This excerpt from Psychosemantics. Jerry A. Fodor. © 1989 The MIT Press. is provided in screen-viewable form for personal use only by members of MIT CogNet. Unauthorized use or dissemination of this information is expressly forbidden. If you have any questions about this material, please contact [email protected].

Epilogue

Creation Myth

In principle , functionalists assure us, you can make intelligence out of almost anything . In practice, however , the situation is a little more complicated . For example: A material substrate capable of sustaining so much as a rudimentary intelligence is likely to be quite remarkably complex in its physical constitution . Given the very nature of matter , such complex systems are invariably unstable . They come apart in what seems, when viewed under the aspect of Eternity , to be a ridiculously short time . This means that the totality of experit:~nce that any biologically feasible embodied mind has time to acquire j.s really quite trivial . And yet, acquired experience is surely a key to the higher forms of mentation ; if experience without intelligence is blind , intelligence without experience is empty . Bernard Shaw suggested that the best \Jvay to get a lot smarter would be to live a lot longer ; he thought that one might do this simply by deciding to . But as a matter of fact- more precisely, as a fact of matter- one hasn't got the option . Embodiment implies mortality , and mortality constrains the amount of information that a mind can come to have. How , then , is embodied intelligence possible? We might call this the lEI problem .' The birds of the air and the beasts of the field never solved it ; considered as experiments in EI, they are .:1 dead end . The trouble with these lower forms of creation is that th ~e cognitive achievements of their species don 't accumulate. Because learned adaptations are not communicated , each individual is forcedl to recapitulate the intel lectual history of its kind . Correspondingly , what is required for the higher manifestations of intelligence is the capacity td pool experience. Shakespeare wouldn 't have gotten around to writing Hamlet if he had had first to rediscover fire and reinvent the wheel . What ' s wanted is that each ne~ generation should somehow inherit the accumulated wisdom of its predecessors. Ideally , each individual should be the beneficiary of the learned adaptation ~) of all of its conspecifics, cohorts, and predecessors alike .

130 Epilogue What makes the EI problem especially interesting from an engineering point of view is that the obvious- Lamarckian - solution proves to be unfeasible . A natural way to pool the experience of conspecifics would be to allow it to affect the transmission of traits from one generation to the next. But this turns out not to be biochemically possible (and, on balance, not to be ethologically desirable either : substantial insulation of genetic material is required to main tain the genotypic stability of a species). So if , in short, the engineering problem is to design a kind of embodied intelligence that will , in the fullness of time , write Hamlet, then the practical bottleneck is to devise a non -Lamarckian mechanism for the transmission of acquired traits ; one that is capable of preserving and communicating cognitive achievements, to put the matter in a nutshell . Homo sapiensimplements the best solution to this problem that has thus far been proposed . Given language (and other cultural artifacts), each generation can record and transmit what it has learned; and each succeeding generation can assimilate the information that is so recorded and transmitted . It turns out that this process can rapidly and cumulatively modify the characteristic behavioral phenotype of a species without resort to Lamarckian modifications of its genotype . Moreover , the same linguistic vehicle tha t permi ts Homo sapiensto transmit acquired information between generations also serves to effect its diffusion among contemporaries . Specialization of labor- and with it the development of expertise- follows as the night the day. The implications of all this for EI have turned out to be positively startling . Early skepticism to the contrary notwithstanding , some biologically embodied intelligences are now able to play very decent games of chess (quite a few of them can beat computers). And , against all odds, a biologically embodied intelligence actually has written Hamlet. (Philosophers at Berkeley deny, however , that these initial successeswill be sustained . The point out , for example, that no biologically embodied intelligence has yet succeeded in writing Matilda, a play that is said to be much better than Hamlet. Why , they ask, should we believe that a biologically embodied intelligence ever will ? I stand neutral on such questions. Further research is required ; speculation is premature .) As just remarked , if Homo sapiensachieves moderately respectable manifestations of EI, that is largely because of the availability of lin guistic and cultural mechanisms for the pooling of learned adaptations . This is, of course, an ethological commonplace and will surprise no one. But it has some implications that are less widely advertised and well worth considering . EI presupposes a culture ; and a culture presupposes a social

Creation Myth

131

animal, one that is capable of integrating its own behavior with the behaviors of others of its kind . But how are such integrations to be achieved? This problem has the form of a dilemma . On the one hand : The more intelligent the animal , the more it needs to be a social creature; and the more social the animal, the more it needs to coordi nate its behaviors with those of conspecifics. But also: The more intel ligent the animal , the more intricate its behavioral capacities; and the more intricate the behavioral capacities, the harder the problem of coordinating conspecific behaviors is going to be. You do not want to have to spend your life reinventing the wheel . But neither do you want to have to spend it learning enough about your neighbor ' s psychology to permit you to exploit his expertise (more generally, to permit a division of labor to be worked out between you ). In short , socialization will not solve the EI problem unless a feasible solution to the coordination problem is also somehow achieved. We are not , of course, the only species for which the coordination problem arises; every animal that moves- certainly every animal that reproduces sexually- is more .or less social, so mechanisms to effect behavioral coordinations have often been evolved before. But they are typically - to borrow the inelegant AI expression- kludges; their very feasibility presupposes the stupidity of the organism that employs them . Male sticklebacks, for example, establish territories which they defend against the encroachments of other males. The result is a solution of the ecological problem of segregating breeding pairs . In order for this solution to work , however , each male must be able to determine which of the objects in its environment is a potential rival . This is a bare minimum sort of coordination problem , and it is solved in a bare minimum sort of way . Sexually active males (but not females) develop characteristic red markings ; and, reciprocally , a red marking in the perceptual environment of a sexually active male prompts a display 'of its territorial behavior . Since this coupling of a fixed releasing stimulus with a correspondingly fixed released response is innate , male sticklebacks don 't have to waste time learning how to detect rivals or what to do about the rivals that they detect. They can instead proceed directly to the important business of generating more sticklebacks. The point to notice, however , is that this solution works only because the ecology of sticklebacks is so impoverished . Male sticklebacks don 't get around much ; they breed and have their being in a world in which de facto the only things that display red patches are rival sticklebacks. The stupidity of the whole arrangement is im mediately manifest when an experimenter introduces an arbitrary red

132 Epilogue object into the scene. It turns out that practically anything red elicits the territorial display ; a breeding stickleback male will take Santa Claus for a rival . Kludges work as solutions to coordination problems, but only w hen the behavioral repertoires to be coordinated are exiguous and stereotyped , and only when the environments in which the behaviors are exhibited are relatively static. Whereas: If social coordination is to lead to higher forms of embodied intelligence , it must be achieved across very rich behavioral repertoires , and across environments that keep changing as a consequence of the behavior of the very organisms that inhabit them . Moreover , the conditions for accurate coordination must be achieved rapidly compared to the length of an individual life . There is, as previously suggested, no use designing a social organism with a long prematurity if most of its apprentice years have to be spent learning the commonsense psychology of its species. Here is what I would have done if I had been faced with this problem in designing Homo sapiens. I would have made a knowledge of commonsense Homo sapienspsychology innate; that way nobody would have to spend time learning i,t . And I would have made this innately apprehended commonsense psychology (at least approxi mately ) true, so that the behavioral coordinations that it mediates would not depend on rigidly constraining the human behavioral repertoire or on accidental stabilities in the human ecology. Perhaps not very much would have to be innate and true to do the job; given the rudiments of commonsense Intentional Realism as a head start, you could maybe bootstrap the rest. The empirical evidence that God did it the way I would have isn't, in fact, unimpressive (though , for the present speculative purposes, I don't propose to harp on it ). Suffice it that : (1) Acceptance of some form of intentional explanation appears to be a cultural universal . There is, so far as I know , no human group that doesn't explain behavior by imputing beliefs and desires to the behavior . (And if an anthropologist claimed to have found such a group , I wouldn 't believe him .) (2) At least in our culture , much of the apparatus of mentalistic explanation is apparently operative quite early . Developmental psychologists now admit that at least J'a rudimentary awareness of the existence of the mental world is present in toddlers and preschoolers [viz ., by age 2.5]" (Wellman , CTM, 176), and the trend - in these increasingly nativistic times- is clearly toward revising downwards the estimated age of the child 's earliest displays of this sort of cognitive sophistication . The recent history of developmental psychologists ' second thoughts on this matter is quite strikingly similar to what ' s been happening in developmental

Creation Myth

133

psycholinguistics : as our experimental techniques get better , infants seem to get smarter . The more we learn about how to ask the ques tions , the more of the answers infants seem to know . (3) I take the lack of a rival hypothesis to be a kind of empirical evidence ; and there are , thus far , precisely no suggestions about how a child might acquire the apparatus of intentional explanation 'from experience .' (Un less by 'introspection ' ?!) Wellman (CTM ) remarks that " language acquisition must be an enormous source of information : there are mental verbs to be learned such as remember/ believe/ know / expect/ and guess." How , precisely / a child who had no idea of what remembering is would go about learning that 'remember ' is the verb that means remember, we are not , however , informed . The advantage of making a theory innate is that what is innate does not have to be acquired . The advantage of making an innate theory true is that , quite generally / true theories license more reliable predic tions than false theories do . God gave the male stickleback the idea that whatever is red is a rival . Because this idea is false, the stick leback ' s innate psychological theory mediates only stereotyped behavioral coordinations , and those only while adventitious ecological regularities obtain . God gave us such - rather more complicated ideas as that if x wants that P / and x believes that not -P unless Q/ and x believes that x can bring it about that Q/ then ceteris paribus x tries to bring it about that Q . Because this idea is true , our innate psycho logical theory mediates vastly more flexible behavioral coordinations than the stickleback ' s, and will continue to do so as long as human nature doesn 't change . That is one reason why we wrote Hamlet and the stickleback didn ' t . Homo sapiens is , no doubt , uniquely the talking animal . But it is also , I suspect , uniquely the species that is born knowing its own mind .

This excerpt from Psychosemantics. Jerry A. Fodor. © 1989 The MIT Press. is provided in screen-viewable form for personal use only by members of MIT CogNet. Unauthorized use or dissemination of this information is expressly forbidden. If you have any questions about this material, please contact [email protected].