Why UAE in­vestors should start look­ing to AI ALEXAN­DRE MOUTHON

Alexan­dre Mouthon ex­plores the grow­ing in­flu­ence of AI

Gulf Business - - CONTENTS -

Ar­ti­fi­cial in­tel­li­gence is one of the fastest­grow­ing ar­eas of dig­i­tal tech­nol­ogy – set to trans­form our lives, and our in­vest­ment port­fo­lios.

The lat­est de­vices from the likes of Ama­zon, Mi­crosoft and Google al­ready en­able us to switch on the lights, check the weather fore­cast, con­sult our di­aries, lis­ten to mu­sic, boil the ket­tle or book a taxi – all by voice com­mand.

Soon, AI will be able to achieve much more, sup­ported by in­vest­ment from dig­i­tal com­pa­nies who now see ma­chine-learn­ing sys­tems as a ma­jor source of fu­ture growth. The UAE has the po­ten­tial to lead this growth in the Mid­dle East due to bold new leg­is­la­tion, a net­work of tal­ent and the grow­ing num­bers of start-ups that are creat­ing unique data sets to solve prob­lems that are ripe for ma­chine learn­ing.

The AI mar­ket is fore­cast to bal­loon to $127bn by 2025 from $2bn in 2015, with the US and China lead­ing the way so far and the bulk of the in­vest­ment com­ing from high tech, tele­com and fi­nan­cial ser­vices sec­tors within those coun­tries.

For the global econ­omy that could mean tril­lions of dol­lars of ex­tra growth, through both in­creased pro­duc­tiv­ity and via a boost to con­sump­tion as peo­ple spend money on new or im­proved goods and ser­vices. AI is a per­va­sive en­abling tech­nol­ogy that will boost the for­tunes of com­pa­nies across mul­ti­ple in­dus­tries, but cause the demise of oth­ers.

The dif­fer­ence be­tween com­mer­cial suc­cess and fail­ure could be how ef­fec­tively and how quickly firms de­ploy AI and how ef­fec­tively they are able to an­a­lyse a lot of data. The an­nual amount of data we pro­duce is fore­cast to reach 163 tril­lion gi­ga­bytes in 2025 – 10 times more than in 2016.

Through ma­chine learn­ing, al­go­rithms can use his­tor­i­cal data to spot pat­terns and pre­dict what might hap­pen in the fu­ture. Next comes deep learn­ing, where com­put­ers learn from their mis­takes and re­fine pre­dic­tions with each new frag­ment of in­for­ma­tion, be­com­ing in­creas­ingly adept at recog­nis­ing images and speech, as well as at nat­u­ral lan­guage pro­cess­ing (NLP) – un­der­stand­ing spo­ken and writ­ten words just as hu­mans do and re­spond­ing within con­text.

There are two di­rect ways for in­vestors to ac­cess AI in­vest­ments. One is through hard­ware – the phys­i­cal de­vices used in ma­chines. The other route is through soft­ware – the col­lec­tion of code in­stalled into a com­puter's hard drive.

On the hard­ware side, de­mand for pow­er­ful graph­ics pro­ces­sors (GPUs), which can en­able com­put­ers to an­a­lyse and make use of vast quan­ti­ties of data, is fu­elled by the pro­cess­ing power re­quired for deep learn­ing.

Driver­less cars, for ex­am­ple, have a lot of learn­ing to do: both on their im­me­di­ate en­vi­ron­ment and how to re­act to dif­fer­ent si­t­u­a­tions. To bet­ter ac­com­mo­date this, Tesla re­cently used new GPUs to up­grade the pro­cess­ing power of its au­topi­lot sys­tem by 40 times. Any fur­ther tech de­vel­op­ments are likely to be met with strong ap­petite from man­u­fac­tur­ers.

Given that most GPU chip pro­duc­tion is out­sourced (with GPU com­pa­nies mainly fo­cus­ing on the de­sign rather than man­u­fac­tur­ing), this should lead to bet­ter growth prospects in the foundry and out­sourced semiconductor assem­bly and test (OSAT) in­dus­tries.

The more so­phis­ti­cated AI ma­chines get, the more mem­ory they will likely re­quire to func­tion. Just one sec­ond of au­ton­o­mous driv­ing can gen­er­ate as much as 1GB of data, for ex­am­ple. All that in­for­ma­tion needs to be stored some­how, some­where; creat­ing de­mand for mem­ory chips, as well as for cloud stor­age so­lu­tions.

AI could ac­count for around 25 per cent of to­tal semiconductor de­mand by 2020, from 10-15 per cent today.

The AI-driven trans­for­ma­tion in soft­ware is ar­guably pro­gress­ing even faster. Dig­i­tal soft­ware com­pa­nies will ben­e­fit the most from the AI era as they can ex­tract re­cur­ring rev­enue from sub­scrip­tions as well as new prod­ucts. This is in con­trast to semiconductor com­pa­nies that will mostly profit from a one-time pur­chase cy­cle.

Com­pa­nies such as Face­book, Baidu, Sales­force.com or Me­di­data have ac­cess to mas­sive amounts of con­sumer or en­ter­prise data in their re­spec­tive fields and are able to of­fer value-added ser­vices based on AI to their cus­tomers.

The scope for what data-crunch­ing AI soft­ware can do is vir­tu­ally lim­it­less, and com­pa­nies are feel­ing the ben­e­fits al­ready. Net­flix, for ex­am­ple, es­ti­mates that it is pre­vent­ing over $1bn of po­ten­tial rev­enue loss per year from can­celled sub­scrip­tions by of­fer­ing tai­lored search re­sults and rec­om­men­da­tions. Ama­zon man­aged to re­duce its ware­house oper­at­ing costs by at least a fifth by em­ploy­ing au­ton­o­mous ro­bots. Ul­ti­mately, the key to cor­po­rate suc­cess in the dig­i­tal world will be the abil­ity to har­ness data and turn it into busi­ness op­por­tu­ni­ties. For in­vestors, one of the most di­rect ways to tap into AI is by iden­ti­fy­ing the tech spe­cial­ists who are de­liv­er­ing the best hard­ware and soft­ware. Only by em­brac­ing the ro­botic rev­o­lu­tion can we stay ahead of the game.

AI IS A PER­VA­SIVE EN­ABLING TECH­NOL­OGY THAT WILL BOOST THE FOR­TUNES OF COM­PA­NIES ACROSS MUL­TI­PLE IN­DUS­TRIES, BUT CAUSE THE DEMISE OF OTH­ERS.

Alexan­dre Mouthon is se­nior prod­uct spe­cial­ist at Pictet As­set Man­age­ment

Newspapers in English

Newspapers from UAE

© PressReader. All rights reserved.