Ar­ti­fi­cial In­tel­li­gence: A Force Dis­rupt­ing Many Sec­tors

Arabnet - The Quarterly - - Industry Stories - By Nadine Ka­haleh | @91Ka­haleh

Sev­eral fac­tors have im­pacted the pro­lif­er­a­tion at which the re­search and de­vel­op­ment of AI is pro­gress­ing. The in­crease in com­pu­ta­tional re­sources, the ex­plo­sive growth of data which stands at 48% year-on-year, based on Mary Meeker’s 2017 In­ter­net Trends Re­port, along with the de­creas­ing cost of data stor­age, and the surge of open source frame­works, in ad­di­tion to the shift from broad AI to in­dus­try-fo­cused AI, have en­abled Ma­chine Learn­ing (ML) and Deep Learn­ing to ac­cel­er­ate the evolve­ment of AI.

To­day, dif­fer­ent stake­hold­ers are bet­ting on the ca­pac­i­ties of AI in trans­form­ing the world as we know it. In fact, Global Fore­cast 2022 re­port ex­pected the AI mar­ket to be worth $16 BN by 2022, which will ev­i­dently in­crease the in­vest­ments in the tech­nol­ogy.

AI is al­ready on the track of em­pow­er­ing dif­fer­ent sec­tors. In this ar­ti­cle we will be un­der­lin­ing the main sec­tors where AI is al­ready hav­ing or is pro­jected to have the great­est im­pact: au­to­mo­tive, health­care, ed­u­ca­tion, and bank­ing. WHAT DOES AI STAND FOR? Known as the fa­ther of Ar­ti­fi­cial In­tel­li­gence, English math­e­ma­ti­cian John Mccarthy, coined the term "Ar­ti­fi­cial In­tel­li­gence" in 1956 when he held the first aca­demic con­fer­ence on the sub­ject. He pre­sented his def­i­ni­tion of the word at a con­fer­ence on the cam­pus of Dart­mouth as such "the science and en­gi­neer­ing of mak-

ing in­tel­li­gent ma­chines".

A num­ber of dic­tio­nar­ies de­fine AI to be an area of com­puter science that em­pha­sizes the cre­ation of in­tel­li­gent ma­chines that work and re­act like hu­mans. AI can ra­tio­nal­ize and take ac­tions that have the high­est prob­a­bil­ity of achiev­ing a spe­cific goal. The tech­nol­ogy bases it­self on the idea that hu­man in­tel­li­gence can be mim­icked by a ma­chine that is wired us­ing ap­proaches from math­e­mat­ics, com­puter science, lin­guis­tics, psy­chol­ogy, and more. It is im­por­tant to men­tion that AI is not merely one tech­nol­ogy; it is a group of cor­re­lated tech­nolo­gies in­clud­ing (1) nat­u­ral lan­guage pro­cess­ing, that en­sures a nor­mal in­ter­ac­tion be­tween com­put­ers and hu­mans, (2) ML, that al­lows com­put­ers to evolve once ex­posed to data, and (3) ex­pert sys­tems that are soft­wares pro­grammed to pro­vide ad­vice.

ML al­go­rithms are based on ob­ject track ing and so­phis­ti­cated pat­tern recog­ni­tion .

“Com­puter vi­sion” con­stantly an­a­lyzes the en­vi­ron­ment and feeds per­ceived images into the al­go­rithms. The images are then an­a­lyzed and the na­ture of the ob­jects is clas­si­fied through AI. Th­ese al­go­rithms give the ve­hi­cle ‘ in­tel­li­gence’, al­low­ing the ve­hi­cle to learn ob­ject char­ac­ter­is­tics such as move­ment, size and shape in or­der to clas­sify fu­ture images with higher ac­cu­racy.

Con­nected Cars

AI en­ables cars to com­mu­ni­cate with one another and with the road in­fra­struc­ture. By han­dling back-end com­pu­ta­tions, AI will de­liver ac­cu­rate and timely data, whereas ML al­go­rithm will be track­ing and reg­is­ter­ing data re­lated to the ve­hi­cle’s speed, lo­ca­tion, and des­ti­na­tion. AI will learn the driver’s daily sched­ule, the roads they usu­ally take, and their ha­bit­ual stops to pro­vide the driver with in­sights be­fore their com­mute.


Th­ese sys­tems in­clude the most inno-va­tive in-ve­hi­cle fea­tures like speech recog­ni­tion and vir­tual as­sis­tants.

Speech Recog­ni­tion

Speech recog­ni­tion pro­vides an eas­ier way for hu­mans to in­ter­act with tech­nol­ogy and in this case, pro­vid­ing the in­ter­ac­tion be­tween driv­ers and their cars. How does it work? AI in­ter­prets voices as sound waves that are con­verted into code which the al­go­rithms then an­a­lyze. Fur­ther to that, the speech is com­pared to other sam­ples stored in the cloud to de­ter­mine what the user is say­ing. The speech recog­ni­tion soft­ware will im­me­di­ately start to up­date speech sam­ples the more the driver uses it, tak­ing into ac­count how spe­cific words are pro­nounced and the tone of the driver’s voice. The tech­nol­ogy’s ca­pa­bil­ity of learn­ing a dis­tinct ac­cent and pronunciation of words also pro­vides out­stand­ing ac­cu­racy and pre­ci­sion. AI also helps speech recog­ni­tion tech­nol­ogy rec­og­nize speech con­text and tone.

Vir­tual As­sis­tants

The ad­vance­ments in speech recog­ni­tion have paved the way for in-ve­hi­cle vir­tual as­sis­tants. At first, driv­ers were very lim­ited with what they could do with speech recog­ni­tion tech­nol­ogy, but to­day, vir­tual as­sis­tants al­low driv­ers to ask for di­rec­tions, get gen­eral in­for­ma­tion and even ad­just their seat­ing po­si­tion and A/C set­tings. By learn­ing driver pref­er­ences, habits, rou­tines and even track­ing the user's lo­ca­tion, route and des­ti­na­tion, vir­tual as­sis­tants can make rec­om­men­da­tions on-the-go. It can re­mind its users to pick up items on their way home, rec­om­mend restau­rants in their area or even place their cof­fee or­der while they're on their way to their lo­cal cof­fee shop.


AI health bots are able to cover a vast num­ber of out­pa­tient ser­vices; they will ask you about your symp­toms and pro­vide you with the in­for­ma­tion you need to know about your med­i­cal con­di­tion by look­ing into the out­come of past treat­ments, as well as your per­sonal med­i­cal his­tory. AI as­sis­tants can also sus­tain con­tin­u­ous mon­i­tor­ing and care to the pa­tients who re­quire that sort of at­ten­tion, like in the case of men­tal health­care. More­over, bots can also com­mu­ni­cate with pa­tients on be­half of doc­tors to fol­low up on their progress, and re­vert back to the doc­tors with the feed­back and in­for­ma­tion re­lated to the pa­tient’s re­cov­ery jour­ney by us­ing Nat­u­ral Lan­guage Gen­er­a­tion and Pro­cess­ing (NLG/NLP) tech­nolo­gies.


AI and ML are cur­rently ca­pa­ble of un­der­stand­ing how the hu­man DNA func­tions and im­pacts life. Sys­tems such as Google’s Deep Mind and IBM’S Wat­son can digest im­mense amounts of data - like pa­tient records, clin­i­cal notes, di­ag­nos­tic images, treat­ment plans - and per­form pat­tern recog­ni­tion in a short span of time. By in­ter­pret­ing the hu­man genome, ML can pre­dict the molec­u­lar ef­fects of ge­netic vari­a­tion and iden­tify pat­terns across mil­lions of data points - a task that would take hu­mans for­ever to do. ML al­go­rithms can quickly scan a pa­tient’s per­sonal and fam­ily health records for sim­i­lar pat­terns and come up with sug­ges­tions that can lead to an early de­tec­tion, hence pre­ven­tion, of a dan­ger­ous dis­ease. With this process be­ing put into ac­tion, medicine could de­tect dan­ger­ous dis­eases such as can­cer and Alzheimer’s through very faint symp­toms, which in­creases the sur­vival rate or treat­ment op­tions of the pa­tient.


Ac­cord­ing to tech crunch, new drugs usu­ally take 12 to 14 years to be avail­able for com­mer­cial use. How­ever, with AI/ ML ap­pli­ca­tions on deck, the process is ac­cel­er­ated. Com­put­ers can mine pa­tient bi­o­log­i­cal data to un­der­stand why peo­ple sur­vive dis­eases and ap­ply the re­sults they found to im­prove cur­rent uti­lized ther­a­pies, or cre­ate new ones.


Sim­u­lat­ing a one-to-one hu­man tu­tor­ing ex­pe­ri­ence, ITS lever­age AI to de­liver learn­ing ac­tiv­i­ties that cater to stu­dents’ cog­ni­tive needs; it pro­vides tar­geted and timely feed­back. Many ITS use ma­chine learn­ing tech­niques, self­learn­ing al­go­rithms that ag­gre­gate and an­a­lyze large data sets, along with neu­ral net­works; this com­bi­na­tion al­lows the sys­tems to de­cide on the type of con­tent that should be de­liv­ered to the learner. On another note, the tu­tor­ing sys­tems that are model based uti­lize a num­ber of AI-ED tools that tai­lor the learn­ing ex­pe­ri­ence to the stu­dent’s cog­ni­tive and af­fec­tive states, al­lows them to dis­cuss and ques­tion the sub­ject be­ing taught, and in­clude open learner mod­els that mo­ti­vate the stu­dents by keep­ing them aware of their own progress, along with so­cial sim­u­la­tion mod­els that help the stu­dent un­der­stand the sub­ject by un­der­stand­ing the cul­ture and the so­cial norms be­hind it.


Vir­tual Re­al­ity ( VR) is all about sim­u­lated im­mer­sive ex­pe­ri­ences. It cre­ates an en­vi­ron­ment where learn­ers get the chance to ex­plore, in­ter­act, and ma­nip­u­late cer­tain el­e­ments. They are there­fore ca­pa­ble of us­ing th­ese vir­tual ex­pe­ri­ences in the real world. To­day, stu­dents can ex­plore a nu­clear power plant, wan­der through the streets of An­cient Rome, or or­bit around the outer plan­ets. Cou­pled with AI, VR be­comes in­tel­li­gent and de­liv­ers an op­ti­mized vir­tual ex­pe­ri­ence. It of­fers an en­vi­ron­ment that can in­ter­act or re­spond to the stu­dent’s re­ac­tions. In­tel­li­gent syn­thetic char­ac­ters are in­cor­po­rated to the vir­tual world; they can play roles in set­ting that can be dan­ger­ous or un­pleas­ant to the stu­dent.


Col­lab­o­ra­tive learn­ing has proven it­self to be a rather ef­fec­tive method of learn­ing, as it en­gages learn­ers and mo­ti­vates them. AI-ED tech­nol­ogy sup­ports many col­lab­o­ra­tive ap­proaches such as :

1. Adap­tive group for­ma­tion : Cou­pled with data about each learner in the class­room, AI’S goal is to de­sign a group­ing of stu­dents that share the sim­i­lar cog­ni­tive level and in­ter­ests.

2. Ex­pert fa­cil­i­ta­tion: AI tech­niques pro­vide col­lab­o­ra­tion pat­terns that are used as an in­ter­ac­tive sup­port to the col­lab­o­rat­ing stu­dents. For ex­am­ple, Markov mod­el­ing, an ap­proach us­ing the prob­a­bil­ity the­ory to rep­re­sent ran­domly chang­ing sys­tems, iden­ti­fies col­lab­o­ra­tive prob­lem-solv­ing strate­gies.

3. Vir­tual agents: They can act as an ex­pert a coach or a tu­tor, a vir­tual peer (fel­low in­no­va­tive stu­dent), or some­one the stu­dents have to teach them­selves.

4. In­tel­li­gent mod­er­a­tion: Us­ing ma­chine learn­ing and shal­low text pro­cess­ing tech­niques, they help the teacher in an­a­lyz­ing dis­cus­sions all the way to reach­ing a pro­duc­tive col­lab­o­ra­tion.

Banks en­gage with their clients ei­ther via in-per­son con­ver­sa­tions in a branch, or through tele­phone calls with sales or ser­vice rep­re­sen­ta­tives. Although this sort of in­ter­ac­tion has proven to be ex­tremely efffff­fec­tive, it also presents it­self to be costly. Chat­bots can be very help­ful on that level, as they main­tain quick and high-qual­ity cus­tomer ser­vice, while plum­met­ing banks’ ex­penses. In ad­di­tion to the above, banks can bet­ter con­nect with the mil­lenials by in­cor­po­rat­ing AI into their sys­tems. As they are con­sid­ered the largest users of mo­bile mes­sag­ing ser­vices, such as Face­book Mes­sen­ger, What­sapp, and Snapchat, banks will have a wide open win­dow of in­ter­ac­tion with this gen­er­a­tion and will ex­pand its client-base and con­ver­sion op­por­tu­ni­ties. More­over, the use of AI pro­motes user-tai­lored con­tent; fur­ther to ag­gre­gat­ing data re­lated to their sub­ject’s habits and needs, bank bots will be en­abled to prompt no­ti­fi­ca­tions about new prod­ucts that are avail­able and also send more per­son­al­ized mes­sages. On another note, banks lever­ag­ing voice en­abled chat­bots add an ex­tra layer of bio­met­ric se­cu­rity for their cus­tomers. The bot will rec­og­nize its users’ voice and there­fore al­lows them to ac­cess to their ac­count bal­ance, set up ac­count re­lated alerts, pay bills, and re­port a lost card, among many other func­tions.


Robo ad­vi­sors are dig­i­tal plat­forms that pro­vide au­to­mated and al­go­rithm-driven fi­nan­cial plan­ning ser­vices, and re­quire lit­tle to no hu­man in­ter­ven­tion. As any Ai-fo­cused sys­tem, robo-ad­vi­sors col­lect in­for­ma­tion re­lated to their clients’ fi­nan­cial sit­u­a­tion and fu­ture goals, and then em­ploys the ag­gre­gated data to offfff­fer ad­vice and/or au­to­mat­i­cally in­vest client as­sets.

To­day, robo ad­vi­sors are ca­pa­ble of per­form­ing so­phis­ti­cated tasks, such as tax-loss har­vest­ing, in­vest­ment se­lec­tion, and re­tire­ment plan­ning, all at lower costs and greater in­vest­ment out­come. Ac­cord­ing to In­vesto­pe­dia, client as­sets man­aged by robo-ad­vi­sors is pro­jected to surge to $2TN, af­ter hav­ing reached $60BN in 2015’s Q4. Other than be­ing an eas­ily ac­ces­si­ble tool that is avail­able 24/7, robo-ad­vi­sors offfff­fer a great ad­van­tage - they are low-cost al­ter­na­tives to tra­di­tional ad­vi­sors. By elim­i­nat­ing costs re­lated to hu­man la­bor, they can offfff­fer high-qual­ity ser­vice at a frac­tion of the cost. In fact, most of th­ese on­line plat­forms charge an an­nual flat fee of 0.2% to 0.5% of the client’s to­tal ac­count bal­ance, that com­pares to a rate amount­ing to 1% to 2% charged by a .hu­man fi­nan­cial plan­ner How­ever, de­spite its proven ef­fi­ciency, the robo-ad­vi­sor is still nascent tech­nol­ogy. Although it has au­to­mated some func­tions re­lated to as­set al­lo­ca­tions, port­fo­lio man­age­ment, and more, 40% of bank users would not be com­fort­able us­ing this tool alone in the times of ex­treme mar­ket volatility, as stated by In­vesto­pe­dia and the Fi­nan­cial Plan­ning As­sosi­ac­tion’s re­cent study.


By lever­ag­ing AI in­stru­ments, banks can iden­tify and pre­vent fraud and se­cu­rity hacks in real time. If a cus­tomer is us­ing a debit/credit card, the de­tec­tion en­gine can score trans­ac­tions within 0.3 sec­onds, and then flag fraud or ap­prove gen­uine trans­ac­tions with­out any in­ter­rup­tion of the pur­chase process.

AI and ML tech­nolo­gies can crunch a mas­sive num­ber of trans­ac­tions and flag any anom­alies; they are able to learn from one in­stance, and there­fore im­prove se­cu­rity. Some banks are also em­ploy­ing th­ese tech­nolo­gies to pay­ment providers, sup­port­ing se­cu­rity op­er­a­tions through­out the en­tirety of the pay­ment ecosys­tem. Th­ese tech­nolo­gies’ al­go­rithms can iden­tify pat­terns in the data to rec­og­nize fraud­u­lent claims, and by learn­ing from each case, they can au­to­mat­i­cally as­sess the sever­ity of dam­ages and pre­dict the re­pair costs based on his­tor­i­cal data, sen­sors, and images. Ai-driven tools can pre­vent false pos­i­tives and pro­vide a bet­ter de­tec­tion in place, which en­ables fraud in­ves­ti­ga­tion teams free to per­form tasks of higher value.


By al­low­ing AI to the banks back-of­fice op­er­a­tions, many la­bor-hours will be scratched offff the em­ploy­ees’ task-list. AI tech­nolo­gies can re­view loan agree­ments, iden­tify re­pay­ment pat­terns, and bring Ro­botic Process Au­to­ma­tion to pop­u­late data en­try and in­crease pro­cess­ing speed, specif­i­cally in the case of struc­tured data.

AI al­go­rithms are set to be more pre­cise given their ex­po­sure to great amounts of data. For in­stance, some banks are plan­ning to in­clude non-bank­ing data in the loan­ing process, such as Ama­zon in­ter­ac­tions, so­cial me­dia com­mu­ni­ca­tions, and sen­sors in phones such as GPS and ac­celerom­e­ters, to there­fore un­earth new ways to de­ter­mine the cred­it­wor­thi­ness and pro­vide the ad­e­quate fi­nan­cial help to their users.

Newspapers in English

Newspapers from Lebanon

© PressReader. All rights reserved.