Read The Most Human Human Online
Authors: Brian Christian
When I talk to friends of mine who are actors, they say that this is more or less the answer that the actor in the long-running show must find, in his or her own way. How do you deviate? How do you make it a new show?
It’d be tempting to think that you spend a certain amount of time learning what to do, and the rest of the time knowing what you’re doing, and simply doing it. The good actor will refuse to let this happen to him. The moment it does, he dies. A robot takes his place.
I think of neoteny, of my younger cousin, age four or so, careening into walls, falling over, getting up, and dashing off in the next direction. Of the fact that kids are much quicker studies at learning to ski because they’re not afraid to fall. Fail and recover.
For the architect, it’s site-specificity; for the actor and the musician, it’s
night
-specificity. My friend Matt went to see a songwriter that he and I admire a lot, and I asked him what the show was like. Matt, unenthusiastic, shrugged: “He has a set list, and, you know, he plays it.” It’s hard to imagine what the artist
or
the audience gets out of that. A great counterexample would be a band like Dave Matthews Band, where one night a song is four minutes long and the next night, twenty. I admire the struggle implicit in that—and the risk. There must be this constant gravity toward finding what works and sticking with it, toward solidifying the song, but no, they
abandon
what works as righter than it will ever be again, or come back to it the next night, but only to depart and try something else that could well fail. This is how you stay an artist as you grow up. And both they and the fans get a once-in-a-lifetime thing.
I suppose when you get down to it, everything is always once in a lifetime. We might as well act like it.
I saw my first opera this past year:
La Traviata
, starring soprano Nuccia Focile in the lead role. The program featured an interview with her, and the interviewer writes, “It’s those unexpected moments that blindside a singer emotionally, Focile feels. In performance, a different phrasing of a word can suddenly take the involved singer by surprise and make her gulp or blink away tears.” Focile seems to think of these moments as hazards, saying, “I must use my technical base to approach certain phrases, because the emotion is so great that I get involved too much.” As a professional, she wants to sing consistently. But as a human, the fine attention to and perception of the tiny uniquenesses from night to night, the cracks in technique where we
get
involved, get taken by surprise, gulp, feel things freshly again—these are the signs we’re alive. And our means of staying so.
1.
“Her” name is an allusion to Eliza Doolittle, the main character in George Bernard Shaw’s 1913 play
Pygmalion
. Inspired by the myth of Pygmalion, a sculptor who creates a sculpture so realistic he falls in love with it (which also inspired, among many other works,
Pinocchio
), Shaw’s play (itself the inspiration for the musical
My Fair Lady
) takes this idea and makes it into a tale of fluency and class: a phonetics professor makes a bet that he can train the lower-class Eliza Doolittle in the spoken English of the aristocracy and have her pass as a noble—a kind of Turing test in its own right. It’s easy to see why Weizenbaum drew from Shaw’s
Pygmalion
to name his therapist; unfortunately, he ended up with something closer to Ovid’s story than Shaw’s.
2.
The only difference, which may be an important one, is that the book’s boundaries and
extents
are clear. If you read it front to back, you know exactly which areas it covers and which it doesn’t. The bot’s extents are less clear: you must, probing the bot verbally, find them. One can imagine a bot that contains a useful response that the user simply doesn’t know how to
get
to. Early “interactive fictions” and text-based computer games sometimes had this problem, nicknamed “guess-the-verb” (for instance, a 1978 game called
Adventureland
required the user to somehow know to type the ungrammatical command “unlight” to extinguish a lantern). It might be fair to say that therapy bots are to therapy books what interactive fictions are to novels.
3.
Glenn Murcutt, whom we’ll hear more from later in this chapter: “We are taught that creativity is the most important thing in architecture. Well, I don’t believe that. I think that the creative process leads to discovery, and discovery is the most important thing.”
4.
In fact, the etymology of “sabotage,” which comes from the French word
sabot
, meaning a type of wooden clog, is said (perhaps apocryphally) to be rooted in a story of workers destroying mechanical looms by throwing their shoes at, and into, them.
5.
See also, for instance, the Army chatbot SGT STAR, who gives very Army-like yet strangely endearing answers like the following: “Brian: What is Delta Force? SGT STAR: Delta Force is rumored to be the Army’s counter-terrorist operations unit. In line with operational security we cannot discuss, nor ponder the unit’s abilities.” He even admonishes you in an Army-like (yet strangely endearing) way: “Brian: Wow, that sounds pretty badass. SGT STAR: OK Private, watch your language or I’ll have to shut you down.” The pathetic fallacy be damned—er, darned—I couldn’t close the window without apologizing. “Sorry about that.” “That’s OK,” he said, and I felt better.
6.
There’s a very real downside to this particular style of luxury. Princeton psychologist Daniel Kahneman notes that arguments between couples are worse in luxury cars than in crappy cars,
precisely
for the things that they’ve paid top dollar for about the car. It’s soundproof, so the noises of the world don’t get in. It’s comfortable, it runs smoothly and quietly, the suspension treats you gingerly. And so the argument goes on and on. Most disagreements are not 100 percent resolvable so much as they can be converted into more or less satisfactory compromises that then assume a lower priority than the other issues of life. They are terminated more from without than from within. Carl Jung puts it nicely: “Some higher or wider interest appeared on the person’s horizon, and through this broadening of his or her outlook the unsolvable problem lost its urgency.” Interruptions can be helpful.
7.
I do tend on the whole to think that words mean most when they’re composed freshly. E.g., that the sinner’s improvised confession in the booth does more, means more, than the umpteen un-site-specific Hail Marys he’s prescribed and recites by rote.
Success in distinguishing when a person is lying and when a person is telling the truth is highest when … the lie is being told for the first time; the person has not told this type of lie before
.
–PAUL EKMAN
For Life is a kind of Chess …
–BENJAMIN FRANKLIN
Entering the Brighton Centre, I found my way to the Loebner Prize competition. Stepping into the contest room, I saw rows of seating where a handful of audience members had already gathered, and up front what could only be the bot programmers worked hurriedly, plugging in tangles of wires and making the last flurries of keystrokes. Before I could get too good a look at them, or they at me, the test’s organizer this year, Philip Jackson, greeted me and led me behind a velvet curtain to the confederate area. Out of view of the audience and the judges, four of us sat around a table, each at a laptop set up specifically for the test: Doug, a Canadian linguistics researcher for Nuance Communications; Dave, an American engineer working for the Sandia National Laboratories; and Olga, a South African programmer
for MathWorks. As we introduced ourselves, we could hear the judges and audience members slowly filing in, but couldn’t see them around the curtain. A man flittered by in a Hawaiian shirt, talking a mile a minute and devouring finger sandwiches. Though I had never met him before, I knew instantly he could only be one person: Hugh Loebner. Everything was in place, we were told, between bites, and the first round of the test would be starting momentarily. We four confederates grew quiet, staring at the blinking cursors on our laptops. I tried to appear relaxed and friendly with Dave, Doug, and Olga, but they had come to England for the speech technology conference, and were just here this morning because it sounded interesting. I had come all this way just for the test. My hands poised hummingbird-like over the keyboard, like a nervous gunfighter’s over his holsters.
The cursor, blinking. I, unblinking.
Then all at once, letters and words began to materialize—
Hi how are you doing?
The Turing test had begun.
And all of a sudden—it was the strangest thing. I had the distinct sensation of being
trapped
. Like that scene in so many movies and television shows where the one character, on the brink of death or whatever, says, breathlessly, “I have something to tell you.” And the other character always, it seems to me, says, “Oh my God, I know, me too. Do you remember that time, when we were scuba diving and we saw that starfish that was curled up and looked like the outline of South America, and then later, when I was back on the boat and peeling my sunburn, I said that it reminded me of this song, but I couldn’t remember the name of the song? It just came to me today—” And the whole time we’re thinking,
Shut up, you fool!
I learned from reading the Loebner Prize transcripts that there are two types of judges: the small-talkers and the interrogators. The latter are the ones that go straight in with word problems, spatial-reasoning
questions, deliberate misspellings … They’re laying down a verbal obstacle course and you have to run it. This type of thing is extraordinarily hard for programmers to prepare against, because anything goes—and this is (a) the reason that Turing had language, and conversation, in mind as his test, because it is really, in some sense, a test of everything, and (b) the kind of conversation Turing seemed to have envisioned, judging from the hypothetical conversation snippets in his 1950 paper. The downside to the give-’em-the-third-degree approach is that there’s not much room to
express
yourself, personality-wise. Presumably, any attempts to respond idiosyncratically are treated as coy evasions for which you get some kind of Turing test demerits.
The small-talk approach has the advantage that it’s easier to get a sense of who a person
is
—if there indeed
is
a person, which is, of course,
the
if of the conversation. And that style of conversation comes more naturally to layperson judges. For one reason or another, it’s been explicitly and implicitly, at various points in time, encouraged among Loebner Prize judges. It’s come to be known as the “strangers on a plane” paradigm. The downside of this is that these types of conversations are, in some sense, uniform: familiar in a way that allows a programmer to anticipate a number of the questions.
So here was a small-talk, stranger-on-a-plane judge, it seemed. I had this odd sensation of being in that classic film/TV position. “I have something to tell you.” But that something was … myself. The template conversation spread out before me:
Good, you? / Pretty good. Where are you from? / Seattle. How about yourself? / London. / Oh, so not such a far trip, then, huh? / Nope, just two hours on the train. How’s Seattle this time of year? / Oh, it’s nice, but you know, of course the days are getting shorter …
And more and more I realized that it, the conversational boilerplate, every bit as much as the bots, was the enemy. Because it—“cliché” coming from a French onomatopoeia for the printing process, words being reproduced without either alteration or understanding—is what bots are made of.
I started typing.
hey there!
Enter.
i’m good, excited to actually be typing
Enter.
how are you?
Enter.
Four minutes thirty seconds. My fingers tapped and fluttered anxiously.
I could just feel the clock grinding away while we lingered over the pleasantries. I felt—and this is a lot to feel at “Hi, how are you doing?”—this desperate urge to get off the script, cut the crap, cut to the chase. Because I knew that the computers could do the small-talk thing; it’d be playing directly into their preparation. How, I was thinking as I typed back a similarly friendly and unassuming greeting, do I get that lapel-shaking,
shut-up-you-fool
moment to happen? Once those lapels were shaken, of course, I had no idea what to say next. But I’d cross that bridge when I got there. If I got there.
The biggest AI showdown of the twentieth century happened at a chessboard: grandmaster and world champion Garry Kasparov vs. supercomputer Deep Blue. This was May 1997, the Equitable Building, thirty-fifth floor, Manhattan. The computer won.
Some people think Deep Blue’s victory was a turning point for AI, while others claim it didn’t prove a thing. The match and its ensuing controversy form one of the biggest landmarks in the uneasy and shifting relationship between artificial intelligence and our sense of
self. They also form a key chapter in the process by which computers, in recent years, have altered high-level chess forever—so much so that in 2002 one of the greatest players of the twentieth century, Bobby Fischer, declared chess “a dead game.”
It is around this same time that a reporter named Neil Strauss writes an article on a worldwide community of pickup artists, beginning a long process in which Strauss ultimately,
himself
, becomes one of the community’s leaders and most outspoken members. Along the course of these experiences, detailed in his 2005 bestseller,
The Game
, Strauss is initially awed by his mentor Mystery’s “algorithms of how to manipulate social situations.” Over the course of the book, however, this amazement gradually turns to horror as an army of “social robots,” following Mystery’s method to a tee, descend on the nightlife of Los Angeles, rendering bar patter “dead” in the same ways—and for the same reasons—that Fischer declared computers to have “killed” chess.