Blog

Not knowing what we don’t know, do we really care?

Good [pause] morning, how may I help you? I’m having trouble with my computer I see. What kind of trouble are you having? It won’t read the CD. I see. The PC won’t read the CD, is that correct? … and off you go into the black hole. For the more sophisticated among us, at some point in such a ...

Robert Dow

dogGood [pause] morning, how may I help you?

I’m
having trouble with my computer

I
see. What kind of trouble are you having?

It won’t
read the CD.

I
see. The PC won’t read the CD, is that correct?

… and off you go into the black
hole.

For the more sophisticated among us, at some point in such a
conversation, you will ask, “Where are you? What time of day is it where you
are?”

Welcome to the Twilight Zone, or is it?

It could be you have entered into a Turing test.

Call centers may be the ultimate Turing test, how would we
know? What question could you ask me, on email or IM, that would prove to you
that I was not a very sophisticated AI? Some people think a joke might do it.
Jokes are one way secret agents test another agent for their authenticity. But,
theoretically, an AI could have a huge database of all, or maybe most, jokes
and their derivatives. (There’s a priest, a rabbi, and a consultant in a boat
…). It’s the same with sports; sports are all about statistics and players’
names, easy database stuff.

robotHowever, with a call center (the verb being “call”) one can
hear the other person, and deduce from that that it’s a human on the other end,
or can you? I called United Airlines to change my ticket. A pleasant-sounding
man asked me questions, took down my frequent flyer number, asked if I wanted
an upgrade. I asked him where he was. He said he didn’t understand my question.
I could see the ambiguity in it, so I asked, what part of the world are in
right now? He said, “I’m sorry, I don’t have that information. Would you like
to start over?” I said “Go stuff yourself.” He said, always pleasantly, “Hmm,
let me get an agent for you.” When the agent came on, I said, I want the name
of that last guy I was speaking to. She said, “I’m sorry, I don’t have that
information.” I hung up, they were smarter than me; but we weren’t IM’ing or
email’ing, we were speaking.

I then called Delta Airlines. I listened to some music and
then a lady came on with a really strong Southern accent. I said, where are you?
She said, “I’m sorry, we’re not allowed to give out that information.” I said,
“What time is it where you are?” She said, “May I help you?” I said, “I want to
know what time it is where you are.” She hung up.

Searle’s Chinese Room

Now I was confused and wondered if I had entered John
Searle’s Chinese Room, or if I was looking for, or may be at, Schrödinger’s
cat.

Searle argues that emulating the functional behavior of the
brain, or some part of it, is insufficient grounds for attributing a machine or
computing device with cognitive states. Ah, but what the hell is a cognitive
state? Who knows? Some folks with MRI machines are trying to find out, but,
basically, it’s whatever you want to call it. Is speaking a cognitive state? Do
lower form animals like dolphins and monkeys speak? Are they cognitive? Does a
computer speak? You see the problem?

So when I ask, “Where are you?” or “What time is it?” I
might get a more reasonable answer from a robot than from a human, and with
such information, can I tell which is which?

Probably not. So then, why do I care? It’s one of the other
great paradoxical arguments: tell me what you don’t know. That’s supposed to
put a robot into brain meltdown. But guess what? That, or any other paradox, if
not loaded in some paradox database, will shut down any human or robot. The
only answer to such a question is another question—either that or
meltdown—unless of course you are a politician, who will never answer any
question, and so we’ll never know if politicians are cognitive, human or
otherwise; but I digress.

But the real question is why do we care?

I think it’s because we don’t want to be defeated by a
machine. Whether it’s our car, a call center, or some deadly AI in a game. “No
damn machine is going to beat me”—Garry Kasparov 1996. On May 11, 1997, Deep
Blue defeated Kasparov in a six-game match held in New York.

“I never considered Deep Blue intelligent in any way. It’s
just an excellent problem solver in this very specific domain”—Murray Campbell,
Deep Blue team member.

So it’s not about if the opponent is human or not, it’s
about losing to a machine. But, if we don’t know it’s a machine, do we really
care? If you got creamed in a game, and then learned that the opponent was an
online player, would you feel as bad if it was just AI? Probably not.

But wait a minute—what’s the difference? After all, some
human (or at least we think so) programmed the AI. So in fact you weren’t
beaten by an AI, you were beaten by a human’s proxy.

There now, does that make you feel a little better?

Reliving the past