Why Computers Can't Be Poets
There is no royal road to learning, the ancients said; but in our culture, now, all roads seem, in one way or another, to lead to the computer. The transition from mathematical logic to the computer is an easy and natural one. Ludwig Wittgenstein's protest in his last years that mathematical logic had become a bad influence in philosophy was directed against the threat, as he saw it, that reliance on a linguistic mechanism could replace real insight and thinking. When the logical machinery gets embodied in an actual physical apparatus, the temptation becomes all the greater to let the machine do one's thinking for one. The question, the overwhelming question, then becomes: how far can the machine go in taking over human thought?
The tendency toward materialism is perhaps a permanent one in human nature, and within its limits a valid one. With the advent of the science of mechanics, in the 17th century, the materialistic inclination turns toward mechanism: the tendency to see phenomena everywhere as bits of machinery incarnate. Thus we get in La Mettrie, the 18th-century philosophe, those quaint illustrations of the human body as a system of imaginary gears, cogs, and ratchets. Man, the microcosm, is just another machine within the universal machine that is the cosmos. We smile at these illustrations as quaint and crude, but secretly we may still nourish the notion that they after all are in the right direction, though a little premature. With the advent of the computer, however, this temptation toward mechanism becomes more irresistible, for here we no longer have an obsolete machine of wheels and pulleys but one that seems able to reproduce the processes of the human mind.
Can machines think? This has now become a leading question for our time. It was first proposed in direct and explicit fashion by the British logician and computer expert Alan Turing in 1950. Turing himself, in addition to his gifts as a logician, was an unusual and interesting personality. During World War II he was part of the British team of intelligence experts that succeeded in breaking the German code and thus freeing a good deal of Allied shipping from the menace of the U-boats. After the war he returned to Cambridge, and continued his researches on logic and computer theory. But his life thereafter became rather beclouded and unhappy; he insisted on being an indiscreet homosexual and fell foul of the authorities. In 1954 he committed suicide, at the age of forty-one. For a man whose mind had been continuously engaged with the question of how the machine might guide and regulate life, he seems to have been sadly incapable of managing his own.
Turing's imagination leaped beyond the actual state of the computer to envisage its future possibilities. Writing in 1950, he predicted that in fifty years we would have computers with a storage capacity of 109 bits. Well, we are near the end of that period, and we do already have machines that approach that figure. We are thus in a position to test some of his prophecies about the future range of the computer's operations.
Among other things, Turing claimed that a future computer could very well write poetry. I propose here to center on this claim, because poetry would seem to be one use of the human mind that the machine could not duplicate. Indeed, the creation of a poem would seem to be at the opposite end of the mental spectrum from the additive and combinatorial operations of a machine. Turing, however, imagined a machine that had actually written a poem: and to be specific, he imagined the poem was Shakespeare's sonnet “Shall I compare thee to a summer's day?” Then he put the machine to a test.
For this purpose Turing devised what he called the imitation game. The question “Do machines think?” he held to be too vague; to give the question sense he replaced it by a more behavioral test.
In the imitation game, the machine and human being are placed in a room, and nearby is an examiner, E, who can put questions and receive their answers. For a machine with a suitable storage capacity, Turing holds, the examiner in 70 percent of the cases would not be able to judge which of the respondents was the human and which the computer, so connected and reasonable would the responses of the machine be.
Here are some of the questions that Turing puts to the machine that has just written Shakespeare's sonnet:
Examiner: In the first line of your sonnet, which reads, “Shall I compare thee to a summer's day,” would not “a spring day” do as well or better?
Witness: It wouldn't scan.
Examiner: How about “a winter's day”? That would scan all right.
Witness: Yes, but nobody wants to be compared to a winter's day. . . .
Thus, presumably, by giving such coherent and sensible answers the computer would prove that the poem to which it had given birth was not a freak accident.
Turing's argument moves seductively, but if we pause for a moment, we begin to find it very questionable. In the first place, the question at issue is whether a machine could ever write a poem, and Turing's method of handling this question is to say, let us assume that a machine has actually produced a poem. Then we proceed to test it. In short, he assumes the very point at issue. If indeed (and it is a very big if) a machine had actually produced a poem, then we should expect that it would be able to answer very elementary questions about it. But the question at issue is the ability to produce the poem in the first place.
Then again, the questions Turing suggests are rather curious, for they omit the first and overwhelming question that would arise if a machine actually were to grind out Shakespeare's beautiful sonnet: is this an original poem of yours? If the machine is only quoting, then it is merely disgorging data already fed into it. But if the response is that the poem is indeed its own, and original, we are in for more serious trouble, for we move into the dimensions of style and history, from which poetry cannot escape. The next question would be: why have you written a poem in a style that was valid over three hundred years ago? In other words, we have to deal with the poem not merely as a manipulation of symbols, but as an act of human consciousness within time and history.
It is a curious twist of irony that this same point can be argued against the school of literary critics known as the deconstructionists. Just as for the partisans of the computer a poem is simply the adding of one symbol to another, for the deconstructionists, too, a poem is merely a collection of signs or symbols to be unraveled by the critic from any point or in any direction his ingenuity can supply. It might seem curious that these two groups—the literati of the avant-garde and the somber partisans of the computer—should here converge toward the same attitude. But if we reflect for a moment we shall not find it so strange. We have to recall that for a long time now the labor of a good part of our culture has been reductive: the effort to undermine in one way or another the spiritual status of the human person. And when thinking becomes generally reductive, we can expect that there will be surprising convergences of differing groups. When you dig the pit deep enough, waters from opposite directions will flow down the same hole.
But we have to insist that the poem is not merely a collocation of signs or symbols. If we take poetry seriously, if the experience of poetry is really a part of our life, then we do not merely read single poems. When the poet matters to us, when he really involves us, we read the body of his work—or as much of it as we can manage. The poet himself becomes a kind of spiritual presence in our life, a personality present to us through and within the poems.
Of course, we have to distinguish between this sense of personality and the trivial and accidental features of “personality” that figure in gossip columns. And here the example of T.S. Eliot becomes especially relevant. In his early criticism Eliot spoke against the poet's flaunting of personality. The genuine poet, he said, is one who seeks to escape from personality—and Eliot even uses a much stronger expression: the poet seeks the “extinction” of personality in his poems.
Yet there seems a rather ironic contrast between the critic's pronouncements and his actual performance as a poet. The body of Eliot's poetry, now that we have it all before us, strikes us as the work of a single personality—a unique and individual mind and sensibility. And this unity is there from beginning to end, through the changes of style and tone, through the long journey from despair into the affirmations of faith. It is always Eliot himself, a unique and individual soul, who speaks to us through and within the poems.
Now suppose that this poetry had been produced by a computer. What would the machine have to be capable of in order to produce this particular body of poetry?
It would first have to have a grasp of the contemporary state of the language, of the idiom, that would be vital and charged for modern readers of poetry. To be sure, our language is still English, and in that general sense is the same as Shakespeare's. But the language also changes from generation to generation: different words and rhythms of speech become charged for contemporary ears. Eliot, in his early poems especially, was one of the apostles of modernism, intent on writing a kind of poetry that would not be a repetition of outworn 19th-century idioms and styles. Then, we should have to presuppose in our imaginary computer an intuitive tact, a creative sensitivity, toward the living language. It is hard to see how one could install these qualities of mind in a machine, however vast you make its storage capacity. The writing of a poem is not merely the combination of discrete units of language.
But more than this: there is the relation of the poet to the past, to dead writers and their traditions. This fusion of past and present is one of Eliot's most original and remarkable achievements, both in his poetry and in his critical prose. One cannot, for example, grasp the full resonance of the poetry without some understanding of his critical explorations of the Elizabethans, the metaphysical poets, and certain French poets of the end of the 19th century. And Eliot did not add these influences one to another, like so many discrete units; they were part of an individual sensibility seeking to define itself, and what he saw in these predecessors was something that had not quite been seen before in the same way. His appropriation of the past was also a transformation of it.
Can we imagine a computer capable of even simulating these acts of mind? Make its storage capacity as vast as you wish, we would still need to equip it with some unique historical sense, an ability to see the pastness of the past as well as its presence, and to respond to the piercing actuality of the present as well as to its evanescence. This is a sense of time and history that cannot be achieved by the addition of units of information; otherwise, every encyclopedic pedant would be able to qualify as a creative historian.
Finally, to bring this tedious business to its conclusion, there is the fact of what may loosely be called the poet's development. The poet changes, ages, matures—and sometimes ripens into wisdom. He is, after all, a man of flesh and blood. That is a fact of which the partisans of the computer take too little note in their search for a mechanical substitute for the hu: man mind.
How much of our consciousness is embedded in and inseparable from this fleshy envelope that we are? Certainly it is not the poet's business to write as a disembodied spirit. He falls in love, suffers, and his body ages—sometimes into the ripeness of vision: “Bodily decrepitude is wisdom,” wrote William Butler Yeats, who turned the afflictions of old age into great poetry. But a machine cannot age in this way. Properly speaking, indeed, a machine cannot mature, for it is not an organic body, growing and ripening through time. As a piece of equipment, it becomes used and defective, its wires frayed and its circuits burned out, and shortly ready for the scrap heap. That might be a metaphorical description of some human lives, but only a very nihilistic and reductive one.
I have no intention here of launching a diatribe against the computer as such, a tool that has its valuable and now indispensable uses. My quarrel, rather, is with the fantasies that have taken shape around this instrument. More often than not, the enthusiasts of the computer are unaware that they are speaking from a particular point of view, a particular philosophy, in the light of which they see the whole phenomenon of mind. And this view of mind is not new; we encounter it in David Hume and the British empiricists. This is the view that the nature of our human consciousness is essentially additive and atomistic. Its function consists in combining one discrete datum, or bits of data, with others; and mind itself is but an aggregate of such data.
It is not hard to see why users of the computer should easily fall into this view. They speak, for example, of the storage capacity of a particular machine, meaning the number of discrete units of information that can be fed into it and again extracted from it by the human attendant. And this way of speech becomes congenial when they turn their talk to the human mind. But their own disposition to see facts in a certain way is not merely one more separate datum in the list of facts; it is, rather, a point of view that provides the structure for the whole.
The reality of consciousness as we actually experience it is more than a grocer's list of disparate items. Its presence is more total and engulfing, and it can move backward and forward in time. In memory, for example, my mind may be jogged at times by some isolated fact or facts from the past of which I merely take incidental notice. But there is another experience of memory more total and vital than this. Something comes back to me from the past, from long ago, but it does not remain an isolated factual datum; I am suddenly back in the house where it happened, feeling as I once did, living in that world I once knew. Memories like this reintegrate us into the past. I look through time, and meet myself as I was; I am the same and other. The memory is the emergence into view of that enduring self that I am through time.
If we turn our eyes in the other direction in time, toward the future, we shall also encounter a consciousness that is not, or at least not yet, a mere listing of discrete data. I have, let us say, the vision of a particular project for the future. It comes to me not as an aggregate of ready-made items, but as a whole of which I have as yet only an intuitive grasp and which I must now proceed, with much sweat and toil, to articulate in its details. If our consciousness could not be groping in this way, it would cease to be genuinely creative, and it could not then be the powerful instrument that it has been in shaping human history.