When the en­tire Sil­i­con Val­ley aims to build AI chips

Technowize Magazine - - Our Team -

You have caught up on the cur­rent af­fairs of the world. News filled with pol­i­tics, war, and crime is all that you read. Along with this, you feel that it is im­por­tant to stay up­dated in terms of tech­nol­ogy as it has be­come an in­te­gral part of our lives. In or­der to make sure that you can con­trib­ute to such tech savvy con­ver­sa­tions in a so­cial gath­er­ing, it’s time to read some­thing about tech­nol­ogy. In­stead of open­ing the crisp Sun­day pa­per, you un­lock your tab to scroll through the news app. This in it­self is a prac­ti­cal ex­am­ple of how im­por­tant tech­nol­ogy has be­come in our lives. You are go­ing through var­i­ous top­ics to read but a term keeps pop­ping up way of­ten than nor­mal. The term is Ar­ti­fi­cial In­tel­li­gence. The con­cept of ar­ti­fi­cial in­tel­li­gence is so pop­u­lar that it

is thrown around in al­most ev­ery con­ver­sa­tion re­lated to sci­ence and tech­nol­ogy. The term rings a bell as this is what Ap­ple and Google were talk­ing about while elab­o­rat­ing upon their new de­vices.

If we talk about ar­ti­fi­cial in­tel­li­gence, it is ba­si­cally a tech­no­log­i­cal con­cept which has now been de­vel­oped on an ex­ten­sive note. In the form of soft­ware, this tech­nol­ogy is in­stalled in var­i­ous kinds of de­vices. To make ar­ti­fi­cial in­tel­li­gence func­tion smoothly in a de­vice, an AI chip is manda­tory. An AI chip is an in­stru­ment which car­ries all the nec­es­sary data for the func­tion­ing of the AI soft­ware. With­out it, it is im­pos­si­ble to in­stall this soft­ware in any de­vice. This can give us an idea about the im­por­tance of an AI chip. As the de­mand for ar­ti­fi­cial in­tel­li­gence seems never-end­ing in this age, the pro­duc­tion of AI chips is also equally im­por­tant. A num­ber of tech gi­ants have un­der­stood this and are ded­i­cat­edly work­ing on the same. As ev­ery­one wants to try a hand at it, it is in­evitable that com­pe­ti­tion is also on the rise. In this race to pro­duce the most ef­fi­cient AI chips, tech com­pa­nies big and small are al­ready on the start line.


It is of­ten be­lieved that ar­ti­fi­cial in­tel­li­gence is a novel con­cept of this decade. This is a case of mis­con­cep­tion as the de­mand for AI in­creased in this decade but its in­cep­tion took place around more than two decades ago. A re­searcher at Bell Labs, Yann Le­cun built an AI chip in the year of 1992. At that time, the im­por­tance of this in­no­va­tion was not un­der­stood. One of the ma­jor rea­sons be­hind this is the lack of knowl­edge about the same. Be­ing an iconic R&D lab sit­u­ated right out­side New York City, Le­cun had a team of ef­fi­cient sci­en­tists to sup­port him in this vi­sion. This AI chip was de­signed in or­der to run deep neu­ral net­works. These net­works are com­plex math­e­mat­i­cal sys­tems. The AI helps these sys­tems learn tasks in­de­pen­dently with the help of data anal­y­sis. This neu­ral net­work worked well in cer­tain cases but had its lim­i­ta­tions. De­spite this, the in­no­va­tion sowed the seed of ar­ti­fi­cial in­tel­li­gence which turned out to be use­ful af­ter 25 years. base, var­i­ous tech com­pa­nies are de­vel­op­ing this AI chip even fur­ther. This de­vel­op­ment is done tak­ing into con­sid­er­a­tion the de­mands of the present day and how ar­ti­fi­cial in­tel­li­gence can en­hance the ex­pe­ri­ence their de­vice would of­fer. As he is said to be the fa­ther of AI chips, Le­cun cur­rently heads the ar­ti­fi­cial in­tel­li­gence lab at Face­book. Le­cun con­fi­dently be­lieves that such neu­ral chips are the need of the hour and need to be worked upon on an im­me­di­ate ba­sis.


One may have no­ticed that in the past few years, ar­ti­fi­cial in­tel­li­gence has be­come a grow­ing trend in the tech­no­log­i­cal sec­tor. Keep­ing this in mind, tech gi­ants are will­ingly in­vest­ing in this field to make sure they lead the mar­ket. As ar­ti­fi­cial in­tel­li­gence is soft­ware, peo­ple as­sume that it will grad­u­ally shut down any scope of hard­ware. What we fail to no­tice is no soft­ware can ex­ist with­out the base of hard­ware in or­der to carry that in­for­ma­tion. In this sce­nario, an AI chip is the nec­es­sary hard­ware. This is why all com­pa­nies of this field are hell bent on cre­at­ing a niche as the best AI chip man­u­fac­turer in the busi­ness. Right from bud­ding star­tups to well es­tab­lished gi­ants, each and ev­ery or­ga­ni­za­tion is work­ing upon it. Com­pa­nies like Google and Mi­crosoft who have earned them­selves a name in the tech busi­ness as pi­o­neers of in­no­va­tion with ex­per­tise are also forced to run in this race to re­tain their po­si­tion. This is the time when the need to build an AI chip for al­most ev­ery func­tion in a de­vice is high. Not only this, but an AI chip has the abil­ity to carry out cer­tain func­tions which would be pos­si­ble by a huge man­ual work­force. For ex­am­ple, Google re­cently pro­duced the TPU chip. It said that this chip could do the job which would usu­ally take 15 ex­tra data cen­ters. This means that trans­fer of data from the source to the de­vices can be much smoother as it will cut down on the long route of first send­ing the data to the data cen­ters and then trans­fer­ring it to the re­spec­tive de­vices. The de­lay of shut­tling data can be elim­i­nated thanks to these AI chips. Due to this, com­pa­nies are jump­ing at the op­por­tu­nity of cre­at­ing AI chips be­cause it not only cuts down on the over­head cost but also pro­vides a more ef­fi­cient so­lu­tion to tech ser­vices.


In­tel, the name which is con­sid­ered the fa­ther of com­puter soft­ware, ac­quired a startup called Nir­vana. This ac­qui­si­tion is in or­der to carry out the pro­duc­tion of a chip which will only fo­cus on ma­chine learn­ing. Sim­i­larly, IBM is work­ing on hard­ware ar­chi­tec­ture which re­sem­bles the the­ory of neu­ral net­work. Along the same lines, Qual­comm is busy de­vel­op­ing chips based on Le­cun’s ex­act prin­ci­ples of neu­ral work. Face­book is do­ing the same with the help of Le­cun him­self. The foun­da­tion that was built 25 years ago is now un­der­go­ing de­vel­op­ment and that too at a rapid pace. De­spite all these at­tempts, there is one or­ga­ni­za­tion that has the up­per hand when it comes to pro­duc­tion of AI chips. nvidia has earned it­self a rep­u­ta­tion in the Sil­i­con Val­ley as the leader of this field. Its ex­per­tise in graphic pro­ces­sors (GPU) can eas­ily crunch mas­sive amount of data with­out any flaws. This GPU chip is mainly in charge of ac­cel­er­at­ing the train­ing stage. Once the de­vice is well ac­quainted with the users’ needs and de­mands, all it has to do so fol­low in­struc­tions. GPUS are con­sid­ered to be the best op­tion for this task. GPUS do the job which usu­ally needs an en­tire ecosys­tem for proper func­tion­ing. Fol­low­ing the trend, Ap­ple Inc. may also soon pro­duce AI chips on an in­de­pen­dent note. It won’t be wrong to say that al­most all tech de­vices use ar­ti­fi­cial in­tel­li­gence. This func­tion­ing also takes place through AI chips. But till a few years back, not all these com­pa­nies pro­duced the hard­ware re­quired for this tech­nol­ogy. They usu­ally handed over the work to an­other or­ga­ni­za­tion and later in­stalled it in their de­vices. But look­ing at the need of the hour, com­pa­nies have de­cided to take the de­vel­op­ment of AI chips in their own hands. This has re­sulted in a num­ber of ac­qui­si­tions in the Sil­i­con Val­ley and will also pre­vent loop­holes in tech­nol­ogy that were ear­lier caused by data cen­ters.

Newspapers in English

Newspapers from USA

© PressReader. All rights reserved.