 Hi, everyone, it's MJ the fellow act tree and in this month's vlog We're gonna be talking about Artificial intelligence and I know what you're thinking Everyone is talking about AI. I mean you go and tick tock. It's chat GPT this chat GPT that it's like haven't we Haven't we spoken enough about this topic and I'm like, no, no, this is such an awesome topic I'm absolutely loving all the videos out there and I thought you know I want to join the discussion and just throw my head in the ring and have a few things to say on such an awesome and Groundbreaking subject. So what we're gonna be doing is looking at, yeah What are my own personal views on AI? How's it gonna be impacting the actuarial profession and job? Basically, I think just those two things will will give us more than enough content to speak about for the next 30 minutes Like I said, this is the vlog video try aim for 30 minutes a longer format Don't have to worry too much about making everything, you know short precise and to the point We can have a little bit of time, you know, just to chat have fun and and think some of these things through So let's maybe start with with y'all. How did I get it involved in AI? And I guess I've always been fascinated with AI sci-fi, you know computers In fact, I should have probably studied computer science instead of actuarial science Just about the amount of time I spent on computers IT was by far my strongest subject at school The parents were very much insistent that I joined a profession and they were like computer science There is no profession attached to it So they they weren't too encouraging on on that front, but interestingly enough my very first I guess you could call it official job because after graduating at the end of 2013 I would spend most of 2014 doing a bunch of small jobs for various different actuarial institutions Around the world of course working remotely It was only towards the end of the year that I actually joined an office corporate kind of job And it was an interesting interview because in the interview in November 2014, you know, that asked me what is what is Bitcoin? I always like to say back then all I could say was internet pirate gold and the people that I would join in this company A lot of them were computer scientists Essentially the company was building back-end systems for insurance companies around Africa And the CEO had the view that it's easier to teach programming to actuaries than actuarial concepts to programmers And what was great about this job is that like I say I got to spend a lot of time with these computer scientists Which very much was my you know my second choice when it came to to what I wanted to study And the interesting thing that we were doing at this at this company is we were creating something called Semantic databases and the big idea behind the semantic database was that you could have a lot more information with a lot less data and this was kind of it's weird because The technology didn't really catch on because data storage never became that much of a problem We were always able to just make bigger and bigger Storage device. I mean, I think this laptop I'm recording on has got like a terabyte of of hard drive space, which is absolutely huge and Semantic compression or whatever it did, you know the database wasn't really needed However, however, it was still very very cool because essentially what we were doing at this company is Instead of just recording all pieces of of data and putting it in a big data You know structured database We could store a limited amount of data in an unstructured format and if we knew the links between different pieces of data We could reconstruct that information when needed. So I'll give you a very very quick example. So you know what I'm talking about Yeah, sometimes what you'll do if you're in a normal database, you'll store the person's ID number You'll then store their data of birth. You'll then store their their age Now what happens is every year you have to run a query and update all the ages You know check when the birth date was and increase it if their birth date has has passed now in South Africa Our IDs the first six digits of our IDs represents our birth date The year the month in the date and I'm not sure I'm not sure if other countries IDs do do the same thing But the idea was that you'd have to store when they were born and what the age of the policy holders were if you had their ID number you simply took the ID number Ran a calculation and you could spit out the age using the current date So you'd say current date minus, you know the first six digits and you could get this person's age Which means you could reduce two fields on the database and they kind of did this not just with personal information It's just a nice easy explanation But when it came to you know, what should the benefit be how much should the premium be, you know All these other things they were able to rather Calculate instead of storing so I was fascinated with this concept of semantics and When I googled semantics I found two very interesting things The one was that apparently they try to make a new internet called the semantic internet and it was a Huge failure which gives me, you know being a big crypto fan. We're like, oh, we're creating a new internet I am aware that we have tried to do this before and we have failed But like I said, I don't want to get too much into the semantic web It was a very interesting project that ultimately ended up failing because people just couldn't I guess agree on what should happen But essentially what they wanted to do was make computers intelligent so that they understood the data that they were using Rather than our current system, which is just you know connect anything together throw the data and then you know have various software Then reading it and and trying to interpret it. They almost wanted to embed that understanding in the infrastructure Anyway, the big thing with semantics is when you start to study semantics you come across a very interesting Individual so this person by the name of Charles Sanders purse now you pronounce his last name purse Even though the way it's spelt you want to say Pierce So sometimes people say Charles Sanders Pierce, but apparently the correct pronunciation was Charles Sanders purse now chances are you haven't heard of this individual the reason why you haven't heard about him is because He was very very intelligent and he's very very intelligent and he was also quite insulting of other Intellectuals so his father was a professor at Harvard So he learned maths and Latin and everything like I think before you could walk note It was a little bit older than that, but you know at a very very early age He would then go on to create a new philosophy for truth known as pragmatism Which says the truth is that which is you know most useful and yet all these cool interesting thought experiments that people couldn't You know unlock unless they adopted pragmatism But and from an actuarial point of view, it's it's fascinating to see the work that he did on statistics Logic hypothesis testing and and all these other things so much so so much so that are actually Seeing that we're talking about chat GPT. I got chap chap GPT up And I said, you know how how has Charles Sanders purse's work? You know impacted the development of artificial intelligence and it said, you know He had an indirect but significant impact on the development of air because Charles Charles was like what? very very early 20th century late 19th century and I mean apparently he wrote a letter to one of his students Outlying how electricity can be used to calculate Boolean functions Which was like 30 years before the first computer came out So it was fascinating, you know how ahead of his time was but he added a lot to to logic reasoning This is coming from chat GPT. It says person's contribution to logic particularly his development of existential graphs and work on deductive Inductive and abductive reasoning have been influential in AI research These forms of reasoning have helped shaped rule-based systems knowledge representations and inference mechanisms in AI Abductive reasoning in particular has been widely used in AI to generate hypothesis and Explanations based on incomplete or uncertain information So as an actually you should be getting quite excited because you know, we use statistics Which is going through, you know data to try and answer uncertain questions. That's what we have hypothesis The one thing I read and I don't know how true this is that apparently he introduced the term likelihood To the whole game and like I said, he did quite a lot on Statistics even did a lot on math with regards to infinity numbers And I know this is completely off topic But he drew the map of the world without Disproportioning any of the continents or any of the sizes and did it in like a square format The only problem is then like New Zealand's in all four different corners of Of the world so not a very practical map, but it has like the North Pole, you know, right in the in the center I always wonder what it would look like if he did it the other way with Antarctica in the middle Like what other countries would then be or constants be cuts up in the corners. Anyway coming back to to Charles It was this other part here where he did semiotics and knowledge representation Piers's work on semiotics had an impact on AI research focus on knowledge representation and natural language understanding He's triad model of science Representation object and interpretation had influenced the development of ontologies and semantic networks Which are used to represent meaning and relationships among concepts in AI You also did symbolic AI and cognitive architecture Okay, Piers's work on science symbols and their relationships had implications for the development of symbiotic AI and cognitive architecture Cognitive architect. That sounds like a really cool word to put through mid-journey and get like some crazy surreal image on that Symbiotic AI relies on the manipulation of symbols and their relationships to represent knowledge Well cognitive architecture aims to model the human's mind structure and processes Purse's ideas on science and symbols have been influential in both of these areas Then like I say he also had this whole idea on on philosophy He introduced quite a lot of things in philosophy, not only pragmatism another thing called tychism Which was this idea that randomness is embedded in reality and it's not just an illusion of our ignorance And he also had this thing called fallibilism, which was this idea that you know the truth can be corrected If new evidence pops up so Crazy crazy guy like I say he was a little bit Gosh, this is something like conspiracy theory. He was like suppressed because he was Can I get the screen back? Yeah It's a little bit like a crazy conspiracy because he was suppressed because he was quite mean to to everybody else I think like y'all he would insult Kepler. Even though Kepler was dead Like he like everybody loved Kepler the guy showed that the orbits are not pure circles But you know eclipses or clip Lloyd things he said, you know if he had followed this Logical approach it would have taken him You know like one week to figure that out the answer instead of instead of years And he was quite like I say aggressive with his with his insults, which Leads me to another thing that chat GPT did which I thought was very interesting I said to GPT out I said what was some of Charles's greatest insults against other intellectuals because I remember reading his book and just being like oh That's a that's a good one Like I should remember that if I'm ever in a in a debate and I want to you know, throw some shade down like yeah I can't remember them exactly but they were like, you know, if you're gonna continue to hold that opinion Then I'm afraid I cannot have a very high You know view of your your intelligence or like I don't know but the way he said it was really cool So I asked chat GPT like what were some of these greatest insults? And it says Charles was known for his strong opinions and occasionally abrasive personalities But this way it gets interesting But it is important to remember that his primary Contributions and it goes on philosophy logic math whatever and it says while there may be instances where he was critical of other Intellectuals focusing on these exchanges might not provide a fair represent representation of his overall Contribution and it goes on to say, you know, like it gives me a few of these things But they were like very much walked it down and it says while pierced did have disagreements and critiques of other intellectuals It is essential to remember that his main contributions where he's got to rework It is these ideas and advancements in various fields that have made the most significant impact and should be the primary focus when discussing his legacy Now what I find interesting about this is Ethics is creeping in to chat GPT chat GPT is telling me You know, I said I just said what are some of the cool insults, you know, like a little fun thing and it's like You should not be focusing on how somebody insults other people You should be focusing on their scholarly work and not their criticisms of others and that I find That I find interesting like I wonder if chat GPT Just came up with that or or if open AI's Researchers have kind of embedded this how would you call it this don't be offensive attitude at all costs Into the thing and that's why it's got these these disclaimers But did you find it interesting how it's telling me, you know, I shouldn't be looking at the insults when discussing his his legacy But like I say big big fan of Charles Sanders purse. He's got a book called chance love and logic really really cool. I think the first chapter is called the fixation of beliefs, which is Amazing, then he's got something called the irritation of doubt and it starts becoming more psychological and Like I said powerful powerful book Yeah, I mean we could make videos just on the content of Charles but coming back to to AI I remember reading up on all of this stuff and it got me really really excited Got me excited for for a couple of reasons The one thing is I was like, okay cool. This is the first Personal this is a bunch of research. That's actually outlining how AI works to me This is these natural language processes This is kind of like the blueprint on how these things kind of get get built It's no longer just a mysterious computer magic that these things are working There is a bunch of logic behind it and I wanted to explore a little bit more So and and I did I did two things The one thing is I made my own little AI so after I left this company I programmed a very basic little AI that plays rock paper scissors But the big thing was that it learned it did two things It learned how to play the game or it learned what what gesture to to show shown on people's past past gestures But it also had a confidence meter So it would if it got the answer right it rewarded itself by being more confident in You know the answer did next so it never had it never gave its answer with pure certainty There was it was always stochastic But the more confident it became the the less uncertain and the more certain it became however the less confident It became the more uncertain and it would always if it got completely You know scared it would revolt back to just being completely random 33.3 percent for each of the gestures But if it became very confident and you know for the next turn if you have played scissors and lost it would know Okay, I'm definitely gonna go rock with such high certainty. So there was a little bit of the AI Because AI even back then was such a such a buzzword. I Submitted it to the South African Journal of Science who they loved it. So they published it in the May 2016 edition But what I also wanted to do was explore AI it was like, okay, cool I made my little AI my little pet thing That could you know think but let's let's go deeper into AI because I could see AI Assisting on two other fronts. The one front was the actuarial front Actuaries when we're building models, you know, we need assumptions We need parameters and sometimes you don't know what those parameters should be now You can use AI to help you determine what those parameters should be This is sometimes when you're training it on a data set sometimes known as as machine learning Which like I said starts getting me into the other things I'm like, okay if we can use AI to make better actuarial models and it's this thing called, you know machine learning Learning I was still learning for my my fellowship exam Which I had failed twice I'd failed this this fellowship exam twice because that was basically They could ask you anything on finance and it was very new on sand. There was like a ton to get through So I thought to myself hold on If this is the logic and these are the the instructions or this is the blueprint on how a machine will go about learning Why don't I? Reverse engineer that entire process and create a study method from it and It's interesting because I made a YouTube video that like was like the dawn of it It was just like the core idea which it was, you know, I think it was Learn less understand more or study less but learn more I don't know it was some kind of thing where it was like spend less time but get actually more extracted from from the activity and The way it went about doing that was saying well, okay, how do you learn like? How do you actually learn if you're very conscious of your learning process? What are the steps that your mind's taking so, you know, you pick up the book you open the book you read You know, this is the input then there has to be something that's happening in the brain You know, you're you're basically seeing markings Okay, and your eyes are interpreting these markings as as letters It's been grouping these letters together and seeing that these letters are forming words now These words are representing Concepts and the sentence is showing how these concepts are connected to each other And once you've read that sentence that thing gets, you know added to your additional knowledge base It gets arranged in into your mind and then when it comes to you answering a question You sometimes take another piece, you know, whatever the question is you link and say oh, yes It was based on this information here with a little bit of processing and my knowledge base and experience I can now write out the answer and we start going through Like I say consciously on what exactly that you're doing that you're learning you create this You find this this thing knowing that Relationships are key when it comes to studying how are different concepts related to each other? And it was this idea of a semantic mind map So mind map which is it was almost there it was almost there mind map You know you draw the little subject in the middle So you say actually in the middle then you draw up and you say, you know You do statistics then you draw up over there and you say insurance, but you just draw and you just link them Semantic maps was name those relationships. Okay, so actually Studies statistics actually works in insurance. So you can see in the mind map It just said both statistics and insurance were associated to to actually but with semantics You say how it is related and if you say actually studies statistics And actually works in insurance just because of that relationship you can now deduce well The fact that I am studying something that something must be a subject The fact that I work in this other something that other something must be an industry And this is where semantics gets quite powerful. So you've got the links But there's also some more general links So there's this thing known as the hypernome and a hypernome is a higher concept So insurance the hypernome of insurance would be an industry Statistics the hypernome of statistics would be subject and what's interesting is these concepts can now Inherent inherit a lot of things that are coming down from their their hypernome. So the example with the subject I know that I can study a subject. I know a subject contains knowledge I know a subject, you know, it can be be used to solve practical problems You know, there's all of these things that I know just from a subject that The term statistics can now inherit and this is powerful from an AI's point of view Because what it can now do is if it figure us the hypernome of something It can then inherit a whole chunk of knowledge for this topic Even if it knows very little about it and this became powerful for me when it came to writing the actual exams If something called a credit default swap popped up in the exam paper Let's say I'm nothing about a credit default swap. I could say well, well, what is it's hypernome? Oh, it's a financial instrument Oh, it's a financial instrument that means that there are probably a couple of parties that you know This financial instrument is connecting because it's a financial instrument. There's some sort of probably, you know value transfer It's a financial instrument is probably going to be Regulated it probably can be price which means there's valuation methods and suddenly you start saying that even though, you know Nothing about they say a credit default swap Just if you know that it is a financial instrument You can start knowing a whole bunch of things now you're in an actual exam They say discuss or describe a credit default swap for eight marks and you don't know what it is But you just know what a financial instrument is and you can say our credit default swap will have parties It'll have this will have this will have that you might not get full marks because you might not you know Realize the underlying thing of exactly what it is doing, but the exam will be like oh, but you know, he's mentioning this He's mentioning that yeah, that that's those things are all relevant to create default swaps And you will get sufficient marks that might actually see you know You pass the fellowship in which case, you know after doing the study method. I was able to pass it I Should remember I went and I did I did a Ted a TEDx talk at at UCT and I was so excited for it I put so much effort into it because of course these TEDx talks were gonna be recorded They were gonna be released to the world and I thought oh, this is gonna be such a great thing to to share with everybody because With semantic relationships There's the hypernum. There's the hypnonym. There's the merinum. There's the hollownam I mean, there's a whole bunch of these different relationships each which are able to extract so much more information and I did this whole TEDx talk they filmed it and then they messed up the sound they messed up the sound so and Then I don't know they had students who are organizing it and then they never edited it I got the raw footage and it's been something to do for a long time has been to you know Try and salvage something of this TEDx talk and and put it out there because this is 2016, you know looking at You know, how how does AI actually work? What was interesting? I remember in in the TEDx talk was I Was I was going, you know researching into AI I came across another Michael Jordan who was publishing things on on AI and He was at Berkeley University in in America. So his name was Michael eye Jordan I thought they were so cool because you know He's studying AI and what better way to differentiate himself from you know, the boss school player then to be Michael I Jordan and And I can't do that because my middle name is Bayman and there's another Michael B. Jordan, you know in in the movies But remember coming across this Michael I Jordan guy and being like no way That's so crazy, you know somebody else with my with my name and then during covert I had ESPN reach out to me to do just you know talk to me because of my name being Michael Jordan And they got me and 16 other people around the world who had this name and we were just discussing You know weird stories This is what we were doing during covert because there was no sports being played and ESPN needed content And one of the other people on this call was Michael I Jordan This old gentleman from from Berkeley and he was telling some stories about how I think he went to Japan and this hotel was So excited for him only to realize that he wasn't the you know the famous basketball player um There was a complete sidetrack, but coming back to to this whole thing with with AI What I find absolutely fascinating is that it is grounded in Logic, you know there there is logic, which I want to encourage especially if you're an actually watching this to go and Explore it because you'll really really appreciate it studying hypothesis testing and you know induction at all of those kind of things I think you'll really be like wow. This is taking steps Which was such a dry subject at university and putting it at the forefront of you know this emerging technology And it's one of these things where I think actuaries look. I don't think our statistics is Is good enough to be you know to contribute to the field But I think we have just enough. I think this is the nice thing about actuaries We have enough knowledge on statistics And enough knowledge on on ethics and enough knowledge on on finance that we can actually Contribute to the discussion, you know, there's this whole thing now with musk and a whole bunch of people Who want to shut down AI training for the next six months, which I personally am very much opposed I think AI is absolutely amazing. I think it's great I think we should put even more of our minds more of our attentions hence why I'm wanting to kick-start this Conversation within the actuarial community For those of you following me on LinkedIn, you see I'm posting all of these like surreal art images created by by mid-journey of Actuaries doing, you know, actually doing what we do best calculating the time value of money matching assets to liabilities I've got another one of coming up like pensions and surplus. It's got all these crazy little images So find me on LinkedIn to to see these images. They're pretty cool. But yes, I'm very much of the opinion that actuaries need to get embedded in this conversation. I think we We can add we're not like I said, we're not the global experts when we shouldn't go there and dominate the conversation We shouldn't be the only ones in the room But I think there should be seats in the room when discussing AI the ethics and all those other Implications I think actuaries have got a lot of value to to add to those discussions. I mean Will AI replace actuaries? I mean We've had calculators. We've had our statistical software, you know Model building is a lot easier calculating and all of these things have been drastically, you know We don't do personally anymore because we rely on technology to do it And I think the same with AI the AI is going to come in and it's going to be a more powerful tool that is going to allow actuaries to be even more efficient and I think once actuaries are more efficient Then we can start adding more value to, you know, other areas of society It won't just be insurance companies that can afford us because we can do a lot of work in a much shorter amount of time so we can start offering our services to to other industries and I think our knowledge of risk and Uncertainty and financial consequences can play a role in all forms of businesses So if AI can make us, you know More affordable because we know we're quicker in less time and actuaries tend to charge on time I think it's going to be a very very powerful thing for for our profession Are they going to be jobs impacted? Yes, you know And that is something that we need to to consider from a societal point of view and that's why like I say We shouldn't be the only ones in the room. We should have, you know sociologists in there We should have unions in there, you know, we should have people That are representing, you know, other other areas in that whole conversation and you know, let us all be in there Let's all discuss it, but I definitely think actually should be part of that AI Discussion, like I said, we're probably going to be on the optimistic pro AI side of things But like I said, very very excited with this technology. We have hit 29 minutes So this is the last 50 seconds. I guess this is where I'm going to end off by just saying that I am playing with AI So I mentioned in the last post, you know, what's my little secret project secret project is using AI to to generate content Religious content because I think this is where the AI is very powerful the more data It has the better its output is going to be and with religion. It's had got what? 6000 years of content, you know of writings. So I think AI will probably perform best when it comes to topics like spirituality, you know, religion worshiping You know churches and those kind of things and I think those are areas where you know, the logic and reasoning and rationale Will be greatly appreciated, especially given, you know, the modern context and times that we find ourselves in But that that said we have hit 30 seconds. Sorry 30 seconds 30 minutes So we are going to conclude the vlog. Thank you so much for watching Let me know your thoughts in the comment section below and I will see you next month I don't know if this is the a the actuarial vlog or the philosophy vlog or maybe this month We're just going to combine the both but I'll see you next month for for another one. Keep well everyone. Cheers