The Im­mutable Con­nec­tion be­tween Moore’s Law and Ar­ti­fi­cial In­tel­li­gence

Technowize Magazine - - Front Page -

In­tel co-founder Gor­don Moore noted a pe­cu­liar trend in the semi-con­duc­tors and elec­tron­ics in­dus­try; an out­line for the whole the semi­con­duc­tor in­dus­try to ad­vance. The four and a half page ar­ti­cle by Gor­don Moore in the Elec­tron­ics Trade jour­nal was the most im­pact­ful ar­ti­cle ever that led the ba­sic foun­da­tion for the sil­i­con based cir­cuitry in­dus­try.

MOORE’S law, THE EVO­LU­TION AND THE AD­VANCE­MENT

Moore was a Cal Tech PHD, co­founder of Fairchild Semi­con­duc­tor and head of its re­search and devel­op­ment lab­o­ra­tory. Fairchild has been founded to de­velop tran­sis­tors from sil­i­con rather than ger­ma­nium.

From 1958, since the in­ven­tion of the first in­te­grated cir­cuit till 1965, the num­ber of com­po­nents or tran­sis­tor den­sity in an in­te­grated cir­cuit has dou­bled

every year, marked Gor­don Moore.

The state­ment which was ini­tially an ob­ser­va­tion was la­belled as Moore’s law later in 1970 by Cal­tech pro­fes­sor, VLSI pi­o­neer and en­tre­pre­neur Carver Mead. The phrase then was caught by the elec­tron­ics In­dus­try as a core prin­ci­ple.

To in­crease the ac­cu­racy of the law, spe­cific for­mu­la­tion of law has been de­vel­oped by Moore after­wards with more pre­cise ob­ser­va­tion. In 1975, Moore re­vised his pre­dic­tion as the num­ber of com­po­nents in the in­te­grated cir­cuits dou­bling every year to dou­bling every two years. The lat­est pre­dic­tion of Moore was fac­tored by In­tel ex­ec­u­tive David house stat­ing that com­puter per­for­mance dou­bles every 18 months fol­low­ing the Moore’s lat­est pre­dic­tion. So when In­tel, the pi­o­neer of chip de­vel­op­ments adapted Moore’s law as stan­dard prin­ci­ple for ad­vanc­ing the com­put­ing power, the whole semi-con­duc­tor in­dus­try fol­lowed this out­line on their chips.

Moore’s law is ba­si­cally about tran­sis­tor den­sity, but there are many ver­sions of the Moore’s law for other ca­pa­bil­i­ties of dig­i­tal elec­tron­ics. Pro­cess­ing speed, RAM’S, sen­sor strength, pixel den­sity or res­o­lu­tion, all have some di­rect re­la­tion with the num­ber of com­po­nents that fits into the in­te­grated cir­cuit. Since their in­cep­tion, en­hance­ments in all these ca­pa­bil­i­ties have also in­creased ex­po­nen­tially.

The chip mak­ing com­pa­nies used to es­ti­mate that in­creas­ing the com­po­nents den­sity will lead to an in­crease in com­put­ing power by x times. The com­put­ing power re­quired for the en­hance­ments till now were the con­se­quence of adap­tion of Moore’s law. The ef­fects of Moore’s law have far fetch­ing im­pli­ca­tions on tech­no­log­i­cal devel­op­ment. Since last five decades, the ad­vance­ment from slow run­ning, huge-sized com­put­ers to the de­vel­op­ments of ar­ti­fi­cial as­sis­tants and AI pow­ered de­vices, Moore’s holds the credit.

MOORE’S law TO halt

Since last five decades, Moore’s law has been a re­cur­ring set of dead­lines that the chip

de­vel­op­ers ex­celled to meet to dou­ble their com­put­ing power. But the soar­ing de­mand for smaller, faster and more ef­fi­cient com­put­ers bound a limit line for Moore’s law to ex­tend any­more.

we have seen an ex­po­nen­tial in­crease of tran­sis­tor den­sity for last half cen­tury fol­low­ing Moore’s law. But now, pack­ing the tran­sis­tors in the sil­i­con wafer is no more an op­ti­mised so­lu­tion to power the com­puter pro­cess­ing. A re­port from In­ter­na­tional Tech­nol­ogy of roadmap for semiconductors (ITRS) claimed that by 2021, tran­sis­tors will get to a point where they can­not shrink any fur­ther. The big-name chip mak­ers like In­tel and Sam­sung ar­gue that mak­ing chips smaller is not eco­nom­i­cally vi­able now.

Moore’s pre­dic­tion was kept alive in the in­dus­try with In­tel lead­ing the change. But ac­cord­ing to In­tel, the tech­nol­ogy roadmap for Moore’s law will scrap soon. In­tel since 1970 has fol­lowed Moore’s ex­po­nen­tial curve and fit twice as many the tran­sis­tors in the same space on a sil­i­con chip. The con­tin­ual shrink­ing has helped com­put­ers get more pow­er­ful and en­ergy ef­fi­cient. But now ac­com­mo­dat­ing the tran­sis­tors in the same space has be­come a short­hand idea, eco­nom­i­cally and prac­ti­cally. with the de­creas­ing dis­tance be­tween the tran­sis­tors, the heat re­lease will in­crease. So mak­ing tran­sis­tors fur­ther any smaller or de­creas­ing the dis­tance would not bring in­crease in com­put­ing ca­pa­bil­ity. we have to­tally got down the piles of Moore’s law to form a grain.

how­ever, Moore’s law is in­di­vid­ual ex­po­nen­tial but the com­put­ing tech­nol­ogy to­day de­mands par­al­lel ex­po­nen­tial. To run deep learn­ing al­go­rithms and de­velop AI ar­chi­tec­ture we need to lever­age the com­put­ing po­ten­tial. GPU’S and re­pro­grammable chips are a step to­wards de­liv­er­ing par­al­lel com­put­ing power.

MOORE’S law will Al­ways be AN IM­POR­TANT base

Moore’s law has helped us bring smart­phones, high-speed in­ter­net ser­vices and ma­jor break­throughs in the field such as ar­ti­fi­cial in­tel­li­gence and ge­net­ics.

In 1965, Moore pre­dicted,

“In­te­grated cir­cuits will lead us to such won­ders as home com­put­ers-or at least ter­mi­nals con­nected to a cen­tral com­puter-au­to­matic con­trols for au­to­mo­biles, and per­sonal por­ta­ble com­mu­ni­ca­tions equip­ment.”

Moore had a pioneering vi­sion and all what he pre­dicted is prac­ti­cally real to­day.

But ex­perts now say than the ten­ure of Moore’s law as the base for ad­vanc­ing the com­put­ing power is about to end. The rate of AI progress

will not in­crease with Moore’s pre­dic­tion. Is it a mere moot now for the devel­op­ment of ro­bust pro­cess­ing chips com­pat­i­ble to put AI in im­ple­men­ta­tion?

Moore’s law is no more a prac­ti­cal so­lu­tion to cre­ate su­per­in­tel­li­gent tech­nol­ogy is a strong point. Even, Moore ini­tially in his ar­ti­cle pre­dicted the prac­ti­cal ap­pli­ca­tion of Moore’s law to be for next ten years from 1965. But then with the con­stant ad­vance­ment, the elec­tron­ics in­dus­try ben­e­fited from the Moore’s stan­dard method of de­sign­ing pro­ces­sor chips till 50 years.

The tech­nol­ogy to­day is tend­ing

to de­sign ar­ti­fi­cial in­tel­li­gence tech­nol­ogy that matches the su­per in­tel­li­gence of hu­man brain. we agree to the fact that Moore’s law can­not cre­ate such pow­er­ful AI. But we can­not point Moore’s law as the rea­son for not hav­ing the com­put­ing ca­pa­bil­ity and hard­ware for de­vel­op­ing self-evolv­ing, nat­u­ral in­tel­li­gence tech­nol­ogy. Moore’s law still gives an im­por­tant ca­dence for the devel­op­ment of AI. The ba­sic AI that we have to­day has de­vel­oped with the power ob­tained from the cur­rent pro­ces­sors; de­vel­oped us­ing Moore’s prin­ci­ple. And that sets the base of the idea of su­per in­tel­li­gence tech­nol­ogy.

The rate of AI progress will in­crease when the com­puter pro­cess­ing speeds will match to that of the hu­man brain. The best brain sim­u­la­tions by the best su­per­com­put­ing clus­ters could hardly fetch 1 per­cent of the brain and 1/10,000th of the cog­ni­tive hu­man brain speed. So all we are lack­ing is that su­per com­put­ing power which is must for the self-evolv­ing tech­nol­ogy.

The naysay­ers of Moore’s law need to un­der­stand that AI will not spon­ta­neously erupt from nowhere with the end of Moore’s law. But we need su­per­com­put­ing pro­cess­ing power that will make the AI pos­si­ble; as we got Moore’s law years back which made pos­si­ble com­put­ing power we have to­day.

we need to fig­ure out how un­su­per­vised the learn­ing would work, how the com­put­ers would de­tect what hu­man brains could or how we could in­cul­cate the hu­man sens­ing abil­ity into the ma­chine; all that with small, ef­fi­cient and op­ti­mised pro­ces­sors.

The founder of Spacex and Tesla, Elon Musk is com­ing up with Neu­ralink which will merge hu­man brains (neu­rons) with ma­chines to de­velop self­e­volv­ing tech­nol­ogy. The startup will find ways to fig­ure out how learn­ing with ma­chines work.

we need to copy the brain and make evo­lu­tion that in­cul­cates ba­sic hu­man think­ing in the ma­chines. Once this tech­no­log­i­cal break­through oc­curs, the world will be a tech­no­log­i­cal won­der­land as we see in the movies and the earth will have an om­nipo­tent God.

Newspapers in English

Newspapers from USA

© PressReader. All rights reserved.