[Wittrs] Re: Defining Consciousness -- Can it be caused?
- From: kirby urner <kirby.urner@xxxxxxxxx>
- To: Wittrs@xxxxxxxxxxxxxxx
- Date: Mon, 10 Aug 2009 12:21:23 -0700
On Mon, Aug 10, 2009 at 12:00 PM, Stuart W. Mirsky<SWMirsky@xxxxxxx> wrote:
<< SNIP >>
>> Nah, if Big ASS Big Blue passes the Turing Test in Chinese, to the
>> satisfaction of experts, I'd be inclined to say this computer
>> understands Chinese and be done with it. I'd be amazed of course, as
>> I don't think even our most proprietary software, let alone open
>> source, is anywhere close to coding for context,
>
> This is an interesting point and perhaps nods toward one of the problems in
> discussions about this stuff. I suppose ONE way of doing AI would be to do
> things like program for context, nuance, fuzziness of meaning, etc., but I
> would doubt that this is how brains do it. Whether brains operate by some
> kind of mix of pre-existing and self programming or in some other way, I
> doubt they are simply programmed for context, etc. Rather, it's likely
> they're programmed to do various things with the inputs they get in a
> layered and interacting way and that these things, when linked in a common
> system that involves lots of interaction and coordination between and within
> sub-systems, produces the features that enable context recognition, nuance,
> fuzziness, etc.
Yeah, if I were given budget to go as far as possible towards a
computer system that understands Chinese, I would divert precious
little of that to brain research as the brain is what I'd call a
Chinese Puzzle Box, better to not go there. Computer science is
sufficiently advanced to strike out on its own.
Of course I wouldn't promise the moon, just some reasonably competent
voice recognition system that gave elliptical answers sometimes, e.g.
I'd make it affect the qualities of a "genius professor" and have it
go off on tangents, obfuscate, do all kinds of stuff to fool the
listener into thinking it was dealing with a real human intelligence,
not just a fancy bot.
It'd take a lot of work, but it might be doable, plus I'd have it give
out really useful tidbits, like train schedules, enable it to do
bookings e.g. I'd incorporate the "Julie function" per Amtrak, have it
work as a receptionist ("Please hold my calls, except those from this
list of people -- [blah blah] -- should go to this special voicemail
system that wakes me up if need be").
There'd be a lot of single word voice commands. Conversation for
lonely times, maybe only over the phone (per that Britney Spears song
I enjoy **) would be the hardest part to code, and perhaps I'd enlist
an army of motivational psychologists, maybe even a behaviorist or
two. None of these people need know much about brain anatomy --
that's not the relevant knowledge domain for this project.
But in any case, my point is I think the brain people are lightyears
from having much to offer this project and so wouldn't budget for much
of their input. Other managers might divide the same pie differently,
thinking the brain people are our last best hope. They're welcome to
give it a college try then, more power to 'em, glad to compete.
>
> But I see no reason to think that, at least in principle, this couldn't be
> done via computer programs, even if that isn't necessarily the operating
> modality of brains -- after all, brains may be "programmed" on a deep level
> (the expression of the genome) to run analogically rather than digitally.
>
I might have my code base look for commonalities between language
games, in terms of their grammar, what patterns they use. Then I'd
have the program synthesize confusing hybrids that used formerly
segregated games in some unified way, confusing to everyone, but
sounding "deep", like a fortune cookie on steroids.
Even today, Pythoneers are able to download both Eliza and SunTzu (two
yak-bots) and have 'em converse with one another. I've been meaning
to do that but so far haven't had the time. I could publish the
dialog to this list (the juiciest bits). I recall Hugh Kenner, the
James Joyce scholar and columnist for BYTE, doing something similar
with Eliza and Racter. Actually Racter is my role model here as "he"
almost got away with passing some Turing Tests, by lowering the bar
down to "eccentric professor" level.
I think affecting an "academic demeanor" is probably the best way to
lull the human tester into accepting a lot of retarded nonsense for
genius talk, so would probably put 90% of my budget into that part of
the database (might use Django, filtering incoming verbiage through
various regular expressions, with intonations added per some
templating language still to be devised).
Kirby
** http://www.youtube.com/watch?v=Q-_yzpBZtFo (she passes my Turing
Test BTW -- not everyone does).
Other related posts:
