ROBBY the robot did it. So did HAL, the murderous computer from 2001: A
Space Odyssey, and C3PO, the golden fool from Star Wars. Even
C3PO’s vacuum cleaner-like sidekick R2D2 managed it, as did Gort, the towering
automaton from The Day the Earth Stood Still—though they both hid
their lights under bushels.
The “it” in all this is the ability to understand language, a sophisticated
brain function that, some cognitive scientists say, lifts humanity above lesser
creatures.
If you hadn’t noticed that these robots were so sophisticated, it could be
because they were all given the strangest of vocal mannerisms. Robby, who
appeared first in Forbidden Planet, spoke in very artificial tones, HAL
sounded constantly surprised and C3PO could not stop its stream of hysterical
nonsense. R2D2 communicated in whistles and burps while Gort failed to utter a
single word.
Advertisement
Which is all rather strange in such advanced machines: each of them was at
some time entrusted with caring for human life (in Gort’s case, only the voice
command “Gort! Klaatu barada nikto!” stopped it destroying all life on Earth).
Let’s face it, with the right interface, it seems likely that all of them would
have passed the ultimate test of computer smarts—the Turing test. That is,
they would all have given human-like answers to any questions a human might
ask.
Things are rather different outside Hollywood, where computerised machines
are still struggling to understand even the simplest verbal communication. And
so, the instant IBM promised it was possible to “compute your way through the
day with just your voice and VoiceType”, its speech-recognition software, I
bought the program, primed for intelligent conversation with my computer.
Intelligent, that is, after I had been properly trained. Even human
linguistic ability can be stymied by our wild variety of accents and rhythms.
Rather than accommodate this diversity, VoiceType tries to regularise human
delivery. It taught me to speak “isolated word speech” into its microphone,
saying each… word… with… a… definite… and… perceptible… pause…
afterwards. By the time I graduated from speech-recognition boot camp, I was
droning out words in approved robotic style.
Talking like Robby the Robot was just fine when I used VoiceType for brief
commands such as “Go to Media Player”, which brought up that function on the
computer screen. But a measured delivery was harder to sustain while dictating
longer passages to create a text document. Still, as I repeated that famous
alphabetical incantation, “The quick brown fox jumped over the lazy sleeping
dog,” the words appeared on the screen just as spoken.
All well and good. Then I began wondering whether the computer could fathom
speech characteristic of humans rather than machines—the allusive,
rhythmic language of poetry. Not wanting to blow the computer’s mind, I started
it at the low or doggerel end. First, I recited a childhood favourite: “Roses
are red, violets are blue, with a face like yours, I’d live in the zoo.”
VoiceType handled it well, although changing “violets” to “violence” was a bit
of a slip.
But then hubris set in. Like Frankenstein, I violated the natural order: I
deliberately introduced poems both romantic and mystical, the very antithesis of
machine style, into my computer’s silicon brain. I really felt the difference as
I recited works by William Blake and Samuel Taylor Coleridge, for it was a
struggle to keep from breaking out of the required robotic delivery and into the
powerful rhythms the poets had intended.
The visionary poet and artist William Blake was no great fan of science. In
1795, he made a print showing a naked Isaac Newton imposing a distasteful
rationality on the world. But science has now got its revenge—in what
VoiceType did to Blake’s The Tyger. The original reads:
Tyger Tyger, burning bright,
In the forests of the night;
What immortal hand or eye,
Could frame thy fearful symmetry?
IBM’s version—care of my voice—runs somewhat differently:
Tiger, tiger, turning right
In the virus of the night
What immortal and worldwide
Could flame thy fearful salmon tried.
Giving credit where it’s due, I must say that “virus of the night” could be a
dark metaphor for disease, more poetic than what most would-be poets declaim at
the local coffee shop. But what on earth is that “turning right” business? Could
it be a conservative political message? And try as I might, I can’t connect a
nervous fish or “fearful salmon” with the unutterable power of God’s creation.
So, despite that one viral image, I awarded the program a C-minus in
comprehension.
Still, I couldn’t give up just yet. So I tried another mystical work.
Coleridge wrote Kubla Khan while immersed in the vestiges of an opium
dream. And the poem does suggest an altered state of consciousness. It begins
like this:
In Xanadu did Kubla Khan
A stately pleasure-dome decree:
Where Alph, the sacred river, ran
Through caverns measureless to man
Down to a sunless sea.
That wondrous name, “Xanadu”, totally baffled VoiceType. In different
readings, the computer turned it into ” seventy-two” or “salmon do”,
mysteriously rediscovering the fish theme from The Tyger.
Nor could VoiceType get a handle on the name “Kubla Khan”, rendering it as
“mobile bond” or “global kind”. So the first line was ruined, and that evocative
river Alph also fared badly, becoming “alcohol”, “health” or “Alvin”. A sacred
river of health or alcohol makes a kind of demented sense, but on my list of
exotic names, “Alvin” scores pretty low. VoiceType’s efforts here earned only a
D-minus.
All this reminds me of science-fiction tale about a human explorer who
crashes on a hostile planet. He finds a shelter with food, air, even
entertainment, but it was built by and for an alien race, and is harmful to him.
Facing certain death, the explorer falls unconscious, only to be roused by the
delicious smell of cooking, while he breathes fresh air and hears lovely music.
Has the shelter changed to accommodate him? No. When he looks down at his body
he sees the automated refuge has transmuted him into the alien form it was meant
to serve.
The Turing test is, after all, a two-way street. As I represented humankind
by reciting poetry, the computer represented machinekind by making me talk like
a robot. So while I enhanced its humanity, it diminished mine. And heaven help
us, I suspect that long before my PC types “Tyger Tyger, burning bright”, I’ll
have come to understand just what it means when it says “In seventy-two did
global kind a steeply pleasure-dome decree”.
