GPT
Posted on
July 5, 2023
As published in the Time of India
ChatGPT has the distinction of being the fastest-growing consumer application in history; it reached a hundred million monthly users, a mere two months after launch. Why are we all so enamored by ChatGPT? Language - from time immemorial was considered a uniquely human capability. Computers powered by Generative AI seem to have learned language capability and are able to converse with us fluently and intelligibly about almost any topic. Even from a long-term evolutionary perspective, this is a big deal.
When I was about 13 yrs old (the 1970s), my mother took me to the Canara Bank on our very own street in Mysore and opened a “kiddie’s account” which came with a nice kid-friendly passbook. I remember her depositing Rs12 into the account and handing me the passbook - as though to say “now go figure out banking”. I diligently deposited the small birthday cash gifts or withdrew a few rupees to buy wicket-keeping gloves etc. Quickly I learned about fixed deposits and interest income and generally how banks worked. Banks were designed for the urban middle class, with forms and transactions conducted largely in English.
Fast forward to 2023, While India’s literacy rate is at 77%, with English speakers around 250 million which is less than 18% of the population, and much of them are still concentrated in the urban centers. Large swathes of our rural areas naturally don't speak English, nor do we have a strong branch banking network. How do we bring full banking access to every Indian? Can we deliver banking in all 22 major Indian languages in this increasingly digital world? Can the bank’s computers understand, converse and transact in all these languages fluently?
I think we can! Finally, we have GPT-like Generative-AI technology that seems to have the language capability including language translation and text2speech and speech2text capability to converse in human-like ways in various languages to provide full banking access to every Indian - towards true financial inclusion.
As banks worldwide try to embrace digital technologies and cloud computing to service a large customer base on diverse client devices, generative AI promises to play an important role in improving banking and lending in particular. How exactly can these GPT-like Large Language Models (LLMs) be useful to banking, particularly in lending?
At its most basic Generative AI can provide a conversational UI interface that can be used by customers to “talk” to the bank app in their own language and get an accurate response about specific type of loan products they are looking for. They should be able to ask in their native language all their queries about the loan type, interest rates, qualification criteria, or processing fees.
Secondly, Generative AI can also complete the loan origination journeys to onboard the customer by filling out the forms, using conversational UI - speech. Meaning the Generative AI system could ask the customer in their own language for their name, date of birth, address, etc. which is simply spoken out. Aadhaar or video KYC should be integrated into smartphones in a simple and convenient workflow.
Third, Generative AI can be a much more effective “Customer Support” tool, since without getting tired, it can go over the customers' entire support history, understand all the financial products s/he is using, have a completely updated understanding of the financial products offered, and then respond to the customer’s query or issue - in the most accurate and personalized manner in the customer’s own language. It can at once improve customer support and reduce the cost of such high-quality service. This can be a game changer.
Trustt Inc the SaaS Digital Banking startup was one of the first to roll out a GPT-based banking product and has launched the Trustt GPT suite in 12 Indian languages, including Conversational Product Inquiry, Conversational Customer Acquisition, and Conversational Customer Support.
Generative AI is not without challenges. Firstly GPTs hallucinate - meaning makeup facts and confidently respond to questions with untruths and made-up information. Trustt has had to spend a lot of effort in addressing this hallucination problem. Secondly in order to ensure full compliance with the privacy and data localization needs, one cannot use the US-hosted API GPT-like services. While the cloud players are all gearing up to offer privacy and data-localized solutions, On-prem deployments need to also be considered.
In summary, generative AI can transform banking and truly bring about financial inclusion for every Indian, by bringing an easy-to-use conversational interface in all Indian languages. I think Generative AI should become a key capability of IndiaStack’s Digital Public Infrastructure that powers Aadhaar (identity), UPI(payments), eSign(digital signatures), DigiLocker(verified certificates), or ONDC(eCommerce). This will enable not only financial inclusion but also inclusive development.
Srikanth Nadhamuni was the founder CTO of Aadhaar, he’s also the Chairman of Trustt a SaaS Digital banking startup, and Managing Trustee of eGovernments Foundation an NGO he co-founded along with Nandan Nilekani
