[Wittrs] Mirsky on Watson and Seale
kirby.urner at gmail.com
Tue May 22 18:53:29 CDT 2012
On Tue, May 22, 2012 at 1:17 PM, SWM <swmirsky at gmail.com> wrote:
>> These are not especially "metaphoric" nor "anthropomorphic" uses, just
>> more everyday operations with our tools of expression.
> No dispute about the range of uses. I will just restate here that I am using the terms as we use them with regard to ourselves, i.e., that they denote whatever it is we think we are denoting when talking about our own mental lives and the mental lives we read in others' behaviors.
A problem in getting a handle on Watson's mental life is some
empiricists may suspect a subconscious feedback loop wherein the
desire of the experimenter to tilt the needle, as it were, actually
gets picked up within the mechanisms and biases the results, a kind of
This is perhaps easiest to control by making sure Watson looks at
random cards, states what he sees, and no other humans are witnesses
at the same moment.
Of course card reading is easy and lots of machines do it, but Watson
might go on to guess the painter or perhaps conjecture as to where the
photograph was taken.
But is it right to call it "a conjecture" just because it might be
wrong, or must there be some mental process?
In mathematics, you don't necessarily know where the conjecture came
from, man or machine, so that means Watson might be said to conjecture
even today. Chess playing programs look ahead.
>> Words are devices, best *not* imagined pointing sticks with their
>> "meanings" at the far end (the bewitching view).
> No reason to even point that out. No one here is doing that. I am using the two words in the way I describe above. That's the whole of it.
OK. As long as you're not imagining mental states as like fish, with
you standing on the shore, skewering them with words like "thought"
That's a funny picture. Some people, when they think "denote" have
this idea that they're actually jabbing a sharp stick through an
Thanks to our philosophical training, we don't have to put up with
that mental picture if we find it misleading in some way.
>> In the age of "corporate persons" it's hard to have it both ways and
>> say "IBM can't be proud of Watson because IBM has no feelings".
I was just imputing mental states to a subculture, which I think is
grammatically OK (people do it a lot).
The individual human need not be considered the repository of every
eddy in the zeitgeist.
Sometimes the picture only becomes more clear as you zoom back and see
Individual fish do not a school make.
I heard 7% of Greeks are voting Nazi if they vote. Is that true?
What's the mental state there?
Can we speak of the psychology of nations and even a mental process of
a kind? I think we do that, and call it psycho-history (no wait,
that's Asimov -- call it intellectual history).
I'm just thinking Watson's mental life might be more like that of a
nation-state, wherein we're talking about shifting weights, alliances,
majorities, genetic algorithms, agents.
You could build the layers like you would a city, thinking in urban terms.
Before you know it, Watson is a pulsing electronic metropolis, setting
trends in music and fashion much as London does, or Rome.
Another world capital, that Watson (renamed ZION by IBM) with new
polling numbers always coming in from around the world for number
crunching -- no wonder she seems to mirror our own thinking.
You can't expect Watson to develop intelligence in a vacuum, any more
than we do.
Feedback loops must be closed.
The solo individual is insufficient for an awakening to self
awareness. Watson will need us to help nudge him along, if he is to
partake of our humanity, our sentience.
>> We give personhood to dolls, to cartoon characters. Bugs Bunny thinks
>> and feels. So why not Watson? Why not God?
> Why not? But I am not proposing we use the words in a different way that is somehow special to Watson-like machines or systems. I am proposing that we use the words in relation to such things as we use the words in relation to ourselves and that, in doing so, we are speaking about a real issue, addressing a genuinely interesting question which can be answered in a scientific way.
Of course you're talking about an average self.
Some selves consider themselves in conversation with other mental
beings or will imagine different parts of their brain each having
voices, meaning they live in a kind of dollhouse or soap opera of
contending brain parts.
That's not average (not sure what is though -- not claiming I'm
average myself), but it might be how we need Watson to conceive of
himself, if he's to be at all accurate with his mental model.
We may find we need to build "hemispheres" because the need for real
time feedback of an "other" (at least one) that's similar yet
different and getting a lot of the same information.
Quasi-duality (asymmetric complementarity, chirality) is actually
critical to any really interesting mental computation -- that might be
Watson may have the mind of a married couple, always arguing with
That sounds like a metaphor but there may be no more direct we to say
it. "Duality begets Self" might be the slogan ("Adam, meet Eve. Eve,
> Again, this does not go to the usage I was employing.
>> "Think" and "feel" are not names for species of bird that we find
>> using our introspective binoculars. They are shuttlecocks weaving
>> through the tapestry of our language, meshing, but not pointing.
> Same point. I've made my usage plain here.
Maybe. You often seem to nod nod wink wink to yourself about what we
all "commonly mean" by various words. I sense some overlap in our
respective usage patterns.
>> So those weren't my two choices.
>> > My point in the article was that they may well be nothing more than a certain kind of information processing, contra Searle, in which case, if incorporated into a Watson-type system, there would be nothing left out.
I'm not sure at what point we'd say Watson had developed an
earthworm's level of consciousness. That begs the question how
conscious is an earthworm and we don't really have ways of quantifying
that, although with funding, we might develop in that direction.
My point being: I consider our way of talking rather fragile and not
easily / simply "extended" to cover machine consciousness, not without
breakdown of the machine-biological divide more generally.
If machines become more worm like then worms will become more machine like.
I'm not saying that would be bad or won't happen, just that we don't
have the means to go forward at all quickly.
Mixing cellular and electronic components in the same device seems a
likely direction in which to continue.
What if Watson were an array of 80K rat brains that could seem to
predict political election outcomes with significantly greater than
chance probability? Is this Watson "thinking about" elections?
>> Not clear what this would even mean. I process information using
>> these words (along with many others). I do work in my office. That
>> doesn't mean these words "are" information processing. They help with
>> the work, as do staples and paper clips.
> It's to the question of how brains do what they do, what consciousness, the state of being conscious, amounts to.
Brains seem to need community, according to Dr. John Cacioppo.
The high bandwidth connections between them suggest they're not
designed to work by themselves, though some can if they have to, at
least for awhile.
Watsonites should probably go with a multi-brain model and immediately
replicate core constructs with partial mirrors (not perfect mirrors,
which'd be redundant).
>> I could draw the comic book version, but I don't think there's a
>> computer scientist alive who could get the layers right.
> Maybe. On the other hand, you never know. Maybe someone will at some point and then all the debate will go away. No, wait, then we can just debate about what it means to get the layers right!
Yeah, I don't think an 80K rat brained Watson that predicts elections
and sets trends in music would satisfy the debaters.
Does it feel? Does it think?
>> If I were a Foundation, I'd be leery of funding such experiments as
>> they could only be part of a plot to bilk me out of my money.
> Like lots of projects I expect. So what else is new?
What you can get funding for are programs designed to change states of
Once consciousness is in the picture, changing it or influencing it is
what money is for, basically.
Money provides an incentive (as an eminently tradable good) to get
people thinking and acting in such and such a way.
Inducing consciousness in an inanimate device begs the question of
what are the criteria of success.
Clearly if Watson can pass the Turing Test we've come a long way (he can't yet).
But surely there'd be levels in between. It can't be all or nothing.
>> The projected scene is reminiscent of when spiritualists would shake
>> down widows for research into breakthrough devices to contact the
>> dead. Wasn't Thomas Edison involved in that business?
> Don't know about that but certainly Chris Columbus had a hell of a time finding investors -- and he was only talking about sailing a ship to the other side of the world.
His backers wanted gold and stuff, so he had to be really cruel to the
natives if they didn't produce it in sufficient supply.
Lots of blood chilling stories about the guy, a horror movie for sure.
By today's criteria, I'm guessing he'd be considered severely
psychotic but that's speculation.
Hard to retroactively judge by the standards of another civilization.
More information about the Wittrs