Ar­ti­fi­cial In­tel­li­gence and the Fu­ture of Pri­vacy

Distinguished Magazine - - CONTENTS - KOEL CHAT­TER­JEE

As AI and ma­chine learn­ing rely com­pletely on data, gath­ered and an­a­lysed from var­i­ous sources, it is not tough to imag­ine the risks emer­gence of so­phis­ti­cated AI sys­tems, poses for the on­line pri­vacy of an in­di­vid­ual.

The ef­fec­tive­ness of data pro­tec­tion has been chal­lenged re­peat­edly over the last decade with in­ven­tions like cloud shar­ing, so­cial me­dia, on­line be­havioural tar­get­ing on mo­bile, etc. Ev­ery time tech­nol­ogy takes a new step to­wards as­sist­ing hu­mans, it poses an all-new kind of threat to in­di­vid­ual pri­vacy. The lat­est among them is In­ter­net of Things and Ar­ti­fi­cial In­tel­li­gence. As AI and ma­chine learn­ing rely com­pletely on data, gath­ered and an­a­lysed from var­i­ous sources, it is not tough to imag­ine the risks emer­gence of so­phis­ti­cated AI sys­tems, poses for the on­line pri­vacy of an in­di­vid­ual.

So, what are the ways in which AI can com­pro­mise the pri­vacy of an in­di­vid­ual? Pri­mar­ily, the promi­nence of AI has been pos­si­ble for three fac­tors, speed, scale and au­to­ma­tion. The an­a­lyt­i­cal speed of AI, for the com­pu­ta­tion of large data sets, has been far ahead of the ca­pa­bil­i­ties of hu­man anal­y­sis. Also, the ef­fi­ciency of AI for anal­y­sis of big data, and the per­for­mance of the tasks des­ig­nated to it, with­out any hu­man su­per­vi­sion have made it es­sen­tial. This has also made it sus­cep­ti­ble to ill us­age by cor­po­ra­tions and large-scale agen­cies, which have reg­u­lar ac­cess to in­di­vid­ual pro­files.

In gen­eral, peo­ple re­main un­aware of the im­mense amount of data that can be gen­er­ated by elec­tronic de­vices – from smart home ap­pli­ances to so­phis­ti­cated com­puter and phone ap­pli­ca­tions, with the in-built soft­ware. As a re­sult, in­di­vid­u­als are ig­no­rant of how vul­ner­a­ble they are to data ex­ploita­tion.

The big­gest prob­lem with AI lies here. An AI As­sis­tant can func­tion prop­erly only when it has ac­cess to the user’s per­sonal data, which ranges from lo­ca­tion and con­tacts and to ac­count in­for­ma­tion and pho­tos. Also, mes­sages and data and all the in­for­ma­tion on a de­vice, have to be un­en­crypted for the AI As­sis­tant to func­tion and as­sist the users with their daily ac­tiv­i­ties. Even if a user anonymizes his per­sonal data, once that be­comes a part of the big data set, an AI can de-anonymize it, based on in­fer­ences from other data sources. Also, as more and more so­phis­ti­cated AI sys­tems are be­ing built, iden­ti­fi­ca­tion through voice and fa­cial recog­ni­tion, across mul­ti­ple de­vices is be­com­ing a man­date.

This makes per­sonal data a part of the non-per­sonal big data pool, which can be ac­cessed with ease, for the pur­pose of pro­fil­ing, sort­ing or eval­u­at­ing peo­ple, with­out their con­sent. As AI can de­duce and anal­yse sen­si­tive data from the non-sen­si­tive, it can be uti­lized to ef­fec­tively pre­dict and in­fer emo­tional be­hav­iour, sex­ual ori­en­ta­tion, po­lit­i­cal views, and cur­rent health con­di­tions, from seem­ingly in­sipid data like so­cial met­rics, lo­ca­tion changes, ac­tiv­ity logs and even, typ­ing speed and pat­terns. And if the in­for­ma­tion col­lec­tion and anal­y­sis done by AI do not seem dan­ger­ous enough, a re­cent study by Black Hat has shown how ar­ti­fi­cial hack­ers have com­pletely out­per­formed their hu­man coun­ter­parts, and what can be the re­sult when AI is weaponized for steal­ing iden­ti­ties.

As the world of Ar­ti­fi­cial In­tel­li­gence con­tin­ues to boom, and the world’s ten­dency to be com­pletely de­pen­dent on tech­nol­ogy keeps es­ca­lat­ing, it is fast be­com­ing unattain­able to pro­tect one’s pri­vacy on­line. It is more of a con­cern for those, who are not ac­quainted with the nec­es­sary mea­sures they can still par­take in or­der to pro­tect their pri­vate data. Though not fool-proof, one can in­crease their re­sis­tance to un­in­ten­tional data shar­ing and shrink their on­line foot­prints. For starters, us­ing VPN (Vir­tual Pri­vate Net­work) or anony­mous net­works, such as TOR (The Onion Router) or Freenet, while brows­ing the net help a user pro­tect their on­line pri­vacy. As these net­works of­fer end-to-end en­cryp­tion and hide the real IP, the data trans­ferred dur­ing a ses­sion can­not be tapped into. Open source Op­er­at­ing Sys­tems, like Linux dis­tri­bu­tions, pre­vent col­lec­tion of data with­out per­mis­sion, un­like op­er­at­ing sys­tems by Ap­ple or Mi­cro­soft. Lastly, brows­ing through open source browsers like Fire­fox, or brows­ing in Incog­nito mode, can re­duce the on­line foot­print of a user.

Ar­ti­fi­cial In­tel­li­gence, be­ing a rel­a­tively re­cent tech­no­log­i­cal in­no­va­tion still has miles to go. Although, it has al­ready been used to make sub­stan­tial pos­i­tive con­tri­bu­tions to the lives of in­di­vid­u­als, and to the world as a whole. Un­doubt­edly, its po­ten­tial will be fur­ther uti­lized to tackle var­i­ous evils of the so­ci­ety and the world. How­ever, just like any other tech­no­log­i­cal in­no­va­tion, its fu­ture will be de­cided by the in­tent of the peo­ple who use it. And while none can pre­dict its suc­cess, it might be safe to take pre­cau­tions, oth­er­wise, the tool in­vented to as­sist hu­mans can very well lead us to our doom.

Newspapers in English

Newspapers from India

© PressReader. All rights reserved.