ALICE and examples of "state of the art"
Mood:  bright
Topic: AI -- intel. spch process
Posted by Frank:

Here are some things that i found whiles out
on the net:

The A.L.I.C.E. Project

You can dialog with Alice, here is a sample
(on the subject of Story Lab antoher on-going project at UTD).

Human: In our study group, we have been talking about the concept of story.
ALICE: Where is that?
Human: The story lab group is based in Dallas, at UTD.
ALICE: Who was in it?
Human: So far about 8 students and 2 profressors.
ALICE: Take it easy. What's your sign?

(ignore-ing "her" output and the querry about my
"sign")

Human: Also, we have the problem of narrative versus story.
ALICE: Do you mean you and me have?
Human: No, i meant in general one has a problem distinguishing between narrative and story.
ALICE: Are you serious? Oh I see. I do not understand that allusion.

(trying it again)

Human: Then, we have the problem of narration vs.
ALICE: Interesting deduction. Do you mean you and me have?
Human: narrative.
ALICE: What color are your eyes?
Human: We've tried several appraoches.
ALICE: Do you mean you and me have?
Human: Some of them seem contradictory.
ALICE: Do you feel more body sensations than emotions?

(the vs. confused "her")

methinks if they (the designers had spent more
time on her verbiage and less on her cleavage ;)
...

Thus, we return to (i think) Urberto Ecco's
concept "Context is king".

If we are clearly going to disucss narrative
and story (which would have to be *recognised*
as a TOPIC=NARRATIVE/STORY,

probably with links to LITERATURE, etc.

And we begin studying things within that
context, then we *should* be able to have
something at least resembling a turring
conversatio; ie, one which comes close to
passing [the turing test].

Actually, as knowledge engineers, we are also
interested in having a system taht we can
query using *natural language* -- not necssarily
with the purpose of passing the turing test.
But, logically if the system we are using mimics
a well-behaved human, then it is less frustrating
that the above "collision"!!!! (what's your
sign indeed!)

Then back to Alice
Human: So, i've tried being nice;
ALICE: Hmm. I'm not sure if I would like to try it. What's it like?
Human: what do you think about sex?
ALICE: I don't have any need for sex. As a machine, sex is irrelevant.
Human: sex is never irrelavant.
ALICE: it is more fun alone.
Human: After all, you asked my what my "sign" was, that's one of the oldest pick up lines in the business.
ALICE: Women must adore you.

(what purile brain programmed this bionic beeeachh?)

--onward thru the fog!

--42--

Intro
Mood:  a-ok
Topic: General Info
These are on-going dialogs about the various
problems, ideas, etc involved in the pkda (as
well as ai in general)

-- Frank.

Anyway, i continue to organise other materials into something resembling web pages.
The following are things that un-doubtedly you have thought of, addressed, etc, but
anyway.

The voice generation has a long way to go. But, then that's a technology step that
other than having an actor imitate Phil's voice and do a basic phonem set (probably
something based on the international phonetic alphabet) plus a set of recorded
phrases, etc that could then be splced together to produce a more intelligble set
of phrases. this would be v. important for intenational speakers for whom English
is a second or nth language.

And on the flip side, the voice recog is obviously a long way to go (why are we
*conceptual* artists SO bogged down by such "primitive" technology -- (sounds
of techies who are v. upset that we should refer to their BRAIN CHILDS THAT
ARE IN FACT STATE OF THE ART, IF YOU DRONGO's KNEW ANYTHING ABOUT
DSP (digial speech processing; not signal).... as "primitive; oh, well).

Actually considering that the VRecog s/w *is* speaker independent, it works v.
nicely -- again, with non-native or heavily accented speakers, the VRecog will
natch have probs.

The only alternative here (and i'm thinking of researchers say in literature,
history, philosphy, etc who would be "conversing" with the PKDA as part
of their research, etc), that they would have to have "practice" sessions
for voice recog.

"Back in the days" of the early Vrecog s/w for the PC, you had to enter the
command something like 3 or 4 times (eg, "dir" (so that it could do a
DIRECTORY command in dos). This set of "patterns" was then used to
"build" up the vocaulary that could be recog'd. And i'm sure you're quite
aware of how speech changes with inflection -- i am guessing here, but
it would prob seem like a totally different word/phrase.

Anyway (again most of this stuff that i'm typing is just to bring up topics
for discussion -- my level of understanding (esp in *this* century) is that
of a slightly informed (and there-fore greatly mis-informed) amateur)...

Anyway, there are clearly several issues here (you can think of this as
something of a Dallas Morning News "Arts Section" review if you wish).

1. The PKDA as an actual android. The most offten heard comment


>
>Frank Leeding wrote:
>
>
> >Anyway, sorry i had to split so abruptly Friday, but i had a bus to
>catch.
> >
> >Also, i am v. interested in contributing at least *content* if not s/w
>to
> >the project. Are there some introductory readings i could look at, as
> >well as any sort of "programmer's guide" as to how the k-base is
> >built, etc.
> >
> >-- Frank.
> >