ma­chine power

Dr Pradeep Pendse, Wel­ingkar In­sti­tute, draws the dis­tinc­tion be­tween au­to­mated and au­ton­o­mous en­ter­prises.


..what is ma­chine learn­ing, ex­actly? Stan­ford Univer­sity com­puter sci­ence pro­fes­sor An­drew Ng de­fines it as “the sci­ence of get­ting com­put­ers to act with­out be­ing ex­plic­itly pro­grammed.” In fun­da­men­tal terms, ma­chine learn­ing is a branch of ar­ti­fi­cial in­tel­li­gence that is meant to repli­cate the way hu­mans take in in­for­ma­tion from their en­vi­ron­ment to make bet­ter-in­formed choices for the fu­ture*. But real­is­ti­cally, will ma­chine learn­ing trans­form the way busi­nesses are man­aged?

It was a reg­u­lar work day—I had en­tered my office premises, showed my face to the small screen on the at­ten­dance sys­tem and it marked my at­ten­dance. Once in­side my cabin, I was re­flect­ing on what just hap­pened and I re­al­ized that face recog­ni­tion and other forms of bio­met­ric so­lu­tions, in­clud­ing fin­ger­print scan­ning and retina scan­ning, have been in­deed a part of a larger wave called Ar­ti­fi­cial In­tel­li­gence (AI) which be­gan a few decades ago but seem to have grad­u­ally en­tered our daily lives. House­hold ro­bots, ro­botic arms in au­to­mo­bile fac­to­ries and haz­ardous en­vi­ron­ments, rec­om­men­da­tion en­gines used in web­sites and searches on net, and self­driv­ing cars are but some ex­am­ples of how AI is pen­e­trat­ing our lives.

what is AI and ma­chine learn­ing

AI, which traces its roots to re­search done by sci­en­tists for over a cen­tury, has been ex­plored in many dif­fer­ent ways—though not al­ways suc­cess­fully. Mim­ick­ing phys­i­cal tasks done by hu­mans (ro­bot­ics), repli­cat­ing [the func­tions of] sense or­gans (mo­tion sens­ing, face recog­ni­tion, etc.), cog­ni­tion, rule-based de­ci­sion-mak­ing, solv­ing puz­zles,

play­ing games (IBM’s Deep Blue against Gary Kas­parov), repli­cat­ing prac­tices of ex­perts in var­i­ous do­mains (ex­pert sys­tems) are ex­am­ples of ap­proaches taken to­wards un­der­stand­ing AI and de­vel­op­ing new ap­pli­ca­tions for it.

AI is pri­mar­ily about cre­at­ing ma­chines ca­pa­ble of demon­strat­ing some as­pects of hu­man in­tel­li­gence. The next fron­tier in AI is to build into the ma­chine the hu­man ca­pa­bil­ity to ac­quire new learn­ing based on past ex­pe­ri­ences, emerg­ing pat­terns, and pre­dic­tions. This is what is known as ma­chine learn­ing. It is this abil­ity which can make the ma­chine adapt­able at the least and au­ton­o­mous in its ex­treme form, as ex­em­pli­fied in part by a self-driv­ing car. AI has so far used sta­tis­ti­cal anal­y­sis, neu­ral net­works, and the like to help in train­ing a ma­chine based on past data and pat­terns. Au­ton­omy, how­ever, means that the ma­chine not only op­er­ates on ev­ery new sit­u­a­tion based on the train­ing data pro­vided, but also self­cor­rects the learn­ing data.

en­ter­prises: au­to­mated or au­ton­o­mous?

With in­creased us­age of IT-based tech­nolo­gies, mo­bile and the cloud, along with re­duc­tion in costs of pro­cess­ing, trans­mis­sion, and stor­age, all en­ter­prises have ac­cu­mu­lated a vast amount of trans­ac­tion data as well as sub­ject-based repos­i­to­ries of the key as­pects of their busi­ness such as prod­ucts, pro­cesses, em­ploy­ees, as­sets, and cus­tomers. This ac­cu­mu­la­tion and cen­tral­ized avail­abil­ity of trans­ac­tion and sub­ject-based data have been a re­sult of im­ple­ment­ing work­flow, busi­ness process au­to­ma­tion, en­ter­prise ap­pli­ca­tions, con­tent man­age­ment so­lu­tions, and data ware­houses.

So­cial me­dia and the IoT have cre­ated the need for a cul­ture of real-time sens­ing and re­spond­ing. This goes far be­yond the tra­di­tional means of uti­liz­ing data in the form of MIS and busi­ness in­tel­li­gence. Many of these or­ga­ni­za­tions have thus reached an ad­vanced stage of what can at best be de­scribed as au­to­ma­tion.

Some of them have gone a step for­ward and brought in an­a­lyt­ics in the form of end-user an­a­lyt­ics (ad hoc query and data vi­su­al­iza­tion), and used an­a­lyt­ics in ar­eas such as cus­tomer seg­men­ta­tion and tar­get­ing for the pur­pose of plan­ning pro­mo­tional cam­paigns. A few of them have built rule-based de­ci­sion sys­tems—these rules can be mod­i­fied but the need for a new rule or a change in an ex­ist­ing one still needs to be done man­u­ally.

Per­haps the more es­tab­lished field of man­u­fac­tur­ing has been us­ing more so­phis­ti­cated forms of au­to­ma­tion such as ro­botic arms or SCADA and process con­trol sys­tems. How­ever, these too tend to­wards au­to­ma­tion and do not qual­ify to be called au­ton­o­mous sys­tems.

Some ev­i­dence of AI and ma­chine learn­ing is how­ever ev­i­dent in so­lu­tions such as chat­bots such as the fe­male voice in a GPS sys­tem which helps us nav­i­gate while driv­ing the car, or a front-end de­vice which af­ter a bit of train­ing starts rec­og­niz­ing text writ­ten by a user us­ing a sty­lus, or a voice-based sys­tem which rec­og­nizes the voice of a per­son and acts on a com­mand. Even ad­vanced an­a­lyt­ics so­lu­tions now be­ing used in cus­tomer [re­la­tion­ship man­age­ment] and risk [man­age­ment] bank­ing and net foren­sics do have an el­e­ment of learn­ing, if not au­ton­o­mous be­hav­ior.

to­wards an au­ton­o­mous en­ter­prise

What dif­fer­en­ti­ates an au­ton­o­mous en­ter­prise from an au­to­mated one is the abil­ity to sense the con­text, pre­dict fu­ture pos­si­bil­i­ties, and act on the ba­sis of these pre­dic­tions with­out any hu­man in­ter­ven­tion. For ex­am­ple, the risk and com­pli­ance process in a fi­nan­cial in­sti­tu­tion could ac­quire

Au­ton­omy, how­ever, means that the ma­chine not only op­er­ates on ev­ery new sit­u­a­tion based on the train­ing data pro­vided, but also self-cor­rects the learn­ing data.

new knowl­edge based on pre­vi­ous risks, and pre­dict fu­ture risk sce­nar­ios based on ex­tra­ne­ous syn­di­cated in­for­ma­tion on con­tex­tual fac­tors. Equipped with this in­for­ma­tion, the risk man­age­ment sys­tem could al­ter its thresh­old lev­els or trig­ger cer­tain ac­tions on its own. This would be con­sid­ered an au­ton­o­mous sys­tem.

Like­wise, an on­line au­ton­o­mous health ad­vi­sor agent who has ac­cess to real-time mon­i­tor­ing data of peo­ple liv­ing in the lo­cal area could pre­dict the out­break of an epi­demic, or at least be able to di­ag­nose a pa­tient more ac­cu­rately based on his or her health data and avail­able com­mu­nity in­for­ma­tion.

What or­ga­ni­za­tions need to do is study their own busi­ness pro­cesses and de­ci­sions, and iden­tify where AI or ma­chine learn­ing could be ap­plied to sub­sti­tute or sup­ple­ment the hu­man ex­pert re­spon­si­ble for mak­ing the de­ci­sions or run­ning the pro­cesses with an equiv­a­lent self-learn­ing ma­chine or al­go­rithm. Ap­prov­ing loan ap­pli­ca­tions; pre­dict­ing de­faults; ad­vis­ing on in­vest­ments, in­sur­ance or health; proac­tive rec­om­men­da­tion en­gines in var­i­ous sit­u­a­tions, and pro­vi­sion­ing of a guest room based on knowl­edge of past pref­er­ences and be­hav­iors of a guest (in the hospi­tal­ity in­dus­try) are but a few pos­si­bil­i­ties.

It has al­ready been dis­cussed that much of the le­gal pro­cess­ing could be au­to­mated—per­haps some­day even judge­ment agents/bots could help clear the back­log of rou­tine le­gal cases which are cur­rently over­load­ing our ju­di­ciary.

The ‘smart’ be­hav­ior ex­pected in smart cities could be an­other great op­por­tu­nity for chang­ing the way we live. A self-pro­gram­ming sig­nal­ing sys­tem which al­ters it­self

Or­ga­ni­za­tions need to study their own busi­ness pro­cesses and de­ci­sions, and iden­tify where AI or ma­chine learn­ing could be ap­plied to sub­sti­tute or sup­ple­ment the hu­man ex­pert.

de­pend­ing on traf­fic con­di­tions, or an au­ton­o­mous road bar­rier which redi­rects traf­fic if the nor­mal route gets choked, are all in­ter­est­ing pos­si­bil­i­ties in the real world.

im­pact of au­ton­o­mous en­ter­prises

While au­to­ma­tion in en­ter­prises helped in de­lay­er­ing an or­ga­ni­za­tion by tak­ing away the bur­den of in­for­ma­tion cap­ture, pro­cess­ing, and re­port­ing thereby im­pact­ing the su­per­vi­sory and white-col­lar cler­i­cal lay­ers, au­ton­o­mous pro­cesses could take away much of the op­er­a­tional and to some ex­tent, even tac­ti­cal de­ci­sion-mak­ing. It will make busi­ness pro­cesses quicker and more re­li­able, ac­cu­rate, rel­e­vant, and scal­able. In its ex­treme, an au­ton­o­mous or­ga­ni­za­tion could well con­tinue to keep work­ing with­out man­ual in­ter­ven­tion un­til the pur­pose for which such an or­ga­ni­za­tion was cre­ated is no longer rel­e­vant in the real world. Re­defin­ing the pur­pose will re­main a hu­man task at least for a long time to come.

the prom­ise

AI and ma­chine learn­ing have a lot to of­fer. The day is not too far when you may stop own­ing a car and just step into the first empty car near your home and ex­pect it to drive you to your des­ti­na­tion; when var­i­ous ser­vices and of­fers are sent to you just when you need it; when some bot or agent proac­tively de­tects your health prob­lem and ac­ti­vates the health­care de­liv­ery sys­tem. ■

Newspapers in English

Newspapers from India

© PressReader. All rights reserved.