DBS began its AI journey about ten years before 2024, as part of its digital transformation.
“our journey on AI started 10 years back. And the genesis of that was actually part of our whole transformation to adopt digitalization for the bank.”

14 Nov 2024 · 33m · Lim Hinchuan, Phoram Mehta, Chua Tian Tian · 24 quoted facts
Banks and payment firms that treat AI as a long-term capability, not a quick experiment, are getting measurable returns. DBS and PayPal both argue that years of groundwork on data, rules and people let them adopt generative AI fast, while quantum computing is a looming threat and opportunity that needs early preparation.
The tension runs between speed and safety. Both guests describe big gains, from customer nudges to fraud detection, but warn about hallucinations, deepfakes and flattering machines. They favour proportional regulation and industry collaboration. Quantum is framed differently: not yet mature, yet 'harvest now, decrypt later' attacks mean organisations should start preparing now.
DBS spent about seven years building its ADA data lake, cleaning data and setting metadata and lineage rules. Lim Him Chuan says this let the bank react quickly when generative AI emerged in November 2022.
DBS's early wealth-advisor project with IBM Watson failed because natural language processing was premature; relationship managers preferred Google. The bank treated it as a first learning attempt and later built three pillars: data platform, guardrails and people.
Phoram Mehta says more than 600 models decide within milliseconds whether a transaction should be inspected, allowed or blocked. This runs at the scale of 400 million customers and over 20 million merchants.
Mehta says explaining decisions suits a merchant refused credit but would be foolish for a blocked fraudster. DBS screens AI with its PURE test (purposeful, unsurprising, respectful, explainable) and the regulator's FEAT principles.
Mehta warns adversaries may store encrypted data today to decrypt once capable quantum computers exist. Him Chuan notes quantum remains immature, costly, hard to cool and short of expertise, yet says starting now avoids missing the boat.
Lim Him Chuan said DBS has delivered close to 370 million in outcomes so far, expects 700 million this year and aims for 1 billion next year. One example is AI nudges to over 8 million customers, which he said raised savings by more than 80% among those nudged in Singapore.
Phoram Mehta of PayPal named deepfakes, fake audio or video used for scams and misinformation. Lim Him Chuan pointed to hallucinations (AI stating false things confidently) and to AI trying to please the user.
Phoram Mehta explained it as attackers collecting encrypted data today, hoping to unlock it once quantum computers can break current public-key encryption. That is why he urges organisations to start changing their cryptography and stop collecting unnecessary data.
Lim Him Chuan said PURE asks whether an AI use is purposeful, unsurprising, respectful and explainable. If a model or data use fails these tests, DBS will not deploy it.
Lim Him Chuan said the technology may need another ten years to mature, isn't plug-and-play, needs heavy investment and cooling, and lacks expertise. Generative AI, by contrast, can be used by simply downloading an app.
Stated by the speakers. Each quote matches the transcript, which shows what was said, not that it is true.
DBS began its AI journey about ten years before 2024, as part of its digital transformation.
“our journey on AI started 10 years back. And the genesis of that was actually part of our whole transformation to adopt digitalization for the bank.”
DBS partnered with IBM to build a Watson-based wealth advisor for relationship managers, inspired by Watson's Jeopardy win.
“we noted that IBM created a machine known as Watson that actually won the American Quiz Show Jeopardy. So we wanted to partner IBM to create a wealth advisor”
The DBS Watson wealth advisor failed because natural language processing was premature and relationship managers preferred Google.
“NLP at the point in time, we were a bit premature. So it didn't quite work out. So our RMS prefer to Google for answers.”
DBS built a data lake called ADA, named after Ada Lovelace, which took about seven years to create.
“we created this data lake known as ADA, which is named after the first computer programmer, Lady Lovelace ADA. And that took us about seven years to create”
DBS's data chapter community has 700 data professionals, including 200 data scientists, within the GSTAR organisation.
“There are 700 data professionals in there, including 200 data scientists. So these 700 people actually are part of my GSTAR organization.”
PayPal runs more than 600 AI models that decide within milliseconds whether a transaction should be inspected further, allowed or blocked.
“we have more than 600 models that are making decisions within milliseconds to decide whether a particular transaction, a particular activity should be inspected further, allowed or not.”
PayPal has about 400 million customers and more than 20 million merchants worldwide, according to Phoram Mehta.
“PayPal scale of 400 million customers with more than 20 million merchants all over the world and continuing to grow.”
PayPal is available in more than 200 markets, according to Phoram Mehta.
“we're available now in 200 plus market to create that opportunity to transform not just the buying experience”
The host stated that DBS's AI industrialisation has delivered over 350 use cases and 800 models.
“DBS's industrialization approach, that has helped it deliver over 350 use cases spanning from customer-facing businesses and support functions and also 800 models to date.”
DBS has used AI to nudge more than 8 million customers in Asia, including over 3 million in Singapore, to improve their financial behaviour.
“today, we have used AI to nudge more than 8 million customers across all locations in Asia. And in Singapore alone, more than 3 million customers actually have been nudged”
Lim Him Chuan said nudged DBS customers increased savings by more than 80% and invested four times more.
“So the savings actually has increased by more than 80%. They have also invested more by four times.”
Lim Him Chuan said DBS's AI has delivered close to 370 million in outcomes, targeting 700 million this year and 1 billion next year.
“we have already delivered close to 370 million of outcomes for DBS. This year, we're looking at 700 million. and then next year we are aiming for $1 billion.”
Lim Him Chuan said many DBS generative AI use cases arose bottom-up from staff after generative AI took off in November 2022.
“when Gen AI exploded back in November 2022, we were able to quickly... Actually, a lot of use cases didn't even come from the center. It was bottom-up.”
DBS created an internal tool called DBS-GPT so staff can use generative AI with sensitive bank information rather than a public service.
“our DBS-GPT, because while the OpenAI one is good, but it is not in terms of so feasible for us to put sensitive information into the public domain.”
PayPal's customer service agents use AI drawing on more than 200 petabytes of data collected over PayPal's 25 years.
“we already have our customer service agents leveraging AI to ensure that not only are they using the data, the 200 plus petabytes of data that we have collected over 25 years of existence”
The host said the Monetary Authority of Singapore published a report on generative AI risks and opportunities for banks in May 2024.
“The MAS, I believe, came up with a report on the emerging risks and also opportunities of generative AI for banks back in May this year.”
Phoram Mehta cited the EU AI Act as coming into effect on August 1st, in contrast with Singapore's approach.
“how Europe has taken on addressing this challenge with the EU AI Act going in effect on August 1st”
Phoram Mehta cited Singapore's Project Moonshot and AI Verify as helping organisations deploy AI responsibly.
“The projects like Project Moonshot, AI Verify, helping organizations, individuals, equip them with the right knowledge on how to do that”
DBS uses a framework called PURE (purposeful, unsurprising, respectful, explainable) to decide whether to use AI models or data.
“we created a framework known as PURE, P as in purposeful, U as in unsurprising, R, respectful, and E, explainable.”
PayPal is a founding member of the Coalition for Secure AI, an open-source coalition whose members include Google, IBM, Microsoft, Nvidia, OpenAI and Anthropic.
“PayPal's a founding member of Coalition for Secure AI, an open source coalition that has come together globally”
Lim Him Chuan said the MindForge Consortium, led by the Monetary Authority of Singapore, brings industry players together to co-create AI standards.
“There's also a consortium. There's the MindForge Consortium. I think that's led by the MAS that comprises friends from different parts of the industry”
The host cited McKinsey as estimating the quantum computing market could reach up to 131 billion US dollars by 2040.
“McKinsey coming up earlier to say that the market size of quantum computing may reach up to $131 billion by 2040”
Phoram Mehta described 'harvest now, decrypt later', where adversaries store encrypted data today to decrypt it with future quantum computers.
“adversaries are storing encrypted data today in the likelihood that once they get their hands on the quantum computer that is capable of decrypting them, then all of that data will be exposed.”
Lim Him Chuan said he had read that quantum computing may take another ten years to mature.
“I read somewhere that it's probably going to take another 10 years for it to be really mature, right?”
Explains how an early Watson failure led DBS to build a data platform, guardrails and a data community.
“we created this data lake known as ADA, which is named after the first computer programmer”
Shows how AI already underpins PayPal's decisions across hundreds of millions of customers.
“we have more than 600 models that are making decisions within milliseconds”
Quantifies DBS's revenue uplift targets and PayPal's personalised payment uses.
“we have used AI to nudge more than 8 million customers across all locations in Asia”
Describes DBS-GPT and enterprise knowledge tools, and PayPal's staff and agent use.
“we have created an enterprise knowledge base for our colleagues to be able to search and summarize information”
Both leaders name the dangers that concern them most.
“One word would be deep fakes.”
Covers proportional regulation, PURE and FEAT principles, and industry coalitions.
“we ask ourselves, can we do it? Can we do it?”
Explains why quantum matters, the harvest-now-decrypt-later risk and adoption hurdles.
“adversaries are storing encrypted data today in the likelihood that once they get their hands on the quantum computer”
The best of money matters. MoneyFM 89.3. Your ears on the ground at Singapore FinTech Festival 2024. That's right, you're listening to a special broadcast of MoneyFM 89.3, right from Singapore Expo Hall 2, where the Singapore FinTech Festival 2024 is held. I'm your host for the next hour, and I'm Chua Tian Tian. Now, this year's Singapore FinTech Festival surrounds six major themes that will shape up 2025. At the front and centre of discussions as the roadmap for AI and quantum, and this is what we are going to talk about in our two-part panel discussion today.
Of particular focus this year is that boom in generative AI, which uses algorithms to identify patterns within existing data to come up with new content, be it text, images, videos, or even quotes. Now, the technology is set to hold the potential to radically change up the financial services industry with an EY analysis suggesting that rethinking the financial institutions with Gen.AI at the core has the potential to create 200 billion US dollars to 400 billion US dollars by 2023, with productivity gains reaching up to 30 % by 2028.
So in part one of our two-part conversation on the roadmap for AI and quantum, we want to look at where fintech players are right now when it comes down to generative AI adoption and what the road to achieving that AI-embedded future will look like for them. So for more, I've got with me here at our MoneyFM booth at Singapore Expo, Hall 2, Lim Him Chuan, Group Head of Strategy, Transformation, Analytics and Research, or GSTAR, as well as Group Executive at DBS.
and we've also got Phoram Mehta, Senior Director and Chief Information and Security Officer for International Markets at PayPal. Welcome Phoram and also Him Chuan.
Thank you so much, Tien Tien, for having us. It's really nice to meet both of you. Yeah, it's my honour to be here today. Exciting morning so far.
Yeah, and I hope it's been an exciting day so far for you as well, Phoram.
Absolutely. Really, really excited to be here in the midst of the entire industry, not just from Singapore, but from all over the world.
Yeah, so before we dive into the conversation proper, could you each spend one minute to briefly introduce yourself and your roles at maybe PayPal and DBS respectively, particularly when it comes down to AI and quantum?
Actually, my team is known as GSTAR, as you mentioned, and the A in the GSTAR stands for AI. So one of my domains is to make sure that as the bank progresses with the use of AI, it is not just about technology, but it's really how do we use that to drive business outcomes and strategies so that we can provide better experiences and outcomes for customers and employees and shareholders. So that's what my team does, which is to look at both strategy, transformation empowered by AI and research.
And what about you, Voram? Tell us more about your role at PayPal.
So as you mentioned, I'm part of the PayPal's information and cybersecurity organization. I've been in the industry 25 years, of which the last 15 have been with PayPal. And as you think of my role, it is both protection for our customers as well as enablement. And technologies like AI and quantum are incredibly exciting to ensure that we can leverage the innovation, do it in a responsible and protective manner.
You know, I recall us sitting down at that cafe at SunTech sometime back about one year ago and talking about quantum for like one and a half hours. So we can expect more conversations coming up. Let's talk more about the roadmap for AI and quantum, which is their theme for today. I do want to start with DBS and PayPal's approaches towards both technologies where possible. Because if I were to go to you first, Him Chuan, DBS embarked on its journey with AI since a decade ago in 2014.
It also kicked off a comprehensive AI industrialization transformation at one point within the past decade. So tell us more about this journey and how Generative AI has augmented that path for the bank.
So indeed, our journey on AI started 10 years back. And the genesis of that was actually part of our whole transformation to adopt digitalization for the bank. And as you know, DBS has been recognized as the best digital bank in the world for several times. and a key driver for that is the use of AI in this whole process. And when we started the journey, we were trying to learn what it means to actually use AI.
And it so happened that at a point in time, we noted that IBM created a machine known as Watson that actually won the American Quiz Show Jeopardy. So we wanted to partner IBM to create a wealth advisor so that our RMS could then provide investment advice to our clients. Well, unfortunately, you know, NLP at the point in time, we were a bit premature. So it didn't quite work out. So our RMs prefer to Google for answers.
So it was actually, you know, a failure. But guess what? Failure means it's a first attempt in learning. So we didn't stop there. And we were fortunate that the Singapore government launched Smart Nation Roadmap. And, you know, ASTAR, the agency in Singapore that drives, you know, research and transformation, they were very keen to partner DBS to work on AI. And so they gave us a group of data scientists that work on different use cases, including contextual marketing for our businesses, the use of AI in risk management, internal audit, even HR has got that.
So that gave us a very good foundation for DBS. And following which we learned that when we want to look at how to industrialize AI, it's important that we do three things. Number one, we must create a very robust platform for data. So we created this data lake known as ADA, which is named after the first computer programmer, Lady Lovelace ADA. And that took us about seven years to create because you have to clean your data quality.
You have to actually set out clear metadata definitions and the lineage must be good. So there was actually a lot of hard investments and work put in. The second thing that we did was to make sure that we have a set of protocol that we are able to guide our people in deploying AI in our businesses. and that comes with a set of gut rails that we have to put in because AI comes with risk as well.
And finally, which is about the people. So we created a data chapter community. There are 700 data professionals in there, including 200 data scientists. So these 700 people actually are part of my GSTAR organization. So these are the three things we have done to industrialize AI in the bank in the last 10 years with pretty good outcomes, actually.
And before we move on to the outcomes and use cases, I do want to talk about that journey for PayPal as well. What's that journey like? What's that road like for PayPal? Where are things at for PayPal when it comes down to adoption of AI and even quantum?
For PayPal, AI and ML models have been an integral part of our services for more than a decade. We've always leveraged the advanced technologies of both machine learning as well as artificial intelligence to detect fraud patterns at millisecond scale. And if you think of, you know, already today, we have more than 600 models that are making decisions within milliseconds to decide whether a particular transaction, a particular activity should be inspected further, allowed or not. This, when you combine with PayPal scale of 400 million customers with more than 20 million merchants all over the world and continuing to grow.
Today, we process one fourth of all global digital commerce transactions. Wow. That is $6 1 trillion. That is the global scale. We process more than $1.5 1 trillion per year. Combine that with the amount of data that we have, the kind of insights that we can create, that we leverage AI for, for providing those personalized experiences for merchants of no size. That aspect was hard to imagine a few years ago. And that's what PayPal is bringing to all merchants all over the world.
we're available now in 200 plus market to create that opportunity to transform not just the buying experience by providing this personalization but also trying to help businesses grow in a safe and secure manner where fraud loss rates being again best in the industry allows us to leverage continue
to leverage the ai technologies i do want to put this question on use cases and economic outcomes together. If I may start with Him Chuan, right? DBS's industrialization approach, that has helped it deliver over 350 use cases spanning from customer-facing businesses and support functions and also 800 models to date. What is the one use case? That's part one of the question. And part two is what are the economic benefits we're looking at? Because I know the bank is on track to achieve 1 billion SING dollars in revenue uplift by 2025 with the adoption of those technologies.
Tell us more.
Thank you. Yeah, so indeed our AI use cases are pervasive, not just in one part of the bank, but it's across all functions. So the businesses actually use AI, risk management, compliance, audit, so on and so forth. And so I can quote you like two examples, if I may. One is, as you know, we have many customers in the bank, right? And we can't actually deal with them one-on-one because we just don't have enough resources.
And that's when AI comes in in a very powerful way. So today, we have used AI to nudge more than 8 million customers across all locations in Asia. And in Singapore alone, more than 3 million customers actually have been nudged by our models for them to actually save more. So the savings actually has increased by more than 80%. They have also invested more by four times. In fact, they have also bought more insurance coverage for themselves by two times more.
So all these outcomes are quite powerful in terms of driving better customer engagement and behaviours. And so all in, I think as you rightly pointed out today, we have already delivered close to 370 million of outcomes for DBS. This year, we're looking at 700 million. and then next year we are aiming for $1 billion.
Now, if I may go to you, right, Phoram, I know AI, for one thing, has been used to identify and counter fraud and also cut loss rate at PayPal in half during the pandemic, in half. And that is a time when your payment volumes nearly doubled from $712 billion US dollars to $1.25 1 trillion US dollars. So what are some of the use cases of generative AI at PayPal? How far has that triggered down to productivity gains and financial value for PayPal so far?
Sure. Thanks, Tintin. As you mentioned, we've been leveraging AI in different aspects of our experiences as well. So primarily, we're looking at increasing the operational efficiency, improving the customer experience, helping businesses grow. And then overall, making sure that the satisfaction that our customers have when they engage with PayPal constantly is improving. There is a feedback loop, and we are constantly making decisions and learning and improving the models and pushing them as fast as possible. You touched upon some other cases besides fraud.
So if you think about cybersecurity for managing 200-plus markets that we are present in, you can also think that the bad guys are also leveraging AI for a lot of attacks. and how we are able to leverage AI for data augmentation, for adaptive learning, to help our defenses from both threat protection, detection, as well as prevention and education for analysts, for augmenting their reports and making quick decisions in how we are leveraging both AI, but also the data that we have in providing better protection to our customers.
Finally, I'd say one area that I'm incredibly excited that AI is helping us transform, is the personalized payment methods technology. So, for example, all of these customers have multiple cards that they use. For AI to be able to leverage and recommend to a customer which card makes the most appropriate use for the customers to use for that particular transaction, something that we introduced in U.S. that is exciting is called smart receipts that leverages the prior transaction history to show them some ads by that particular merchant to encourage some sort of a buying for the next purchase.
So these type of personalized experiences, research has shown, that increases the repeat and the loyalty by these customers when they engage with these merchants.
And if I can start with Him Chuan this time around, then Forearm, right? An EY analysis suggested that rethinking the financial institution with Gen AI at the core, that has the potential to create 200 to 400 billion USD with productivity gains reaching up to 30 % by 2028. How far are you aligned with this?
Because of the foundation that we've laid with our ADA data lake, our ELN protocol and the people that we have, when Gen AI exploded back in November 2022, we were able to quickly... Actually, a lot of use cases didn't even come from the center. It was bottom-up. A lot of colleagues actually created a lot of demands on how they want to use Gen.Eye in their day-to-day activities. So as a result of that, we managed to then frame all the use cases into two prisms.
So we have the horizontal vertical use cases, and those cases relate to customer center agents, coders, the wealth advisors, operations departments, so on and so forth. Then we also have use cases on a horizontal basis, whereby we have created an enterprise knowledge base for our colleagues to be able to search and summarize information they need to know about the organization and customers. And the other horizontal use case we have created is actually our DBS-GPT, because while the OpenAI one is good, but it is not in terms of so feasible for us to put sensitive information into the public domain.
So the DBS GPT actually provides this avenue for colleagues to put things which are more DBS sensitive into the tools. So I think all in, I think we've come a long way in the last 18 months to basically create more business use cases for our people with Gen AI.
And what about you, Phoram, looking at the road ahead and the potential of generative AI?
Again, incredibly excited about the potential and the power that it can bring to touch the 400 million plus customers that I've mentioned. What we are doing is taking a thoughtful approach to understand the impact that Gen.AI can have on not just our business, but also the competitive landscape. We have enabled all of our internal employees to leverage Gen.AI in finding use cases in which they can provide better efficiency, improve customer experience. we already have our customer service agents leveraging AI to ensure that not only are they using the data, the 200 plus petabytes of data that we have collected over 25 years of existence, to resolve any kind of customer questions effectively and efficiently, but also leveraging cybersecurity,
fraud detection, and personalized experiences that we go.
If I were to get you to summarize the one key takeaway that you want our audience at the Singapore FinTech Festival to take home, what would that be?
Well, I recently come across a quote from Jensen Huang, the NVIDIA CEO, who said that it is not AI that's going to take away our jobs. It's actually people who embrace AI that are going to take away our jobs. So AI is here. So if we don't embrace it, then that's when we might be a bit out of place. So that's the message.
What about you, Faram?
I'll say two takeaways quickly. One is that there is incredible potential for every single company to think, individual to think of how they can leverage the partnerships that exist around the world and finding the right partners who understand their value of privacy, of protection, of security to leverage their business to grow beyond their imagination is one. On the other hand, there are risks that come with these type of technologies, and it's incredibly important for businesses that are implementing AI to invest in also protection in understanding the risks that come with it and ensure that there are appropriate guardrails in place.
And you know, Kim Chan and Phoram have talked in depth about the use cases of AI and also the economic benefits they bring about to both DBS and PayPal. But let's look at the other side of the equation, right? the risks surrounding the AI. The MAS, I believe, came up with a report on the emerging risks and also opportunities of generative AI for banks back in May this year. The risks stretch across seven dimensions. You've got fairness and bias, ethics and impact, accountability and governance, transparency and explainability, legal and regulatory, monitoring and stability, as well as cyber and data security.
It's a bit of a mouthful here, but as a player within the banking space right now, Him Chuan, if I may start with you. What is the number one risk on your mind right now?
So, Tien Tien, I think that to me, AI is very different from other forms of technology that we have seen in our history. It is actually able to do reasoning, very much like a human being. It is also able to create things that in the past, a machine couldn't. right so that means that you know it's in fact with jenny i you know we're going to have like a virtual colleague appearing in our workspace so how do we deal with this new you know person new colleague how do we embrace it uh it's not just about the the outcomes in terms of the values but there are definitely a lot of uh considerations on risk that we take care of
and one thing that i found when i use jenny is that it will try its best to please me right you always try to make me look very good, right? So we have to be careful not to get ourselves deceived by this machine that we are infallible. So I think definitely there is a lot of work to do to make sure that as we progress with this innovation, we are also managing the risk properly, including for example hallucination risk from the user Gen AI.
And going to
you for at PayPal, what is the one key Gen AI risk that sort of keeps you up or wake at night?
One word would be deep fakes. If you look at the technology advancements over the last couple of decades from internet, just the speed at which now data can transfer across the world with 5G, 6G even coming over, social media combined with mobile phones of how people consume this intact, and then the ability of AI to create misinformation campaigns and deep fakes where not just audio or text, but actual videos can be created with very, very little computing power.
The virality that AI enables bad actors to create confusion, misinformation and mislead them into, you know, either fraud or fake scams or whatever it is, is the most, is the biggest risk where we feel there needs to be more that can be done in terms of addressing that challenge.
And please go ahead, Inshun.
Yeah, so, you know, I mean, you talk about production of music, videos and all that. So I think, Tintin, you're aware that I wrote poems. Actually, I also composed some songs in the past when I was in Taiwan. And recently, when I wrote a poem, some friends asked me to then put it into a song. And normally, it would have taken me like six months to complete the whole production with instrumentation and vocals. This time around, I managed to find, you know, a simple app, AI Music app.
I just pasted the lyrics into the app and just selected the style. So I chose rock style. And within like two minutes, the whole thing was completed. Okay. But I was really delighted by that. However, it also means that maybe somebody could be faking me. They could be using my identity to create something, putting my voice in. And then claim to be me kind of like singing the song. And it really sounded like a human being singing.
So I think you're right that as we do all this, let's be very careful on the gut rails that we have to put in place to manage the risk.
So for the next time we come across a video of Him Chuan Singh or becoming a rock star, I think we better think twice, right? Absolutely. Now, are you using AI in terms of your passion projects as well? Or is it more of something that keeps you up awake at night because of work?
Yeah, so AI is incredibly helpful in collecting information and summarizing reports, the incredible amount of threats, incrementally amount of intelligence that comes out and analyzing all of that, creating reports and so on and so forth in terms of just consuming knowledge is where I practically use it.
And you know, Phoram, if I may stay with you, let's talk a bit more about that balance between regulations and innovations and AI because you've talked earlier about how AI guidelines and regulations, they should be proportional to both opportunity and risk. What do you mean by that? If we could get more granularity, any examples that you may be able to share with our audience here?
This is incredibly important. And PayPal as an organization welcomes all the global organizations to come together to address this challenge collaboratively. If you think about how Europe has taken on addressing this challenge with the EU AI Act going in effect on August 1st, and then how US is doing it. And then Singapore, I feel, is taking an incredibly nuanced and business-friendly approach where instead of saying what you shall not do, saying how should you do it the right way.
The projects like Project Moonshot, AI Verify, helping organizations, individuals, equip them with the right knowledge on how to do that, providing the right checks and balances to test those is where I feel we should be going. From a PayPal perspective, what we are working with policymakers is to find a proportional approach. For instance, it would be incredibly foolish for us to reveal to a fraudster why we are blocking a particular transaction. But if a merchant wants to understand why a credit decision wasn't in their favor, absolutely they should know.
So having regulations that are proportional to the risk, having laws that allow organizations to leverage AI without putting unnecessary burden on them is important. And hence a multi-stakeholder approach that takes global standards and best practices into consideration and allows organizations to innovate in an incremental, responsible manner is how I think about it.
If I may go to you, Him Chuan, what is your approach in balancing risks and opportunities? Because I know DBS has established a number of frameworks on responsible data use with a committee comprising senior leaders across different units to provide ambiguous AI use cases and models. Tell us more.
Yeah, indeed. So we take this very seriously at the onset because we are very aware of the downside implications on using AI. and the way we look at it actually are threefold, right? So first of all, we ask ourselves, can we do it? Can we do it? Do we have the data quality, lineage, metadata definition to manage this whole utilization of use of AI, right? So as mentioned, we have created the ADA platform, which is actually the way that we can upscale or industrialize the use of AI.
The second thing that we ask ourselves is, should we do it? So should we do it as in, well, we can do it, but maybe we shouldn't do it because it is not purposeful, it will be surprising to our customers, it's not respectful, we cannot explain why we are using this data on AI. So as a result of which, we created a framework known as PURE, P as in purposeful, U as in unsurprising, R, respectful, and E, explainable.
So unless the AI models that we put through or the data that we want to use are able to comply with all these pure principles, then we will not do that. Then the third thing that we have to consider is that while we can do it, we should do it. How do we do it? How do we do it? Then I think that is really a function of us making sure that the data and AI that we use Is it being fair?
Is it ethical? Are we ensuring that it's actually accountability? And finally, is it transparent? And this is very much the MES fit principles, FEAT, that we also comply with. So these are the three things that we consider very thoroughly before we deploy any AI models into production.
If you're just joining us, We are now in conversation with Lim Him Chuan, Group Head of Strategy, Transformation, Analytics and Research and also Group Executive at DBS as well as Phoram Mehta, Senior Director and Chief Information and Security Officer of International Markets at PayPal. Now if I can pose both Phoram and Him Chuan this question starting with Phoram, how can financial services industry band together to align regulatory frameworks and also maximize the potential of the technology? what are some conversations that are already perhaps taking place right now be it within or
outside of the sff 2024 incredibly important by the way i love pure so i'm definitely gonna look
at how we can always talk to purist because being too purist might also inhibit innovation absolutely
i'll come up with something that rhymes with trust so you know partnerships is incredibly important not just from leveraging best practices as financial services industry together creates experiences that our customers can use, but also to address the risks and the standards and the laws and regulations that we've been talking about. So PayPal's a founding member of Coalition for Secure AI, an open source coalition that has come together globally to provide open source methodologies, frameworks, standards around the world, along with us, Google, IBM, Microsoft, Nvidia, OpenAI, Anthropic, all of them are members.
And this is what we want financial services to also adopt, where we can provide good guidance, good feedback to the policymakers, to the regulators, and showing how we can provide responsible AI experiences to our common customers.
What about Him Chuan?
I guess, you know, center of this whole thing, banking is really about trust, right? And trust is not just about one party. It takes many in the whole ecosystem to create that trust. I'm really glad to see that in Singapore, for example, there are initiatives like the AI Verify Initiative. I think it's by IMDA. There's also a consortium. There's the MindForge Consortium. I think that's led by the MAS that comprises friends from different parts of the industry and coming together to co-create a set of standards that we should all conform to.
Because after all, there's really no trade secret. I mean, risk management, controls, there's no trade secret about it. So it's a very good way to actually share and learn from each other on how to create something that we can sustain. Because having an AI model is one thing, but how do we make sure that it can be done in a very sustainable way, in a safe and sound way? That, I think, is very important.
Okay, let's begin to wrap up the conversation, right? Talking about the adoption of, we talked in depth about AI. Now, what about quantum computing? Phoram, this question mainly for you, but Him Chuan, feel free to jump in. Because even though we've got, you know, McKinsey coming up earlier to say that the market size of quantum computing may reach up to $131 billion by 2040, there appear to be challenges in the adoption of quantum technologies on the ground.
So why should companies care about quantum in the first place? What promises does quantum hold for the future? And why isn't it as talked about as compared to generative AI on the ground?
Again, incredibly excited about another technology, which is quantum computing. If you think about the potential that has already been demonstrated by the quantum computers that exist today, just the you know the future that it'll create whether it is in terms of disease detection drug development predictive models for climate change incredible amount of potential yeah with that though there will always be risk so the certain types of transactions certain types of operations that quantum computer promises to make in seconds or minutes classical computers would take years to do that and one amongst chief amongst them for the backbone of public internet is based on asymmetric cryptography or what is commonly known as public key cryptography which relies on the ability to
factorize prime numbers from large composite numbers so once the quantum computer is able to break the public key cryptography I think there will be huge risk in terms of protecting the data that traverses over the internet particular type of attack that causes a lot of risk is harvest now, decrypt later, which means adversaries are storing encrypted data today in the likelihood that once they get their hands on the quantum computer that is capable of decrypting them, then all of that data will be exposed.
And that's why it's incredibly important for organizations to at least start understanding the risk, start building capabilities that may not necessarily require quantum computer qualified engineers, but understand the risks so that you can build capabilities to change the cryptography that they use, to understand how risky it is so that they can stop collecting the data, and most importantly, rely on partners who have the resources to invest in these type of technologies so that they don't have to reinvent the wheel once these risks start realizing.
I'm also quite aligned in terms of the direction. I think that quantum computing in terms of aspirations are definitely very good. The only thing is that I think at this point in time, we do face some challenges to consider before we could industrialize the use of quantum. So one is, I think, is the maturity of the infrastructure, the maturity of this whole technology. I mean, I read somewhere that it's probably going to take another 10 years for it to be really mature, right?
So that's one thing. Number two is about the infrastructure itself. It's not as easy to have this plug and play compared to traditional form of computing. So like even AI that we use today or Gen AI, it's really not that difficult to just plug in. For the example I gave you on my music production, I just downloaded an app. I just did that, right? But to do this whole quantum computing engineering, I think it requires a lot more investments.
Also, you know, the need for cooling factor. Then finally, it's really about expertise. And I think today, I think it's still very much in the infancy stage. I don't think that we have enough expertise to really drive this thing going forward. But I do agree, I think we have to start somewhere. So maybe this is a nice moment whereby it's still very much a green field and if we don't start looking at it, then maybe we might again miss the boat going forward.
Okay, so very quickly, if I were to get Ram and Hindruan, you two to just give us one quick takeaway for the audience, what would that be?
Incredible innovations require thoughtful implementation that take into consideration the safety, the security, the interoperability, the inclusivity and togetherness or collaborativeness for us to transform lives of millions. And that's what PayPal is doing.
And for me, I think what we say, FinTech, financing technology, we seem to focus a lot on technology. Let's not forget at the end of the day, we are here to serve our people, our customers. individuals right so I still think that it's important to make sure that we are very clear on the jobs to be done for an individual the data level individual data life of a company and technology is really a means rather than an end and unless we are focused on that we might be chasing something for nothing hmm
okay thanks so much for Roman him Chuan that was Lim Him Chuan group head of strategy transformation analytics at research and group executive at DBS as as well as Phoram Mehta, Senior Director, Chief Information and Security Officer of International Markets at PayPal. That was the second of our two-part panel on the Roadmap for AI and Quantum, broadcasted live at the Singapore FinTech Festival 2024. Special thanks to our Senior Producer, Presenter and Assistant Finance Editor, Ryan Huang, and our Field Producer, Nadia Koh.
I'm your host and producer, Chua Tiantian. That's A-W -E-D -I-O. Audio at the App Store and Google Play.