Do Computers "Have Syntax, But No Semantics"?

See also Searle (1982); Searle (1984, p. 33).

In a recent paper, for example, Bozşahin (2018) writes not even about computers "having" syntax/semantics, but about them "being" syntax/semantics. For further examples see Figdor (2009), Ford (2011), Rapaport (2019), Lyre (2020) etc.

Though, as it was pointed out to me by a reviewer of this paper, Oxford English dictionary records an employment of this term, in Northern American Review, already from 1874.

See Posner (1986) for a general historical overview.

Similar arguments were put forward by a number of authors, including Rey (1986) or Rapaport (2000).

Of course, these examples are slightly odd, for a normal computer need not be taught how to add – on the contrary, addition is one of the few things it is able to do "by itself".

This, in effect, is the plot of what Searle famously presented under the name of the "Chines room" (Preston & Bishop, 2002; Searle, 1980) as his challenge to functionalists such as Newell & Simon (1963) (and, indeed, Turing, 1950).

Which, of course, is a concept with a venerable philosophical history: see Jacob (2019).

See also Searle (1983).

This is not to say that this notion must be devious, nor that it is peculiar to Searle. There are certainly other philosophers who want to erect semantics on similar foundations (from Husserl and his followers to Schiffer, 1972; 1987, or Fodor, 1975; 2008). But many semanticists are adamant that semantics is a public business not to be sealed within minds. Thus Quine (1969) urges that language is "a social art we all acquire on the evidence solely of other people’s overt behavior under publicly recognizable circumstances" (p. 26) and therefore "the question whether two expressions are alike or unlike in meaning has no determinate answer, known or unknown, except insofar as the answer is settled by people’s speech dispositions, known or unknown" (p. 29). A similar approach is taken by Davidson (1984) and his followers (Lepore & Ludwig, 2007) and by various exponents of the use theories of meaning to be discussed later (Brandom, 1994; Dummett, 1993; Horwich, 1998) etc. Also partisans of denotational semantics, such as the exponents of post-Carnapian formal semantics (Cresswell, 1973; Montague 1974; Cann 1993; and also some followers of Davidson, like Larson & Segal, 1995) build theories which are not based on intentionality.

Let us leave aside the objection that knowing the Peano axioms is not enough, for our human understanding of the language of arithmetic involves, as shown by Gödel, also having the knowledge of the truth of some sentences that are not derivable from the axioms. If we were to accept the objection, then nobody, save a few mathematical logicians, could ever be said to understand "4" (unless we count also those who "know" the truth of the sentences only because they have been told they are true, which again can then validate the case for computers).

See Rapaport (2000) for a similar argument.

As Dennett (1998, p. 24) puts it: "But, of course, most people have something more in mind when they speak of self-consciousness. It is that special inner light, that private way that it is with you that nobody else can share, something that is forever outside the bounds of computer science. How could a computer ever be conscious in this sense?".

In case of programming languages, the internal semantics has been presented also in the denotational form known from formal semantics of the formal languages of logic and subsequently also of natural languages. See already Gordon (1979).

Rescorla (2012, p. 707) works with a related, though different opposition: "inherited" vs. "indigenous" semantics. While the latter is close to the "internal" semantics of Piccinini and "syntactic" semantics of Rappaport (a semantics that is generated alone by the system), it is opposed not to an "external" semantics, but to one that not only comes from without, but is conferred on the system by some other system.

As Dummett (1993, p. 37) puts it, a theory of meaning is "to present analysis of the complex skill which constitutes mastery of a language, to display, in terms of what he may be said to know, just what it is that someone who possesses that mastery is able to do; it is not concerned to describe any inner psychological mechanisms which may account for his having those abilities".

See Brandom (1994); see also  Peregrin (2014).

See, e.g., Kusch (2006).

This is what Wittgenstein illustrated by his famous "beetle-in-the-box thought experiment"—see, e.g., Stern (2013).

For why this difference matters see Peregrin (2017).

Carnap (1934) claims that language is based on formation rules (the rules of well-formedness) and transformation rules (those of deduction). What I call syntax in the narrow sense amounts to the former only, whereas that in the broad sense comprises also the latter. See also Peregrin (2020).

Another discussion within computer science engaging the concepts of syntax and semantics concerns the very nature of computation and the nature of concepts required for its characterization (Rapaport, 2018; Shagrir, 2020). This discussion, however, is somewhat orthogonal to the current one. In it syntax and semantics are used to distinguish features that are purely formal or structural from those that are a matter of content or have to do with the instantiation of the structure. But though I have doubts that syntax and semantics are the best conceptual tools to resolve this issue, here they are used in the clear sense, not obscuring the problem.