AI sce­nario in bank­ing

A fo­cused in­ter­view with Van Baker, Vice Pres­i­dent, Gart­ner:

Banking Frontiers - - Deep Insights - Manoj@bank­ingfron­tiers.com

Manoj Agrawal: How many magic quad­rants for AI/ML/DL? Which was the last and which is the next?

Van Baker: On AI there is none. We have magic quad­rants on an­a­lyt­ics. The mar­ket is too im­ma­ture. We try and avoid DL as ter­mi­nol­ogy and focus on DL net­work. We won’t do magic quad­rants on AI over­all, but we have guides on vir­tual cus­tomer as­sis­tance, hosted AI ser­vices and con­ver­sa­tional plat­forms. The ear­li­est we can have a magic quad­rant on AI is at least 12-18 months away. Lots of AI companies have good aware­ness and rev­enue… but not on prof­itabil­ity or cash flow.

What are the dif­fer­ences be­tween AI of­fered by big tech, well-funded star­tups and less funded star­tups?

From a ML per­spec­tive and broad ap­pli­ca­tion per­spec­tive, you need a lot of re­sources – CPU, mem­ory, data. Es­pe­cially so for un­su­per­vised learning. That lim­its it to large companies like Google, IBM, etc. Or else, some­one should have the abil­ity to tap re­sources on cloud. Even a small AI re­quires big re­sources for the de­sign, but not for pro­duc­tion. For pro­duc­tion, the al­go­rithm could run on a phone.

Early adopts will face the chal­lenge of hav­ing adopted a weaker AI tech­nol­ogy in some of their cases. How can they fig­ure out when their choice is no longer the best and what chal­lenges will they face in shift­ing from one AI tech to an­other?

Some providers in the mar­ket tightly cou­ple chabots with their own NLP en­gine. We think there is risk as­so­ci­ated with that. We are start­ing to see the emer­gence of mid­dle­ware in the API space. Chat­bot providers like Liveper­son have a lot of in­te­gra­tion ca­pa­bil­ity and bot build­ing ca­pa­bil­ity, but you can bring bots built us­ing other frame­works and bring them into their en­vi­ron­ment. So flex­i­ble frame­works are emerg­ing. When I talk to Ama­zon, Google, etc, about how of­ten they com­pletely re­build the AI en­gine and cre­ate a new algo, the an­swer is every 6 months. So, it is im­por­tant to have flex­i­bil­ity.

Same AI tools are avail­able to every­one – so will AI change the na­ture of com­pe­ti­tion? How?

It will, be­cause in this era of dig­i­tal biz, where the pace of biz is chang­ing very rapidly and the ap­pli­ca­tion of AI will give in­sights, that will give a brief com­pet­i­tive ad­van­tage. So, the com­pet­i­tive ad­van­tage is not going to be sus­tain­able and there will quickly be a level playing field. Data has no value, so dif­fer­en­ti­a­tion will come from build­ing the neu­ral net model, which is a very chal­leng­ing task. Neu­ral model will be the next fron­tier of com­pet­i­tive ad­van­tage – and that is a chal­lenge for data sci­en­tists and ex­perts to build un­su­per­vised / su­per­vised / re­in­forced mod­els. It is a mix of in­tu­ition and sci­ence.

How can one dis­tin­guish be­tween in­tel­li­gence and wisdom – so as to not lose wisdom to AI – eg wisdom about cod­ing or soft­ware ar­chi­tec­ture or us­age in­sights or vul­ner­a­bil­i­ties or hu­man psy­chol­ogy?

Ma­chines don’t think. AI is al­go­rithms. They are di­gest­ing a lot of data and iden­ti­fy­ing re­la­tions. All this is ma­chine learning. ML can han­dle much larger data sets com­pared to hu­mans. But what is miss­ing is com­mon sense. ML is ac­tu­ally pat­tern recog­ni­tion.

Are VCs us­ing AI to an­a­lyze in­vest­ment pro­pos­als?

It could be ap­plied to a por­tion of the anal­y­sis. VCs look not only at pro­posed prod­uct and its fea­tures and ca­pa­bil­i­ties, but also for lead­er­ship qual­i­ties and peo­ple skills.

Which sec­tor of fi­nan­cial sec­tor is lead­ing the ef­fec­tive de­ploy­ment of AI?

In­vest­ment bank­ing.

Newspapers in English

Newspapers from India

© PressReader. All rights reserved.