Ar­ti­fi­cial In­tel­li­gence Is Not A Tech­nol­ogy

Business a.m. - - TECHNOLOGY & INNOVATION - busi­ness a.m.

PEO­PLE HAVE LONG dreamed of the idea of ma­chines hav­ing the in­tel­li­gence and ca­pa­bil­i­ties of hu­mans. From the early Greek myths of Hephaes­tus and his au­toma­tons to the Golem of Eastern Euro­pean Jewish tra­di­tion to well over a hun­dred years of sci­ence fic­tion sto­ries, nov­els and movies, our hu­man imag­i­na­tions have en­vi­sioned what it would be like to have sen­tient, intelligent, hu­man-like ma­chines co-ex­ist with us. In 1920 Karel apek’s play R.U.R. (Ros­sum’s Uni­ver­sal Ro­bots) first coined the word “ro­bot” and gave us a name to give to the cre­ations of our imag­i­na­tions. In many ways, the quest for the intelligent ma­chine lead to the de­vel­op­ment of the mod­ern com­puter. Ideas by Alan Tur­ing not only for­mu­lated the ba­sis of pro­gram­mable ma­chines, but also the core of the con­cepts of ar­ti­fi­cial in­tel­li­gence, with the name­sake Tur­ing Test pro­vid­ing a means for eval­u­at­ing intelligent ma­chines.

Yet, with cen­turies of tech­nol­ogy ad­vance­ment and the al­most ex­po­nen­tial in­crease of com­put­ing re­sources, data, knowl­edge, and ca­pa­bil­i­ties, we still have not yet achieved the vi­sion of Ar­ti­fi­cial General In­tel­li­gence (AGI) -- ma­chines that can be an equal coun­ter­part of hu­man abil­ity. We’re not even close. We have de­vices we can talk to that don’t un­der­stand what we’re say­ing. We have cars that will hap­pily drive straight into a wall if that’s what your GPS in­structs it to do. Ma­chines are de­tect­ing images but not un­der­stand­ing what they are. And we have amaz­ing ma­chines that can beat world cham­pi­ons at chess and Go and mul­ti­player games, but can’t an­swer a ques­tion as ba­sic as “how long should I cook a 14 pound tur­key?” We’ve mas­tered com­put­ing. We’ve wran­gled big data. We’re fig­ur­ing out learn­ing. We have no idea how to achieve general in­tel­li­gence.

Part of the rea­son for this dis­con­nect is con­fus­ing the var­i­ous things that we’ve de­vel­oped as a re­sult of our quest for the intelligent ma­chine from the quest it­self. Ar­ti­fi­cial in­tel­li­gence is not a tech­nol­ogy.

Ask­ing the ques­tion whether or not some par­tic­u­lar tech­nol­ogy is or isn’t AI is miss­ing the point. Ar­ti­fi­cial in­tel­li­gence is the jour­ney. It’s the quest for the intelligent ma­chine. All the tech­nolo­gies we’ve de­vel­oped on the route to that quest are things that are in­di­vid­u­ally use­ful, but all to­gether, have not yet got­ten us to the goal. This is why it’s im­por­tant to un­der­stand that ar­ti­fi­cial in­tel­li­gence is not a tech­nol­ogy, in much the same way that the Space Race is not a tech­nol­ogy.

Mod­ern AI Roots

In or­der to un­der­stand where we cur­rently are or aren’t with AI (de­pend­ing on how you want to look at things) it’s im­por­tant to know the roots of mod­ern AI. The term ar­ti­fi­cial in­tel­li­gence was coined in 1956 at a Dart­mouth Univer­sity sum­mer con­fer­ence. John Mc­Carthy, cel­e­brated AI re­searcher who pulled to­gether the con­fer­ence him­self, said that AI doesn’t have a tech­nol­ogy-spe­cific mean­ing. In his words, “AI is the sci­ence and en­gi­neer­ing of mak­ing intelligent ma­chines.”

The cre­ator him­self saw AI as not the ends, but the means.

AI is not a tech­nol­ogy any more so than physics or civil en­gi­neer­ing. The chal­lenge is, like all sci­ences, there must be ap­pli­ca­tion to bring con­cept to re­al­ity.

Since the 1950s huge ad­vance­ments in tech­nol­ogy have evolved to help us on our quest to­wards the goals of ar­ti­fi­cial in­tel­li­gence. In the early decades of AI from the 1950s through the 1970s, we cre­ated ma­chines that could play games, un­der­stand ba­sic logic, could en­gage in sim­ple con­ver­sa­tions, had ba­sic ma­chine trans­la­tion ca­pa­bil­i­ties, and we even rudi­men­tary neu­ral net­works.

After a de­cline in in­ter­est pre­cip­i­tated by the un­der-de­liv­er­ing of AI over-prom­ises, in­ter­est re­sumed in the late 1980s with wide­spread adop­tion of com­puter desk­tops on knowl­edge worker’s desks and the de­vel­op­ment of ex­pert sys­tems. In the late 1990s IBM’s Deep Blue beat chess grand­mas­ter Garry Kas­parov. It seemed we were back on a roll with AI de­vel­op­ment. But it all came to a grind­ing halt when cor­po­rate and ven­ture in­ter­est waned

Yet, the dream was still alive and in­ter­est in AI re-emerged with a pas­sion in the mid 2000s. With the al­most lim­it­less power of cloud com­put­ing and GPUs, the preva­lence of big data and our know-how to use it well, and the de­vel­op­ment of ad­vanced neu­ral net­work al­go­rithms, we were able to achieve some of the prom­ises un­kept by AI in the past. AI is hip again. It’s in vogue. The money is flow­ing. Com­pa­nies are be­com­ing “AI first” as if the pre­vi­ous stigma never ex­isted. In­ter­est in AI con­tin­ues to heat up and shows no signs of slow­ing. And yet, with all the bil­lions in­vested and thou­sands of the world’s best thinkers on the case, we still have not cracked the nut of ar­ti­fi­cial general in­tel­li­gence. Is our tech­nol­ogy de­fi­cient? Is it a prob­lem with our data? Do we still not have enough com­put­ing power? Or is it more likely that the AI train is rolling along, and we’ve just left the lat­est sta­tion. In this wave, we’ve solved some prob­lems, but the goal of AI is still ahead of us. We need to pro­ceed to the next sta­tion. AI is Not A Tech­nol­ogy

Is AI the tech­nolo­gies peo­ple use to make ma­chines intelligent, or, is it the move­ment to­wards the goal of achiev­ing ma­chine in­tel­li­gence? Ac­cord­ing to John Mc­Carthy, AI is ac­tu­ally a sci­ence. It’s a field of study. But it might be more help­ful to think of AI as a goal. If AI is con­sid­ered to be a col­lec­tion of tech­nolo­gies, then you can ar­gue all day about what is and what isn’t AI. Are soft­ware ro­bots AI? Are self-driv­ing cars AI? Is com­puter vi­sion AI? Is char­ac­ter recog­ni­tion AI? If you think about it as tech­nol­ogy then it’s al­ways sub­ject to dis­agree­ment and in­ter­pre­ta­tion. How­ever, if you think about it as a goal, or a quest, then it’s some­thing we’re al­ways striv­ing to achieve, even if we aren’t quite there yet. Even if you think of AI as a field of study, like physics, those fields of study have goals. The goal of physics is to gain the true un­der­stand­ing of the na­ture of the uni­verse. Ev­ery­thing we’ve de­vel­oped in that quest for un­der­stand­ing in physics are tech­nolo­gies that are use­ful in our ev­ery­day lives. But those tech­nolo­gies aren’t physics -- they are the byprod­ucts of our quest to un­der­stand physics. In the same way, ma­chine learn­ing and com­puter vi­sion and ro­bot­ics aren’t AI, they are the tech­nol­ogy byprod­ucts of our quest to achieve or un­der­stand AI.

The quest for the ar­ti­fi­cially intelligent ma­chine has lead to great ad­vance­ments in the field. As out­lined above, we now have tech­nolo­gies that can rec­og­nize and clas­sify images, un­der­stand and gen­er­ate nat­u­ral language, self-driv­ing ve­hi­cle ca­pa­bil­i­ties, chat­bots that are able to con­verse si­mul­ta­ne­ously in mul­ti­ple lan­guages, sys­tems that can help di­ag­nose dis­eases such as cancer or di­a­betes, and an al­most in­nu­mer­able set of ap­pli­ca­tions across a wide range of in­dus­tries. Many cat­e­go­rize these ap­pli­ca­tions as “nar­row AI” (or heaven for­bid the pe­jo­ra­tive term “weak AI”). It goes with­out say­ing that since no one has yet achieved AGI (con­sid­ered to be “strong AI”), then ev­ery cur­rent ap­pli­ca­tion of AI is nar­row.

This makes the term nar­row AI si­mul­ta­ne­ously un­help­ful and use­less. When peo­ple use the term nar­row AI, they re­ally mean the cog­ni­tive tech­nolo­gies that have been de­vel­oped on the quest for the intelligent ma­chine.

So, in­stead of say­ing nar­row AI, just use the term “cog­ni­tive tech­nol­ogy.” That’s more in­dica­tive of what is meant. When some­one says general AI or strong AI, you should ask - are you re­fer­ring to the tech­nolo­gies that im­ple­ment the intelligent ma­chine or the ul­ti­mate goal it­self? One is a sci­ence, the other is the tech­nolo­gies of ap­pli­ca­tion.

with AI’s in­abil­ity to again de­liver on some of the prom­ises.

Newspapers in English

Newspapers from Nigeria

© PressReader. All rights reserved.