Noam Chomsky Explains Where Artificial Intelligence Went Wrong

While pop­u­lar­ly known for his pierc­ing and relent­less cri­tiques of U.S. for­eign pol­i­cy and eco­nom­ic neolib­er­al­ism, Noam Chom­sky made his career as a researcher and pro­fes­sor of lin­guis­tics and cog­ni­tive sci­ence. In his 50 years at MIT he earned the appel­la­tion “the father of mod­ern lin­guis­tics” and—after over­turn­ing B.F. Skinner’s behav­ior­ist paradigm—founder of the “cog­ni­tive rev­o­lu­tion.” But these are labels the self-effac­ing Chom­sky rejects, in his char­ac­ter­is­ti­cal­ly under­stat­ed way, as he rejects all tri­umphal­ist nar­ra­tives that seem to promise more than they deliv­er.

Such is the case with Arti­fi­cial Intel­li­gence. The term, coined in 1956 by com­put­er sci­en­tist John McCarthy, once described the opti­mism with which the sci­en­tif­ic com­mu­ni­ty pur­sued the secrets of human cog­ni­tion in order to map those fea­tures onto machines. Opti­mism has turned to puz­zle­ment, ambiva­lence, or in Chomsky’s case out­right skep­ti­cism about the mod­els and method­olo­gies embraced by the field of AI.

Nev­er par­tic­u­lar­ly san­guine about the prospects of unlock­ing the “black box” of human cog­ni­tion through so-called “asso­ci­a­tion­ist” the­o­ries, Chom­sky has recent­ly become even more crit­i­cal of the sta­tis­ti­cal mod­els that have come to dom­i­nate so many of the sci­ences, though he is not with­out his crit­ics. At an MIT sym­po­sium in May of last year, Chom­sky expressed his doubts of a method­ol­o­gy Nobel-win­ning biol­o­gist Syd­ney Bren­ner has called “low input, high through­put, no out­put sci­ence.”

Recent­ly Yarden Katz, an MIT grad­u­ate stu­dent in Cog­ni­tive Sci­ences, sat down with Chom­sky to dis­cuss the prob­lems with AI as Chom­sky sees them. Katz’s com­plete inter­view appeared this month in The Atlantic. He also video­taped the inter­view and post­ed clips to his Youtube chan­nel. In the clip above, Katz asks Chom­sky about “for­got­ten method­olo­gies in arti­fi­cial intel­li­gence.” Chom­sky dis­cuss­es the shift toward prac­ti­cal appli­ca­tion in engi­neer­ing and com­put­ing tech­nol­o­gy, which “direct­ed peo­ple away from the orig­i­nal ques­tions.” He also express­es the opin­ion that the orig­i­nal work was “way too opti­mistic” and assumed too much from the lit­tle data avail­able, and he describes how “throw­ing a sophis­ti­cat­ed machine” at the prob­lem leads to a “self-rein­forc­ing” def­i­n­i­tion of suc­cess that is at odds with sci­en­tif­ic dis­cov­ery.

In the clip below, Chom­sky dis­cuss­es a new field in sys­tems biol­o­gy called “Con­nec­tomics,” an attempt to map the wiring of all the neu­rons in the brain—an endeav­or prick­ly biol­o­gist Syd­ney Bren­ner calls “a form of insan­i­ty.” Katz asks if the “wiring dia­gram” of the brain would pro­vide “the right lev­el of abstrac­tion” for under­stand­ing its work­ings.

The inter­view is worth read­ing, or watch­ing, in full, espe­cial­ly for stu­dents of neu­ro­science or psy­chol­o­gy. Chom­sky dis­cuss­es the work of his one­time col­league David Marr, whose posthu­mous­ly pub­lished book Vision has had an enor­mous influ­ence on the field of cog­ni­tive sci­ence. Chom­sky also prais­es the work of Randy Gal­lis­tel, who argues that devel­op­ments in cog­ni­tive and infor­ma­tion sci­ence will trans­form the field of neu­ro­science and over­turn the par­a­digms embraced by ear­ly researchers in AI. While this is an excit­ing time to be a cog­ni­tive sci­en­tist, it seems, per­haps, a dif­fi­cult time to be a pro­po­nent of Arti­fi­cial Intel­li­gence, giv­en the com­plex­i­ties and chal­lenges the field has yet to meet suc­cess­ful­ly.

Relat­ed Con­tent:

Noam Chom­sky Spells Out the Pur­pose of Edu­ca­tion

Noam Chom­sky & Michel Fou­cault Debate Human Nature & Pow­er (1971)

Josh Jones is a doc­tor­al can­di­date in Eng­lish at Ford­ham Uni­ver­si­ty and a co-founder and for­mer man­ag­ing edi­tor of Guer­ni­ca / A Mag­a­zine of Arts and Pol­i­tics.