for the
kind of programming that this step requires than a regular programming
language, such as Lisp, Prolog, Python, or C.
However, such independence is a long way off. An early step is to
create an interpreter or compiler that can handle standard inferences,
such as
He has a weak cough; I am not sure whether it caused by the dust
in this building or an illness; he definately has a runny nose.
He says he feels terrible. I conclude he most likely has a ....
In the process of creating such a tool, we will find out more of what
makes a programming language different from a language such as Lojban
or English that humans can use to speak about many different things.
For example, programming languages appear to me to be based on logics
that provide coherent and consistent mappings across domains.
As far as I know, all programming language use definitions that
directly, or indirectly, are built from necessary and sufficient
conditions. (`Fuzzy logic' and `uncertainty factor' techniques
provide ways to emulate categories that are not strictly defined
directly.)
Humans, on the other hand, do not define all concepts this way. They
only define some concepts necessary and sufficient conditions, such as
those associated with law, physics, or mathematics. For example,
humans do not define the concept of `mother' with consistent,
coherent, necessary, and sufficient conditions. To be called a
`mother', a person need not be a birth mother, need not be an adoptive
mother, need not be nurturant, need not take care of the baby, and
need not be female.
Instead, in every day life we humans use a different categorization
method. However, to settle disputes we often use courts; and Western
courts often redefine the aspect of an everyday category such as
`mother' to be appropriate for the case at hand. Usually, a court
uses consistent, coherent, necessary, and sufficient conditions so
that it can draw inferences and come to a conclusion.
Interestingly, humans consistently and coherently map bit patterns in
a computer to concepts that have meaning for humans -- that is to say,
the humans undertake a metaphorical action, a mapping, that works as
far as the humans are concerned.
For example, I command the computer to move the cursor back to the
beginning of the current sentence by typing an ALT-a keychord. From
the point of view of the computer, there is no meaning to this action.
>From the computer's point of view, it receives a bunch of electronic
pulses and as a result, it changes changes the set of electronic
pulses that it sends to the screen. However, the action is meaningful
to me. I see the cursor jump back to the beginning of the sentence;
that location becomes my new editing point.
--
Robert J. Chassell bob@rattlesnake.com
Rattlesnake Enterprises http://www.rattlesnake.com