Webinar on AI, Ethics and Compliance (moderated by Pekka Dare, ICA)

okay it's um two o'clock Singapore time so I'm I think gonna start bang on time welcome uh and thank you very much for joining us today for a uh a really interesting topic and and I've been looking forward to this session for a long time actually so today we're going to be focusing on on AI ethics and compliance and how we can we can all navigate the really complex new digital world that we're living in um and I'm delighted uh to be presenting today's session in association with the

Asian Bankers Association uh my name is Becca D I'm the president of the international compliance Association and we're delighted to support the work that the ABA does um in supporting its members around the region and also Delight to be bringing you the session today in ass association with Hawk Ai and Robin who is joining us uh also from Singapore Robin Lee who is the general manager for the asan Pacific region with Haw and and Robin we've known each other for for many years now great to see you

again always good to chat with you and of course you should mention I know you're gonna say a little bit about your background but um you know again uh you know great friend of the A and and author of uh specialist certificate qualification in AI for compliance professions we going talk about later as well so Robin great to see you um I'm going to ask you to to set the scene and give us a bit of a a presentation really for the first sort of 30 35 minutes around this this really important topic uh and then what we'd like to do is is

open up to questions from the audience so please do say hello wherever you are in the world in the chat let us know whereabouts in the world you are um let us know your name and where abouts in the world you are say Robin and I are both in Singapore at the moment um and we'd love to hear your questions so please do put your questions in the Q&A uh and we'll Endeavor to leave at least 15 20 minutes at the end to field any questions you have us your comments your thoughts your questions it's always great to hear from you so look this is

such a huge topic we're seeing a um you know a real seismic movement now in the adoption of of AI machine learning in financial institutions and particularly Banks and when I travel around the world and the region I I see some really interesting developments but with that they're also coming some interesting challenges we're seeing new regulatory Frameworks around ethics and AI um solidifying all around the world and of course we're in an arms race with you

know organized criminals using you know AI particularly generative AI to do nefarious things so there's a lot to talk about today and and we've got someone best place really I think to to guide us through this um slightly you know uh well exciting yet also slightly scaring scary new world and that's Robin so Robin please I'm going to pass the beton over to you um PR want to introduce yourself to the audience and if you want to share your screen I know you've got some some interesting slides

to talk through as well over to you Robert sure and and and and first and foremost thank you so much to the international compliance Association the Asian Bankers Association and yourself peka for for for having me here today it really is an honor and a privilege um let me as as you you suggested let me start by sharing my screen um so one moment let me just pull that up okay um and you can see my screen now um so as mentioned my my name is Robin

Lee I'm the general manager vapac for Hawk and today we will be covering AI ethics and compliance navigating the new digital landscape really really exciting stuff um just in advance uh just wanted to to share with everyone that um I'm I I have attempted to make this as as accessible and exciting as possible um so um just apologies in advance for any pop culture references humor etc etc but

engagement really is the name of the game for these webinars and it's something that I I I constantly find challenging and and in a in a positive way okay so um in terms of of what I'll cover I I'll give I'll give a brief introduction we'll then talk talk about AI in in general we we we'll we'll level set and bring things back to First principles we'll then do the same for ethics then we'll take a look at some of the ethical dilemmas in AI specifically and then what I thought would be nice would be to to to Really home in on the

financial crime space which is where I play in today and where Hawk plays in today with respect to AI uh follow it up with some insights SL predictions and then we we'll we we'll go over to the discussion after okay so from an introduction perspective just a little bit about myself I was born in this country here uh I spent 10 years in this country here uh I've been in Singapore now for about 14 years but my my parents are from this country here so a lot of people wonder why I I look like Jet Lee and sound like

Harry Potter I I I hope that that that really um explains that that fact um okay I started my career in Silicon Valley I was there for six years I then went to Wall Street where for four years I worked in the Hench industry as an analyst covering Technology Solutions so during my time in Silicon Valley my my AI exposure or or what what we we call AI today was was was pretty significant from the perspective that I focused on a company which has now been acquired twice and is now under 3M where we

focused on the electronic medical record and part of my role was was actually developing the the algorithms to access the data to that records to those records um as well as some of designing some of the um applications that that we we we revolve that revolved around that that that data set um yes and then went to Wall Street as I mentioned four years worked in the hedge fund industry as an analyst covering technology companies and then since then for the past 14 years or so I've been I've been involved

in the um finan the the compliance um AML Financial crime space in in a number of of areas I've been in the education space I've done the startup space I've done the space from my time at um what's known today as eleg formerly Thompson Reuters so I was responsible for the strategy for World check uh for those of you are familiar with World Che um within southeast Asia and then I've um I've also done the software space with um a company called Navia um AI um and

most recently on the the the uh public space with um the US Department of Justice and their public private partnership focusing on the fintech space called def Tech sector public dialogue today uh I have a number of roles I've um written a couple of books including one on regtech uh I teach um at these institutions here um I am also the APAC Vice chair for the global coalition to fight Financial crime um as well as sitting on the reg tech

subcommittee for the S uh the Singapore fintech Association and my my my um the job I guess that that pays the bills which is being the GM of Hawk uh for APAC okay so a little bit more about Hawk specifically we're a German software company that's uh roughly six years old um head um and with with APAC being the um the the um sorry with Singapore sorry being the APAC HQ but you can see we're we're internationally

um based and um what is it that we do exactly the way I like to describe us is we are the iOS of financial crime we're a platform um similar to your phone where where we have apps that sit on that platform as well as a big data Lake underpinning that platform and therefore the ability to have lots of of um AI machine learning functionality including what you see here is um is um our one of our core competencies okay so that's a little bit

on the introduction let's talk about AI itself and um the business focus on AI has been intense uh we see some quotes here on the left by Google Facebook BYU um whose CEO is also called Robin Lee but the the Lee is Li um and on the right here you'll see the the um acquisition rate of startups as well and look at look at this on the right here you'll see you know 29 over the course of you know 11 years Apple bought 29 AI

startups Google 15 Microsoft 13 Facebook 12 these are these are not small numbers ladies and gentlemen um and then of course we have the all these developments but are they any good for us and we have this debate that that's going back and forth and on the top you can see where um Elon Musk is is very apprehensive of AI and and where it can go and then on the bottom you'll see how um that um Mark Zuckerberg is is quite enthusiastic um about AI okay but let's

level set here and and ask ourselves what is AI exactly what is it what is it and what does it mean um so I I I I I kind of just did a bit of um Googling around and Wikipedia around and I found an official a computer science and colloquial definition but wrapping it all up together having like the the robin Lee definition per se um I think this one sums it up best right using machines to act and think as humanly as possible okay um

so um the main drivers behind AI uh and and why we're seeing it so much today and why it's being so talked about today and and why the the technology has progressed so much today I put it down to three things okay number one data is the new oil and if we look at the amount of data that's being generated today it's exponentially getting larger and larger and larger over the years all you know whether whether it be the the handheld devices we use or the the um

the the hardware the new hardware um um products that have emerge all of them are collecting data okay whether it's your thermometer whether it's your weighing scales etc etc and so the whether it's your social media applications like Facebook Etc you know the amount of data that that is being collected is is just increasing exponentially and so we need to process that these large amounts of data and on the right here you'll see the advancement on in GPU technology has actually superseded that of CPU

technology so just in layman's terms of GPU use a graphical a graphics Processing Unit um companies like Nvidia and AMD manufacture these um they have been um the advancements that have happened in gpus from both a power saving as well as a processing power perspective has just has just developed um significantly we've seen some really really significant gains uh especially on the cloud side where Cloud servers

are starting to leverage gpus and so as a result of that the ability to to Crunch large and large sorry large amount of numbers has just um has just been become super pre prevalent as of late on the bottom here I guess the last one um that I think is interesting is the um is the um uh evolution of programming languages so stuff like python is really really accessible to everyone now and and and so your ability to to leverage um python which is a a

data science focused programming language um also has has helped the cause for AI okay so AI has actually been around since the 50s and how I would describe it is is um sorry the it was it's been around since the 50s and what's happened over the over um since then as many different techniques have emerged and what will happen is that um you know something will happen sorry a particular technique will emerge it will gain um

companies will get a lot of funding there'll be a lot of um a lot of new companies emerging etc etc and then it will Peak out the technology will Peak out at a certain point and then crash and that's what we call an AI winter and we've had two of those recent um two major ones um since the 50s um but the latest drive that has um really really pushed AI is up here generative a sorry machine learning and and as a consequence of that generative AI as well but if you have a look here there's actually quite a number of different

techniques um involved under that definition of of of AI okay I I have to mention generative AI it's it's it's definitely the new hot thing especially um in our industry uh but if you have a look here I've I've put a definition up um the drivers I've really spoken to um sorry the drivers are software and Hardware so I kind of spoken to the hardware part already but on the software part we have um gpts so

for those of you who um you know are una aware of what the chat GPT stands for the the GPT sorry in chat GPT generative pre-training Transformer and basically it's a machine learning algorithm that can handle large data sets without much effort okay and then the second element is large language models right these are models with billions trillions of parameters um that that that that that can get get worked um because of of the the hardware stuff that we see here

okay let's move on to ethics ladies and gentlemen and um again very very basic definition here from the Oxford dictionary I don't want to spend too much time on ethics I think we just need to have a good idea as to you know what what kind of what the what kind of discussions are involved with ethics and so what I've done here ladies and gentlemen is I've put some some theories I've grabbed this this this infographic which I think really speaks really well to it where you see the different theories Within eics um the sub um the

sub theories that um emerge from that and the the very basic questions that that that you know um articulate those those sub theories so what is my best long-term interest what will bring the greatest overall good for all stakeholders what is my duty to others considering these factors here am I being an exemplary person and am I showing care for um others in my community okay but this is probably the beaten potatoes

of of the presentation ladies and gentlemen the dilemmas okay and the best summary I've seen of ethical dilemmas when it comes to AI specifically so bringing it all together that is are the ones that we see here summarized by the world economic forum and what the Ware has done is they've identified nine different um um topics that really when brought together really summarize that whole um ethical challenge when it comes to deploying Ai and we're going to go

through these in a lot more detail but in a nutshell unemployment inequality Humanity artificial stupidity racist robots security evil Genies Singularity and robot rights so let's let's explore these one by one ladies and gentlemen um very quickly I am conscious of time pckers so don't worry oh this this isn't going to take long okay unemployment what happens after the end of jobs uh and what I've got here is a three pictures okay um the first one is of

what we call a baby Breza so for those of you who aren't aware of what this baby Breza is it's a machine where you put um the pow for for for baby milk you put the powdered milk in you put the water in it heats the water to the perfect temperature for the baby and you just press the button and it pours the perfect bottle of water that's really efficient and and a really good example of of how how um you know efficiencies can be gained with with AI and Technology because arguably keeping that

water to uh that particular temperature that's AI right it's being monitored and and and ensuring that the the water stays within that that that set temperature okay um then it can get to the point where we're super lazy so this is a scene from the movie Wally where in the future humans have have become lazy have the screens in front of them and and become so lazy they they they can't even walk um and so need these device the these these chairs to to um

chaperone them around and then last but by no means these this is this is um an image from the movie feris bu's day off so what I'm trying to say here is why work right if the machines are doing doing all the the day to days and and you know we don't they're doing all the work for us then we can live that life like Ferris buer so this is a movie about a guy who a high school kid who takes the day off school to and does whatever he wants right um so

yes inequality how do we distribute the wealth created by machines and um the the kind of Futures that have been um suggested as a result of this could be are are diverse and and these are the EXT these are extreme examples of that right on the left we have Star Trek where everybody they have actually gotten rid of money in the future and everybody lives for the betterment of humanity and then on the right is an ex this is from a movie called elisium where the world is basically um gone to

One Big Mess it's it's it's desolate and and dystopian and all the rich people the 1% have gone to this donut where any disease can be this donut spaceship sorry where any disease can be cured everybody gets the best clothes the best life etc etc Humanity how do machines affect our behavior and interaction so on the left here um it's a very familiar scene right you go to a restaurant and I go to restaurants quite a bit and and just

observe people and this is quite the the common view right everybody's just on their phone um talking to their phones I've even been in a situation actually where I've seen people sitting in a restaurant and they're talking to each other through their phones they're not even interacting in person um and I guess a really extreme example of that would be the Matrix right and Virtual Worlds and we we we we really saw the EM the significant emergence a couple of years ago on the metaverse right so

that's probably where the the puck is heading in terms of our interaction artificial stupidity how can we guard against mistakes so this is a this is a a poster from a really really old school TV series called Buck Rogers in the 25th Century and this is about a 20th century man who's an astronaut he goes up into space and is Frozen in time and then he gets rescued in the 25th

Century and when he gets rescued um you know it's it's a sci-fi movie it's kind of a take on Star Wars uh where Earth is at war with an alien race and um you know somehow when he lands on sorry when they just shortly after they've rescued him and revived him um they just happen to have an attack from these these alien spaceships and so um they rallied the Earth Defense this guy gets allowed to jump into one of the spaceships to fight

the aliens even though he's just recovered like 10 minutes ago and um they all go off into space and they're fighting the aliens but the spaceships are on autopilot and they end up losing that I mean they seem to be losing pretty badly then Buck Rogers is like forget this switches it to manual and takes out the whole alien Fe Fleet with this one spaceship so um you know depending too much on technology and and and and what have you could could have its um negative consequences um racist robots how do we

eliminate Ai buyas and the most famous example of this is we've seen um credit firms in the US um discriminate against black people and and and and and score them lower when it comes to credit just based off of of their race okay and and because the data that that came before that is is is is biased towards that security how do we keep AI safe from adversaries so this is the fourth Die Hard movie and um for those of you

who are action movie fans and and and and have not seen this basically um a terrorist group takes advantage of everything like the the traffic light system the the um airports etc etc and basically hold hold um hold hold I think it's it's is it Los Angeles or is it the whole of America or something like that to Ransom based off that and and so it takes Bruce Willis to come in and and and and and and sort that out um evil Genies how do we protect

against unintended consequences this is from the planets of the apes movie um and the the new series and in the latest one if you look at how the the um the um uh virus is is is is made um it's actually through computers IE AI right and um and then the result is you know Humanity um um gets taken over by by the Apes Singularity how do we stay in control of a complex intelligent system

so this is from the Terminator movies I think most people have seen this is pretty obvious right that with with Sky inet becoming so intelligent so smart IT renders um humans as Obsolete and so the machines take over the world robot RS this is an interesting one right how do we Define the Humane treatment of AI and um I I don't think this is as as prevalent now but the signs are there ladies and gentlemen and so the the first example I've got here

is of of Facebook right Facebook um you know we um knows who you like they know who you hang out with they know who you're friends with and so they target the ads towards you and Target the content towards you so that you're you're constantly engaged on the on the platform okay um another example of that is what you see in the middle here this is a tamagachi so um for those of you who are are too young to remember this this is a a virtual pet and um and you

have to feed it you have to um uh bathe it clean its poop etc etc right and um when these first came out people were really really emotionally attached to these and and and I remember several instances where W with with with um like some of my cousins for example where their tamagi died and they they they really got you know caught up in that okay so um so there's that aspect there

and then this is the most extreme example for and this is this is from the l337 is actually from the Han Solo Star Wars movie and um in that movie we actually see um a revolution of robots against humans right so um um and um they they and where the robots feel that they were treated like slaves and they they should have have have rights as a a result of that okay bit of an extreme one but I I I I can assure you ladies

and gentlemen we are we are heading there um actually the most recent one I I got a a message today um sharing an article on character AI being sued for um by by some of its users um over depression so they they would have a um they would create a virtual girlfriend um a gentleman in the US created a virtual girlfriend on character AI super super attached and then and then something kicked off and and he's he's

he's he's claiming um um um his depression costs against character AI for for them um engaging him in in in in in such a manner okay so let's let's focus now and and talk about um uh the financial industry itself and specifically AI in financial crime and if we take a look at um Financial crime methodologies today the the the if if I was to take uh fraud and AML and and and look at the

different applications that you see this this is a fair summary right kyc KYB are you who you say you are client screening did you do anything wrong payment screening are you sending money or receiving money from sanctioned entities um customer risk rating is a the ability to risk rate your customers Str Str filing filing your Str strs your suspicious transaction reports fraud and within fraud we have an element of transaction monitoring we have um device aspects like um device IDs and and and

and behavior and be and and and actual behavioral um metrics as well um adverse media um transaction monitoring sifting through um transactions and looking for suspicious activity and customer insights as well okay um the challenge that we see in financial crime is a lot of those applications are old and can't scale out accordingly because of the way they're architected um what you saw um each of

these applications are disperate and function independently of one another the data quality availability accessibility isn't isn't always there we see a lot of false positives uh because of the limitation of of those systems uh Cloud versus on Prem um politics is an aruond newer Frankenstein offerings so a lot of the um the applications that we see out there aren't written by the um aren't written by the vendors and so what they'll do is

they'll take someone else else's um solution Stitch it together and uh sorry a whole bunch of other people's solution Stitch it together and offer what what we call Frankenstein products mindset within the FI themselves and the fact that with financial crime we are up against the human mind the human brain is the best computer on Earth ladies and gentlemen and and and fighting that we are we have still got a long long way to go that being said when it comes to these applications AI is being deployed

all over and So within these applications anything that's OCR is optical character recognition um machine learning is is what ml stands for deep learning uh is what DL stands for um neurolinguistic programming is NLP and gen AI is in green so just very very quickly OCR optical character recognition that's taking a a picture with text in it and identifying what that text is machine learning is going

through large data sets and um um apply like training tasks uh that that you conduct on those data sets over and over and over and getting it better and better and better at doing that that particular task um geni I've already defined uh what that is and deep learning is putting all the data in one place and then asking the algorithm what do you see from this what interesting insights can you draw from this okay so um there are challenges and I've

listed these challenges here and I'm not going to go through them all ladies and gentlemen but you can see the the highlights here right data privacy and and by the way um General AI challenges are written in red and gen AI specific challenges are written in green and we can we can um and if anyone um wants a copy of the these slides please do reach out to me I also have a copy of um an AI procurement checklist that that PE people can get a hold of as well so please do email me if you you would like

a copy of that too but if you look at the highlights here right data privacy and security risks Regulatory Compliance risk bias and fairness concerns model accuracy and reliability remember what I was saying about who's writing the code you might not even know that right U operational risks transparency and explainability cost and resource allocation dependency on technology ethical concerns and skill and knowledge gaps so what this leads to ladies and gentlemen is when we talk about AI ethics it's essentially not too

dissimilar to compliance right a ethics is the precursor to compliance and I would say the same is very true for for AI governance AI risk and AI compliance itself right and this is why we understand the the ethical challenges with AI we get better at at conducting those respective um tasks okay so just a couple of very quick insight sites um in terms of what

we're seeing next on the AI in the AI Financial crime space um I'm seeing a lot more people so we we we've already heard a lot about deep fakes right which is um a form of adversarial AI whereby um the bad guys are creating a deep fake image of somebody and projecting that to the end user so for example I could be talking to you here as Robin Lee but what you see on your screen is Jennifer Lopez and there was a case in Hong Kong uh

earlier this year of someone who thought their CEO was calling them and they wired $25 million doll to um to someone else and um and got scammed on that okay um already spoken to adversarial Ai and and P peka mentioned that too right bad guys using AI um and the best example I can think of that is what's happening with the scam farms in Southeast Asia uh previously um Regular IND indviduals were being kidnapped brought over to

these scam farms and made to conduct scams um the criminals have taken it to the next level promising Stanford computer science phds um um you know high paying jobs in Singapore with a quote unquote stop over in a southeast Asia country they'll get kidnapped their passport will get confiscated and they get made they're you know they get forced to create these adversarial AI algorithms such as the Deep fake one that I mentioned earlier okay I've

already spoken to the metaverse this is an interesting one and this is on the the industry side um I've heard two things two two names for it the first is sraml and the second is SciFi and that is you know we're already seeing fraud and AML coming together we think cyber might be next okay and and so you'll have this department where cyber fraud and AML sit together shared data and data sets um the collocation of so this is straightforward you more more data

sharing definitely helps uh the fight against Financial crime uh collocation of public private Partnerships we're really seeing that a lot in the on the fraud space an example of that being the Singapore scam Command Center and then the return of utilities but this time digitized and again in Singapore we have um Cosmic which I think is a a good example of that now going broader here uh I have three possible outcomes from a us as in

the humanity perspective Ai and what we see globally um so as humans we we we maybe we need to become more Street Smart versus book smart um learning becomes a passion versus a a chore uh work becomes more analytical creative versus busy um convenience SL personalization becomes the new norm and takes priority over privacy there's always a balance there right um especially when we replace this with

customer experience we become less emotionally attached to other humans okay and then um on the AI front AI will become more and more involved in our everyday lives if it isn't already uh anyone will be able to have access to AI algorithms and it will become easier to train them um they are becoming more user friendly and and I can just cite this with programming right when I was programming 24 years ago it really was line to product whereas if you look at some of the programming now it it's pretty much drag and drop right in some

of these these applications so um yes um okay and then um let's see here new interfaces and how we interact with AI will emerge I I think um I can be safe in saying that the touch screen and voice is very quickly replacing the keyboard and the mouse and then um AI will get better at creative stuff and and and gen AI is an example of that right writing songs creating movies etc etc okay um the world will continue to

become smaller um elimination consolidation of language I put this one in because there's this example of of Google The Google Buds where someone is talking in Spanish to someone who's talking in Chinese and the buds is Translating that in real time um time will continue to become more valuable Virtual Worlds becoming more important than the Real Worlds I definitely see that happening and then this one's an interesting one and I I put this in as a more of an afterthought energy food shelter so if Ai and Technology can come

in and and and somehow replace all of these um whether that's solar energy here whether that's example um of um um soilent the company soilent where they make a meal for $1 um a bottle uh for food and and and and and the profits they generate from that they were actually sending over to um to needy um jurisdictions and shelter the best example I could think of shelter is there was there's a company in um Eastern Europe that can 3D print you a

house that's 1,200 square feet three bedrooms uh for ,000 us okay so if if if that becomes a norm then then then shelter shouldn't be a problem either okay and then going back to the um um actual measures that have been taken for AI from an Ethics perspective we have um Mas feat principles here UNESCO has some some core principles that you see here uh Harvard Business Review has put out a

practical guide to building ethical Ai and that's a very interesting one to read ladies and gentlemen um but we've also had a form of of of Regulation SLG guidelines and what I try to do here is pick the 10 most significant ones that I've come across um and as you can see here it all starts in 2016 with gdpr um the AI ethics guidelines from the European commission are prettyy interesting to go through um I already mentioned Singapore's mod wayi governance framework but in Europe This

is the big one right now the AI Act and the the the Talk of the Town in my industry is we're seeing many AI startups actually leave Europe uh for other jurisdictions uh because of this okay um and then a couple more here from the the UK and the the United States okay so that's it from me ladies and gentlemen in terms of what I wanted to present I'm just going to hand it over to to Becca to go through the the next couple of slides because this course that we put together was a a

really really um um um awesome experience and we my my my co-author and I worked really really hard to to to ensure that we provide as awesome an experience when it comes to to what what what you do sorry what you learn when it comes to this course so um so sorry over to you Becca yeah thanks Robin um you know really interesting and and a lot to unpack in in your presentation there um so yeah you know look you know we we're all needing to engage with this topic to

upskill ourselves uh and the IC was was delighted to partner with with Robin and his his colleague and help create the new international compliance specialist certificate and AI for compliance professionals and it it's a really great qualification focused on giving you the right tools to ask the right questions to engage in the the you know with confidence on the on the subject of AI whether that's understanding about the implications for regulatory Technology Solutions whether it's understanding on

our the latest ethical considerations and ethical regulatory Frameworks so when you considering implementing an AOA um solution in your organization how do you practically go about evaluating the ethical and practical and operational considerations as well and of course A lot of it is also thinking about the future because this is such a fast moving area as Robin has alluded to um you know so where are we going with AI over the next few years as well uh and again the course is purely online so you

can study at your own time most people will be able to finish it within four weeks as well but you've got up to two months access and at the end of the course there's a a um an online multiple choice examination as well and if you're successful you're able to get an automatic qualification which of course being with the international compliance ass assciation is recognized internationally and all of our qualifications are awarded with the alliance Manchester business school as well so they're uh you know internationally recognized qualification as well so yeah very proud of this and

it's proven very very popular indeed all over the world so I think we have another slide there Robin thank you um if if you want to click on the QR code if you're interested there um for everyone attending the webinar today you can um if you use the code at the checkout if you're registering for the program that you see on your screen there you receive a 20% discount and if you're a member of the Asian Bankers Association you also get a 30% discount as well so yeah we have a QR code so hopefully yeah look forward to seeing

some of you on on that program and and enrolling in that as well um also worth mentioning um that we've got the uh Asian Bankers Association general meeting and Conference coming up I was um privileged enough to go to the the last event in um uh great event really recommend it some really amazing thought leaders from all around the the region of the world uh and in this one in particular um we're going to be focused

uh on Asian Banks transitioning towards a sustainable future and you're going to have Bankers from all over the region coming in to talk about you know the current areas of focus as well um so this is the 40th one uh 40th time it's been held it's in Tai pay 18th to 19th of November and again I know the uh Asian Bankers Association is is delighted to invite participants in this webinar to visit the uh conference

website um at uh https aa2024 tww and I'll put that link in the in the Q&A as well um for you as well so something to to look forward to as well okay uh Robin I think we've got one more slide have we yes so discussion um let me just put this in the Q&A I know we've had some great questions coming in as well um so look I'm gonna throw a few questions and then let's just open up to

the audience's questions so please um I despite me saying that the chat was available I understand it's disabled so please do pop it in the Q&A um I just want to ask you one question Robin and this is bit Li field maybe but I heard the term P Doom recently uh I don't know if you're familiar with that term Fe Doom so I was listen to a podcast called coming storm really excellent I recommend it but one of the things they were talking about was the view of you know AI experts AI Engineers AI

evangelists about the probability that AI will be an existential threat to humanity and the percentage the higher your P Doom percentage is the higher percentage you believe AI is likely to cause an existential threat to ity and you mentioned the concept of Singularity where we have you know ultimately maybe one totalitarian Authority or entity having too much power and and you know bad consequences so apparently um the

average AI engineer their level of P Doom is 40% which is quite High isn't it then you've got people you mentioned Elon Musk so if I had to push you Robin do you have a a p Doom percentage what on a level of uh of of of of you know existential threat do you what chance do you think it is that the a can p can pose for Humanity I'm I'm I'm going to not Dodge the question but but um answer it from a

a different perspective here and and and that is you know if the P Doom is indeed referring to a Terminator scenario right where the machines rise up and and and um take you know eradicate Humanity I I I I don't have as as as as strong a view on that and and and and I'm not really I I I don't feel anyway that I'm in the in the position to to to give a solid

answer on that but the area where I I'm concerned is the the critical thinking aspect the the the creative aspect right if you know I'm I I'm an artist heart right I was I was a songwriter in the former life I like doing a lot of writing and and and and and what have you and and um I you know if if as a songr if I take simple songwriting right

chat GPT is pretty good and if if it gets to the stage where your your favorite favorite artists um in the music industry are are singing you know AI written songs and they're not coming up with the songs that themselves like like princ did back in the day or David Bowe or anyone like that right then you know I I I you know I'll be very very disappointed and and and and very disillusioned as well um and if I if I take that one step further and and and

apply that to the financial crime space right um you know as as Financial crime Fighters we do need to have that aspect we do need to because we're figuring out criminal Behavior Uh and criminal activity which is you know just as intuitive right and if we leave it 100% to the machines then I'll be I'll be very um a disillusion on that front as well yeah I see there was there's a an open letter published by um Tom York

from radio head and and something like 300 leading artists and musicians um raising exactly those concerns okay but look let's go to some of the questions and please do do pop your questions in the Q&A can I start with this one I it's interesting I don't know if you're familiar with this but uh the question here is there's much talk on how AI could improve our lives but is there quantitive results or measurements or has there been a study on how AI has indeed improved the way we supervise financial institutions how financial institution have served as

customers better Etc so we seeing in particular obviously we're focused on banks in this this session with the ABA Robin but is is there much evidence or analysis of the benefits that we're starting to see in in banking or financial services in general are you are you aware of any you mentioned underwriting and insurance decisions I know where we talk about financial and possible any thoughts okay on so there's there's there's a little bit to unpack there so in terms of using AI to

supervise financial institutions I have not seen anything like that myself but maybe I can I can go back and and have a look at I mean if if anyone's going to have anything like that it would be like bis right or or um those kinds of Institutions there so not not that I'm aware of but um don't quote me on that and then how financial institutions has have served its customers better I mean this whole stuff that I mean that

that's kind there's been stuff that have been carried over from the tech space right the the like automated NPS scores for customer success um nps's net promoter scores um etc etc and I see a lot of the binch companies um apply apply stuff like that um I'm just trying to think I mean from from the financial crime space specifically which is where I sit today I guess the the the the best one is is um like the false positive reductions

when it comes the the effectiveness of false positive reductions with respect to transaction monitoring and screening and there's a number of reports done done on that whether it be from the big four whether it be the Consulting companies or even the vendors ourselves have have some of those numbers there yeah okay okay I mean look I think I think the I'm aware of of big Regulators starting to very much look at um using AI in terms of Border made and we've seen a move in the last years

around the world to um automated regulatory submissions from FIS I know that the FCA in the UK automated a lot of that so obviously having that huge pool of data you know of course regulators at this in terms of how they can supervise more effectively and look for pns oh oh no doubt I mean with soup Tech etc etc but in terms of reports or studies I I I I have not seen anything person yeah yeah I I haven't either I have not but it's a good question thank

you um okay let me go back let's see what else we've got Robin uh for you um there's a question here actually can iI be biased how so um so you mentioned that you know obviously some of the challenges concerns so talk us through what we talk about potential for AI to be biased what where does that bias come from um so if we if if I Al to bring it back to First principles when we talk about the the bulk of quote unquote AI that we talk about today is machine

learning okay and and and and what is machine learning exactly it's it's an algorithm you put a bit of data into the algorithm or sorry you put some data into the algorithm and I'm I'm I'm really breaking this down in layman's terms here right the algorithm processes um that data through a task and then spits out the result and learns a little something along the way then you put another bit of data another set of data in and it does the same thing and gets smarter and spits out a more accurate

result and it will keep doing that over and over so the more data that you put in the smaller the algorithm becomes and and um hence you know what I said earlier in the presentation about data the amount of data being produced exponentially increasing that's how these algorithms become smarter and smarter and smarter is because we have so much more data available so that being said if that data has some sort of of bias attached to it then the result

that gets spit out is is going to incorporate that bias into it hence my example with you know um African-Americans in in in the US getting discriminated against for for their credit rating because in the past that certainly happened but today not not not so much right um the and this is an interesting angle on adversar adversarial AI there are there are bad guys out there doing what we call data poisoning so they'll they'll feed they'll deliberately manipulate the data

that's being inputed into some of these algorithms to to manipulate the the result that gets um gets um created from from those algorithms so so um we we we we we do have challenges from from that perspective as well okay thank you um so there's a question here as well and you mentioned um some of the specific applications we're seeing in Banks um in the financial crime space and and I've certainly seen some really interesting developments there but the question here

is what role does AI play in enhancing the accuracy of transaction monitoring systems for AML compliance and I guess before you answer that we are we seeing kind of the death of or a movement away from rules-based transaction monitoring now and you you've seen you know some of the big world's Banks using using you know trumpeting the the reductions and false positives that they've had from adopting AI Solutions so what's your take on that question what is the role of AI in making us better at transaction

monitoring for AML okay so taking it back to First principles here right what how do we do things today for transaction monitoring it's predominantly through rules right and the the easiest example I can site is any rule over us$10,000 needs to get flagged okay uh we've had a fundamental challenge with rule-based systems not not from the fact that they generate a lot of well on one part yes they have generated a lot of false positives but on the other hand it it's the the um

inflexibility of the systems themselves okay so for me to change that $10,000 to $50,000 with the traditional systems what have I got to do I got to go back to the vendor Say Hey I want to make this change to $50,000 the vendor will will um send someone over to hand program hardcode that that rule change into the system right it could take up to two weeks and then uh once you've done it you might be like you know what 50,000 is not capturing enough now let me bring it down to 40,000 and then you

got to go through all of that again and and each time you change that rule it's a $10,000 cost $10,000 cost right so it gets expensive very quickly and it's a very inefficient way of doing that of tuning that that that that rule systems the newer Solutions out there um are able to for enable the compliance officer to do it right so the the the the the the operator can just sit there and and do it themselves without going to their internal it department or the

vendor and do that and you're able to make that change there and then in addition to that um we have what we call a Sandbox environment where you can you can take either a copy of the full data itself data set itself or a sample of the data and test the rule changes out in that sandbox environment and then when you're finished with with what you have you press a button and then everything gets updated automatically okay so so rules base is traditional it has challenges but you'll be surprised number one how much false positive

reduction you can do when you have control over your own rules and can test them out accordingly okay now let's move to the AI part so this is even before AI right um so so so when when we get to Ai and transaction monitoring this is where things get interesting as well right because we have different angles in which we can approach this we can do an overlay over the existing rules based system and eliminate the false posit sorry reduce the false positives um um using machine learning right there are systems out there that will

um um apply a uh deep learning solution to identify false negatives right which is picking stuff up that you should have picked up but but but didn't um through through some sort of insight or some sort and then there's the there are there are solutions out there which which take it um a a whole angle further where you're not looking at the trans where transaction monitoring is just a part of the puzzle and you're doing customer monitoring overall versus just

a a single piece of the puzzle um now to answer the question on on AI Solutions I mean it's definit we're definitely seeing um a lot of vendors that are AI only right and uh um including you know Google right who who have a um an AI only transaction monitoring solution and that that's been that's been a big big hit amongst um HSBC in particular right and and and and has been approved by regulators

um internationally okay yeah um interesting interesting thought so we we got a lot of questions coming in as well um one question here is um do you see I've lost it now but do effectively the question was do you see uh from John John yo thanks John Robin do you foresee Asia to follow Europe and introducing a similar AI rule so you talked about the AI act and obviously the potential for regulatory Arbitrage

and pushing maybe some vendors out of of Europe do do you think Asia is going to go down the same route Asia is the most fragmented place on the planet I I tell this to my my my management team at Hawk all the time right um how we how we think in Australia is so different to how we think in Japan to Indonesia etc etc so my my my short answer to that is fits and spurts right it'll be it'll be it'll be dependent on on on which part of of

of of Asia we're we're talking about here right yes yeah as you say it's a very desperate I mean we're seeing as you you mentioned some of the initiatives that we're seeing in the region but it's very much jurisdictional by jurisdiction um okay so um let me let me let me just just just um answer that like Malaysia has an AI act coming out at the end of the year for example right whereas in Singapore it's guidelines right now so it's just an example of of

you know the differences and and we're right next door to each other so yeah okay uh Monica thank you Monica Monica Cura has asked how do you see the long-term evolution of AI in terms of ethics and responsibility um I mean it all as the algorithms get smarter and smarter and smarter I think that and we already see this by the way in the in the technology companies right so you go

to Google you go to Salesforce there are people going around with the job title Ai ethicist and I think this is going to hit financial services within the next you know five years easily yeah you have ai ethicist working in in Banks yeah definitely and and I think it's already happening I think the FCA consulted last year on whether there should be a mandated role for for banks and FIS aist um okay thank you uh here's one from Yin

Yin Chang thank you Yin um how will current AI developments transform work you mentioned already that perhaps we we need to we're freeing up the opportunities to focus on more value ad roles more creative thinking or critical thinking roles but the question is how can humans really f f future proof ourselves and make sure we can still be relevant when AI achieves AGI or is it even possible for AI to achieve AGI what sorry AGI do you know that artificial general

intelligence as smart as as us right fully kind of equivalent of not Superior to humans yeah okay so what are your thoughts on how can we future proof our CS so I I I first and foremost think we we we do the stuff that we we really go hard on the the stuff that AI um Isn't So effective on today right which is the soft skills right so the people skills ability to to interact to you know sit down and have dinner with

someone without looking at their phone um we we so that that that's that's one aspect the second is I in an ideal world you would have a a a and I've said this a couple of times now like like a Robocop approach where it's humans and AI working together um rather than a Terminator approach where it's 100% AI so that that that would be the the

second aspect on that um and then what's the other part how make sure we can still be relevant when when AI achiev so yeah so answers that and then is it possible for AI to even achieve AGI um you know what that I I I don't have an intelligent answer to that because we're still on that journey and and and where we've gone with machine learning and and subsequently gen AI we we haven't peaked on that yet right so

do I know if it's going to top out and we'll have another AI winter after not not really but um I I I think that that we we we just future proofing ourselves is is understanding it and understanding where we need to put the the guard rails right and and and ensure that that this thing doesn't get too out of control but there's a linked question to that um by May which is I think a really good question which is what would the role

and responsibilities of compliance departments regarding AI in the banking industry be um so you know what you mentioned you know the role of perhaps Ai ethicists and Banks but when where and what should a compliance function be doing in regards to this whole topic Robin so I think I think in in you know first and foremost see where AI is going to be part of your day-to- dat right so we we already spoke about transaction monitoring there are AI applications that can take a regulation and spit it

out into you know um layman's terms and then subsequently from that spit it out into how it would how you would apply those those um regulations to your your day-to-day operations right and I think I I think in general right that that slide I put up with the the questions that you should be asking really are you know um complement the role of a compliance officer um and simultaneously

if you again I'm not just putting it saying this as a promotional thing but but I I really did sit down and and um and think hard when it came to putting that checklist together um not that not that as compliance officers we are you know tick the Box types but I think it's a good guide as to as to to what what you should be looking out for and the questions that you should be asking when it comes to to um um leveraging AI applications yeah thank you I'm

conscious of time Rob we're we're at the allotted uh finish so we we like to be punctal some great questions apologies we not had a chance to answer some the questions about the use of AI in electronic kyc for example which is a topic I've looked at quite a bit happy if you want to reach out to me generally to have a discussion on that um got some thoughts there uh as well LinkedIn perhaps uh but there we go Robin's put his email up um I know we've had a couple of people asking for your email Robin um and do reach out to Robin directly if you're interested in getting

a hold of um C slides and indeed the checklist that Robin's mentioned as well great so yes again thanks Robins