Week 1, Syntax and Semantics

Notes for CS198–79 The Poetry of Computer Science, the Computer Science of Poetry: Philosophy of Computation, UC Berkeley

Background reading: http://www.mirandafield.com/static/classes/triggering-town.pdf

Supplementary Worksheet: https://d1b10bmlvqabco.cloudfront.net/attach/jdf1ibhhu4545d/hz5sgoa7845d1/je0hpd22xjr2/5_poetry_and_vectors.pdf

1.

Today, we’ll talk about syntax and semantics. Syntax and semantics are concepts that occur primarily in semiotics, which is the study of meaning. Syntax in this context means, roughly, the formal or structural relations of signs with other signs; semantics means, roughly, the meaning of those signs. An example of syntax is grammar. An example of semantics is the meaning of the sign “5”, which is supposed to mean the quantity five. In a way, syntax is “objective”, whereas semantics is “subjective”. Syntax is the formal relations between signs, which means that those relations are exactly specified and agreed upon beforehand. Semantics is about meaning, and one might say that meaning is somewhat inherently a subjective thing; two people can see the same sign and experience different meanings.

2.

I bring up syntax and semantics because I want to talk about the Chinese Room argument. The Chinese Room argument is an argument by John Searle, who was a philosophy professor at UC Berkeley for over fifty years. Currently he is “on leave” because he was accused of sexual assault last year. Which brings us to an interesting philosophical question: do we have reason to suspect the philosophy of some philosopher who has done something patently wrong? This is not an isolated issue, either. Heidegger, the towering figure of modern German philosophy, was a literal Nazi; he not only joined the Nazi party when Hitler came to power, but also provided intellectual justification for its policies. There are countless other examples: Hegel provided justification for colonialism, Hume said racist things, and so on and so forth. One response is: yes, if a philosopher’s philosophy leads to wrong acts, then there must be something wrong with the philosophy. Another response says: no, because the philosopher may have interpreted his own philosophy wrong. Yet another response, from the strain of moral relativism, says: who are you to say that the philosopher did something wrong or right? But this seems at least lazy, because wouldn’t we sincerely want a philosophy that enables us to say that killing six million people is wrong?

3.

The Chinese Room argument goes something like this. Suppose there is a man inside a room, let’s call him John, and a man outside the room, let’s call him Bob. John does not understand Chinese. Bob understands Chinese. Bob, however, doesn’t know that John doesn’t understand Chinese. Bob tries to communicate with John by writing Chinese on strips of paper. Now John, again, doesn’t understand Chinese, but he has a little rule book that maps from Chinese input symbols to Chinese output symbols. If the incoming symbols said, in Chinese, “Hello there”, the rule book might map those symbols to “Nice weather, isn’t it?”, in Chinese. So John can simply copy down whatever the rule book says on his strip of paper and give it to Bob. Now, this is Searle’s claim: evven though Bob may get an impression that John understands Chinese, this does not mean that John understands Chinese. All John did was: take some input symbols, flip through his rule book to find where they map to, and write down those symbols. There was no understanding involved. In other words, all John did was: do syntactic manipulations. Syntax, again, means formal relationships between signs. John has a rule book that defines some formal relations between input Chinese symbols and output Chinese symbols. So John’s flipping through the rule book was just syntactic manipulations. There was no understanding, no mental content involved; in other words, there was no semantics involved.

One can distill the essence of this argument in the following way:

1. Computers perform purely syntactic manipulations. (All John did was flip through his rule book and mindlessly find where an input symbol mapped to an output symbol.)

2. Mental contents have semantics. (Presumably, thought has meaning.)

3. Syntax is not sufficient for semantics. Equivalently: semantics is not reducible to syntax.

OK, let’s examine the argument. Number one seems plausible enough: symbol manipulation is in fact all that computers do. Number two seems plausible enough, also: thought has meaning, right? When you eat an apple, that gives you some meaning, right?

N: I don’t agree with that one. After all, what does meaning mean?

OK, great point. So let’s talk about the meaning of meaning. What do you think is the meaning of meaning?

T: it seems like we get meaning by finding patterns. If something “makes sense”. If there’s a reason why something was done.

OK, that’s plausible. So let’s write down:

Meaning is what has reason, i.e. is not random.

But then again, maybe not. Remember, in our first meeting, we talked about worldviews. One of them was determinism; another was a sort of spiritual universe. The first, we said, was meaningless, because everything had a reason. Because we found all the patterns. Then maybe we should write:

Meaning is what escapes reason, i.e. is random.

This is strange, isn’t it? Both make sense, in a way. And both don’t, in a way.

We’ll return to this later. But for now, let’s move on. I think I can safely say: most people, laymen, not philosophers, would agree with number 2 of Searle’s 3-step argument. Number 3, however, is a little more esoteric. What does it even mean, that syntax is not sufficient for semantics? Or that semantics is not reducible to syntax? So let’s prod this notion further.

4.

When we say syntax is not sufficient for semantics, we mean: there is no way to get to semantics through syntax, however much syntax we may have. That is, there is a so-called “metaphysical gap” between them: it is an uncrossable gap, a gap that cannot be bridged by any stretch of the imagination. Now take this with a grain of salt, but we can understand this gap as the gap between the body and the soul. The body, in a way, is syntax; the body is public, it is accessible, it is “objective”. The soul, in a way, is semantics; the soul is private, it cannot be accessed, it is “subjective”. And when we think of the body and the soul, we usually think that the soul is the meaningful part, whereas the body is just some meaningless piece of flesh. So, in a way, to say that there is a metaphysical gap between body and soul is to say that there is a metaphysical gap between the syntax and semantics. But Hugo would beg to differ.

5.

Earlier I said: the sign “5” is supposed to refer to “the quantity five”. But there is at least one easy objection that can be raised towards this idea: the sign, by itself, without anything it references, has meaning. The sign “5”, for example, kind of looks like a fish hook. It kind of looks like a sickle. So maybe that’s what it means. You might complain: well, that’s trivial. Anyone in their right mind would see “5” and think of the quantity five, not a fish hook. But poets are sensitive people. Poets may say that the visual, or more often auditory, effects of signs themselves have meaning, and thus cannot be ignored. Hugo makes a point discussing the following poem in The Triggering Town:

Rattlesnake

I found him sleepy in the heat

And dust of a gopher burrow,

Coiled in loose folds upon silence

In a pit of the noonday hillside.

I saw the wedged bulge

Of the head hard as a fist.

I remembered his delicate ways:

The mouth a cat’s mouth yawning.

I crushed him deep in dust,

And heard the loud seethe of life

In the dead beads of the tail

Fade, as wind fades

From the wild grain of the hill.

Generally, in English multisyllabic words have a way of softening the impact of langauge. With multisyllabic words we can show compassion, tenderness, and tranquility. … In the first four lines of the poem, seven of the twenty-six words … are two-syllable words. This is a fairly high count unless you are in politics. The snake is sleepy. He poses no threat to the speaker.

…

With single-syllable words we can show rigidity, honesty, toughness, relentlessness, the world of harm unvarnished. … In the final five lines the poet kills the snake … and we get no multisyllabic words in the entire passage. (p9)

Hugo says: the number of syllables a word has, which is a purely syntactic feature of the word, has an effect on the meaning of the word, which is a semantic feature of the word. When I say that the number of syllables is a syntactic feature I mean that everyone agrees on how many syllables a word has, that it is, in a way, an objective feature of the word. Even so, it affects the meaning, which seems to be subjective, of the word. Hugo says, in this way, that syntax effects semantics; in this way, the objective world spills into the subjective world.

One might say: well, okay, but that’s a poem. What about more “serious” writings, like news?

I caution against communication because once language exists only to convey information, it is dying.

Let’s take language that exists to communicate — the news story. In a news story the words are there to give you information about the event … By understanding the words of a news article you seem to deaden them. (p12)

In news articles, it seems like the objective does not spill into the subjective, or in any case spills in less than in poetry. In news articles the point is to get some information across, and the fastest way to do this is to use signs that have agreed-upon, “objective” meanings. But Hugo cautions against using language in this way. In other words, Hugo cautions against the strict delineation between syntax and semantics, claiming that “once langauge exists only to convey information, it is dying.” There must be another function of language; what is it? Hugo might say: to explore the space between syntax and semantics, i.e. writing poetry.

Facilitator: Jongmin Jerome Baek (jjbaek.com)