All Things Are Full of Gods, page 29
Then, of course, there’s the nonsense about the “intentional stance.” Well, surely I shouldn’t have to point out to you that any such a stance would itself be a specimen of intrinsic intentionality. I’m not even sure this rises to the level of a good logical circle, it’s so very, very inane. I might be able to extend my sense of my own intentionality to some inanimate object, as a result of the pathetic fallacy perhaps; but I certainly can’t attribute intentionality to myself except as an intrinsically intentional act.
HEPHAISTOS: It’s turtles all the way down, then?29
PSYCHE: If you like. I’ve already said as much, haven’t I? Let’s call it the “testudinal principle”—the principium testudinis: mind in any of its aspects—unity, subjective consciousness, intentionality, rationality, what have you—can rest upon no foundation but itself, and there’s no way into that self-subsistence from outside. That’s why Dennett is forced to use rhetorical sleight of hand—albeit not very deftly—to make his arguments seem cogent. Note how he extracts a reductionist account of “belief” from his little parable about the robot precisely by first introducing the notion of belief into his description of the robot’s functions, like a dove concealed in a magician’s hidden pocket, so that, having first rhetorically animated the machine, he can then reverse his tropes and transform the computer into a metaphor for the inanimate mechanisms he thinks minds to be. After all, so his argument goes, if a computer can be programmed to “believe” it has a visual experience . . . and so on and so forth. Except, of course, it can’t. Computers have no beliefs, whereas we most definitely do. Moreover, we don’t merely “judge” and then “report” to ourselves—whoever those selves might be—that we’re conscious after a sequence of unconscious judgments—whatever that means. Rather, we’re directly, infallibly conscious of being conscious, and for precisely that reason we’re capable of making judgments—epistemological, moral, propositional, and so on. The sākṣi—the witness—within each of us is a pure subjectivity, a pure awareness of being aware, and were that not so we too would be like machines, entirely devoid of mind, judging nothing, reporting nothing, believing nothing. This isn’t matter for any kind of meaningful debate. It’s directly and indubitably self-evident. We may make epistemic judgments regarding perception, but we don’t make phenomenal judgments regarding experience. We possess the phenomenon in se, without need of any propositional supplement, as a private experience that’s irreducible to the functional information available to our neurology. Neural signals may convey all the sensory data there is to convey; but only the subjective mind possesses the phenomenal impressions.
Actually, the analogy is defective at both poles, because computers not only have no beliefs and make no judgments, but also because they don’t “report” that they do, even if we’ve programmed them to produce reports to that effect—any more than an abacus with varicolored beads “reports” that it “believes” that (2 × 10) + 5 = 25. Computers can’t tell themselves they see colors, nor can they tell us. They merely process code that, for them, never comes to possess any semeiotic content whatsoever. It is we alone, through them, who make reports of their functions to ourselves, in a semantic form that possesses content only in a hermeneutical space belonging to minds, into which no machine process can ever enter. And it is we alone who make judgments; computers can’t. Judgment is a deliberative act based on a discriminating knowledge that also exists only in the hermeneutical realm. Computers don’t even employ words; they generate electrical operations that we’re able to understand as words because, unlike computers, we possess minds. And, of course, it doesn’t matter whether the computer is creating visual representations on a screen or not, because representation too is a matter of intentionality. Purely physical systems can translate physical realities into different kinds of data: a camera, for example, stores patterns of light and color in an analogue or digital form, and a recording device can do the same thing with sounds, but nothing that either device produces constitutes a representation of anything at all unless there’s a subjective consciousness present directed toward the picture or the recording as representing something beyond itself, and toward the thing depicted as represented, all of which is an act of awareness—of subjective attention. When Dennett tells us that, for all we know, we might be “zombies” who only believe we’re conscious, he’s talking twaddle. Belief isn’t data; judgment isn’t an unconscious operation of physiology. Each requires an intention and also an immediate affect, an actual “what it’s like”: to believe is to experience oneself as believing; to judge is to experience oneself as thinking. A “zombie” could never experience thought, could never be aware of being aware, and so couldn’t think—just as, in fact, no computer has ever entertained a single thought of any kind.
HEPHAISTOS: [After several moments of pensive silence:] As it happens, I do want to discuss computational models of the mind. You’ve made a great many assertions, after all, as to why they’re defective. I’m going to require a more substantial argument before I concede the point. But the day is passing swiftly.
PSYCHE: Yes, it is. Tomorrow, then. We can . . .
HERMES: Wait, please, let’s not dissolve our little parliament yet. I want to know whether our friend Hephaistos is now ready to grant that subjective experience really is what it is—really is subjective consciousness with its own intrinsic properties, irreducible to mere physical facts.
HEPHAISTOS: [With a sigh:] You’re unrelenting. [Considering for a moment:] What I’ll acknowledge is that Dennett’s arguments to the contrary are circular, as you say, and largely unpersuasive as stated. And yet I find myself unable simply to grant that anything just is as it seems, without absolutely ruling out every other possible angle upon it. Something still tells me that it’s not wholly implausible to say we’re essentially dealing with information states that, through a certain structural arrangement of relations, can generate an appearance of intrinsic subjective experience, above and beyond the information itself.
HERMES: An appearance to whom?
HEPHAISTOS: To themselves, perhaps. I don’t know. Whatever the case, computers are, after all, very impressive analogies of minds, it seems to me. As I said some hours ago, however, I must demonstrate either that mental phenomena can, despite all appearances, be described mechanistically or that matter can, without any violation of physicalism, be described in a non-mechanistic way. If you convince me that the computational analogy fails, I’ll abandon the former task and take up the latter. But first you must convince me.
HERMES: And if . . . ?
HEPHAISTOS: Tomorrow, however. Let’s say around teatime.
[Exeunt omnes.]
DAY FOUR
Machine and Soul
I
Language, Thought, and Code
Late afternoon, verging on early evening. The angle of the Intermundian sun’s rays is already markedly oblique. The breeze, coming from the direction of the rose arbors, is especially delectably fragrant. PSYCHE, EROS, HERMES, and HEPHAISTOS are now seated at the garden’s center in chairs of wrought-iron filigree embellished with ingenious floral designs, arranged around an elegant but intimate table whose circular top is of blue marble resting on an ornate base of the same dark iron. The tea-things have just been cleared away—or, rather, waved away into nonexistence—and everyone seems quite at ease.
HEPHAISTOS: That was delightful, as the hospitality of this house always is. I can’t say I’m especially eager at the moment to resume our conversation of yesterday, but I expect I should. Or would someone else like to start?
PSYCHE: No, you go ahead, please. You’re still the chief inquisitor here.
HEPHAISTOS: I’m not sure I relish that title, but thank you. Yes, then. Remind me, did we discuss the “language of thought” hypothesis of mind?
HERMES: Yes. We talked about it in relation to intentionality. You made it sound like just another version of functionalism?
HEPHAISTOS: Similar—aligned, perhaps—but with certain distinctive features. It’s not an especially interesting theory in general outline, because all it really says is that there’s a structural continuity or similarity between reasoning and speaking, which is hardly an earth-shaking proposition. What’s important about it, and what makes it relevant to computational models of mind, is what it says about language, and therefore what it says about mental states. In a sense, it’s a kind of reverse reductionism: not of language-use to physical events, but of physical events to the structure of language—but, of course, on the premise that language already exhibits properties that aren’t, as such, mental.
HERMES: Then it’s a false theory. All language-use is intentional and conscious.
HEPHAISTOS: You’re somewhat missing the point. I’m not talking about the use of language, but of its basic structure. Yes, when I say, “That was a lovely tea,” I’m aware of what I’m saying and why I’m saying it, but that’s all in a sense consequent upon linguistic functions that precede any intrinsic intentional content. Language is composed of symbols, such as phonemes or graphemes, which in isolation have no essential properties beyond their physical embodiments; they mean nothing except by virtue of their relations to other things—other symbols, objects, processes. There’s nothing actually tea-like in the word “tea.” There’s nothing intrinsically porcine in the word “pig.” So it is, this theory tells us, that there are neurophysiological equivalents of linguistic functions—neural symbols, as it were—that are combined in representations of the world. They have no inherent intentional content, obviously, but they take on an intentional character through our acts of propositional or assertoric judgments—judgments such as “this is true” or “this is false,” that is. Call it “mentalese”: a “syntax” of brain-states that realizes neural symbols in a way that processes . . . well, input into output. Mental states, however—qualia, reasoning, beliefs, and so forth—come afterward, and belong to an emergent, second-order reality; the first-order series of realizations are brought about through the interactions of the semantic properties of this “language of thought” with one another and their environment, and all the actual transitions within this functional process are more or less computational in nature. Or, to put it very simply, none of these mental mechanisms in itself understands the representations it produces; but at the level of second-order processes the mind is a system that, as a whole, does understand. In fact, at that level, producing those understandings simply is the production of representations. That’s what it is to entertain mental “concepts.”
HERMES: So mental states—states we call consciously intentional—emerge from this “language of thought,” which itself emerges from an aggregation of neural “symbols” in interaction with one another and with the world, just as meaning emerges from the interactions of the otherwise inherently meaningless signs composing natural language?
HEPHAISTOS: That seems a fair summary of what many theorists claim.
HERMES: Well, then it also seems wrong.
HEPHAISTOS: Does it now?
HERMES: Don’t misunderstand me. I’m perfectly prepared to grant that thought and language are . . . isomorphic, let’s say. Indeed, I believe thought is language. But I don’t believe language is even vaguely similar to “neural symbols” in interaction with one another, principally because the very notion of “neural symbols” strikes me as just another vague metaphor—a transference of terms from a context in which they have a meaning into one in which they don’t. It might make a kind of sense, I suppose, if language actually arose from symbols that somehow existed in themselves prior to meaning and then, only latterly, combined into meanings through structural relations and interactions. But of course that’s not what happens. Language doesn’t arise from the parts of language; rather, language in its totality, as an intentional semeiotic system, is itself the source and ground of the parts that compose it.
HEPHAISTOS: Is that a structural or a genetic claim?
HERMES: Both. Structurally speaking, language isn’t a bottom-up product of a more primordial level of symbols without intentional content. Genetically speaking, language didn’t evolve out of an ensemble of discrete linguistic functions that somehow pre-existed the languages that give them meaning. Symbols never exist in isolation; they exist only within an entirely replete semeiotic ecology, a complete language, in which syntactic and semantic functions are wholly in operation, both of them sustained by a fully realized capacity for symbolic thought. To understand anything as indicating something else is always wholly dependent upon a complex system of signification. Language is structurally a top-down hierarchy, descending from intentional thought through a system composed of semantics and syntax—neither of which can exist in abstraction from the other—to particular instances of meaning.
HEPHAISTOS: Ah, of course. I almost forgot: the three of you think of language as a kind of formal causality, fashioning the material order “from above.”
PSYCHE: Language, life, and mind, to be precise. None of these things evolved. They’re eternal . . . well, forms.
HEPHAISTOS: Oh dear.
PSYCHE: Rather, material reality has evolved over time into an ever richer participation in their power. They’re the formal reality shaping material reality “from above” into ever more diverse and complex living, thinking, and communicating organisms.
HEPHAISTOS: Oh dear, oh dear.
HERMES: At least we’d agree that thought is language; we’d merely add that language doesn’t arise from any kind of mechanistic system, while thought is always already intentional, and intentionality can’t emerge from the non-intentional. So, what are we talking about here? Neurophysiological symbols? What could that possibly mean? How do neural processes refer to anything as that particular thing? Symbols exist as nothing but intentional content, whatever their physical accidents. That’s their whole ontology. Simply seeding one’s account of the brain’s operations with linguistic terms—symbol, semantic content, syntax—doesn’t render the claim that non-intentional processes are somehow essentially linguistic, or that language rests upon a non-intentional foundation, even remotely credible. I mean, look, when I say—and say sincerely—“That was a lovely tea,” my understanding of what I’m thinking can’t be the posterior effect of physical relations between differing neural functions distributed about my brain. My knowledge of my meaning—which is also my subjective awareness of my knowing my meaning—is a hermeneutical, interpretive act, a top-down formal determination that assumes the lower neural states into the higher mental states. It’s a semeiotic process: I’m using signs meaningfully.
HEPHAISTOS: And yet computers give us a very compelling model of the opposite process: of thought, that is, arising from a host of uncomprehending competences.
HERMES: They most certainly do not. That’s not what computation is. Quite the reverse: a prior semeiotic content on the part of a programmer or a user encodes or interprets the computer’s functions, but those functions—as far as the computer’s concerned—have no meaning at all. A computer neither uses symbols nor thinks.
HEPHAISTOS: Ah, well tell that to the chess-masters and masters of Go who’ve been defeated at their games by these supposedly unthinking mechanisms. Look, before we start freighting our account of things with language about intrinsic meaning, why don’t we proceed from the evidence before us? It’s hard to deny, after all, that computers seem to behave—and ever more so—as if they were thinking machines, which means there’s at least nothing inherently absurd in suggesting that thought might be, at base, mechanistic. Maybe the brain too is only a computer, and mind only a kind of software—just so many flows of data, processed into behaviors by functional connections and discriminatory filters, all erected upon a kind of digital platform in the neurology of the body. Maybe, as Dennett argues, all the brilliance and powers of comprehension that define our mental states and capacities simply emerge—or, rather, develop—out of countless little uncomprehending competences compounded over epochs of evolution into ever more competent—and so ever more comprehending—systems.1 If we were to undertake our aforementioned “homuncular decomposition” of the mind in purely computational terms, we might descend through a symbolic level of operations down to a level of something like simple binary functions, then down further to the even simpler “switches” in the brain that produce binary processes. Now, it’s a matter of indifference to me whether one concludes from this either that one day computers of sufficient complexity might become conscious or one concludes instead that thinking beings like gods and humans and Intermundian squirrels are in fact only computers who merely imagine they possess intrinsic consciousness. Those are really just two different ways of saying the same thing: that what we think of as intrinsic mental states are, in their origins, only functions. You’ve just described the process perfectly, in fact, albeit inadvertently: your seeming “mental states” really just translate certain brain-states into the behavior of saying, “That was a lovely tea.” Yes, yes, I know that doesn’t explain how consciousness exists or seems to exist, but it does explain what consciousness does. We’re accustomed to think we, say, experience a pain state and then, in consequence of that experience, perform this or that action. All that functionalism asks us to do is to entertain the possibility that the reverse is actually the case and that we “experience” a certain state as painful only as a result of its function in the larger system of translating outputs into inputs—the software of “thought.” We experience it as pain, that is, because such an “experience” is part of the process of producing, say, pulling one’s hand away from a fire.
