Creating confidence in finance: How AI is changing the landscape for your finances | John Colthart

Creating confidence in finance: How AI is changing the landscape for your finances | John Colthart

Show Video

Hello, everyone. Welcome to another edition of Curiosity on Stage. This presentation is part of a series where we discuss new and emerging technologies and how they are affecting us as a people in Canada and also worldwide. My name is Michelle Mekarski and I am the science advisor at the Canada Science and Technology Museum. For those of you attending with visual impairments, I'm a woman with shoulder-length brown hair and brown eyes.

And I am joining you this evening from my home office in the city of Ottawa, which is built on Unceded Algonquin Anishinaabe territory. Before we begin, I would like to take a quick moment to thank the National Research Council of Canada for their support in making this series more accessible through translations, captioning and transcriptions. So, Curiosity on Stage. Our goal here is to inspire thought.

We rely on the insights of experts to get us thinking about topics in science and technology that have the potential to really shake things up and really fundamentally change our experience as humans. Certain technologies actually have the potential to revolutionize the very structure and nature of our society, to transform our industry, our culture, the economy, and even our philosophies. If we take, for example, the agricultural revolution. It was driven by sciences and technologies like animal husbandry, irrigation, and the plow. The resulting food surpluses allowed our populations to grow into cities and then into states. And the fact that not everybody needed to worry 100% of the time about what food they were going to eat allowed certain individuals to specialize in things like politics, handicrafts, or art, which created the basis for our modern economy.

If we fast forward to the Industrial Revolution, it was driven by machines like steam engines, which provided sources of power other than humans or animals. These new sources of power made industries more efficient and therefore their goods cheaper. Populations rose dramatically again, and those populations moved to cities, which urbanized our society. Now, the information revolution also known as the age of the Internet.

Here we have things like computers and TVs and mobile phones which demonstrate the advances in electronics, computing, and communications technology that define this revolution. As these integrated systems of technology spread through society and take root, information, innovations, and ideas diffuse far and wide. Fundamentally changing once again, our culture, economy, politics, and our personal philosophies on life. Today, it seems like we're in another technological revolution, an artificial intelligence revolution. In the industrial revolution, machines were able to replace much of the physical work being done by humans. Now we're seeing AI, we're seeing with A.I.

this ability of computers to take on the cognitive work of humans, things that at least historically required human intelligence to do. So as you'll see later in this presentation, AI is an extremely powerful tool. And as a result, it's spreading into every corner of industry, economy, and society. Now, what makes A.I. so useful is it's very good at finding patterns in very large sets of data.

Think satellite images of the entire planet, your DNA, or the world's financial records. Now, financial professionals spend a lot of their valuable time in low cognitive tasks, like sifting through a whole bunch of financial transactions. Now, wouldn't it be great if there was an AI system able to rigorously audit financial data and pick out the key areas that human professionals should investigate further? Well, today I am delighted to welcome John Colthart of MindBridge AI, a company developed to do just that. John has had a whole...has held a series of roles with increasing responsibility at Mindbridge and currently serves as their senior vice president of Strategic Insights and Marketing.

Before joining Mindbridge, John held leadership positions at IBM in brand management, product experience, and design, and he was a member of the team that launched IBM Watson Analytics. Before IBM, John was VP of Sales Operations for Clarity Systems, which was later acquired by IBM. So I know I can't hear you, but I hope you're all clapping with me as we welcome John today to Curiosity on Stage. Cool.

Well, thank you very much for having me. I really do think that it's an interesting time to be in the world when we start thinking of where and how our finances our, our whole ecosystem goes as it relates to artificial intelligence. In fact, most of you probably already use artificial intelligence every single day. The idea and the concept of picking up a smartphone and asking the question of where you want to go, where you want to eat, how to get something.

That's all based in the same logic that artificial intelligence was created. I'm really excited as well because I'm just down the road from the, the Science and Tech Museum. I get to go there with my kids fairly often here in Ottawa or as often as we were able to before. Now, obviously, with our annual membership, hopefully we'll be able to go again.

And we we live fairly close to the aviation museum, which is really exciting for them. So thank you, very much to the team over Ingenium for all their dedication to, to learning and things like that for for kids of all ages. I still consider myself a kid. I really do want to get to a point where you can ask me lots of questions, but there's some things that we need to do to get there. First, we need to talk about what is artificial intelligence, how it's changing everything, and where we go. The world is very, very different today than when I first started in industry back in the late nineties, early 2000s.

And it's really accelerated at a clip that I don't think anyone could possibly have seen or experienced. Right? I don't think we had full understanding of where things would go. And when you look at there's laws like Moore's Law, which is all about the the ability for CPU or computer processing unit size to shrink but double in power every 18 months.

It was like the leading indicator for computer scientists and geeks like myself in the late nineties to try to figure out how small will these things go. Well, it's gone so small. And I'll show you a representation in a minute. It's gone so small that we can now process more information than we ever have humanly thought possible with computing. I think we all maybe assumed we'd get there, but we're doing it now and we're doing it at speed and scale.

When I start thinking about artificial intelligence and having been in the space for for almost a decade now, I really do liken the transformation to be very much like the Industrial Revolution probably was. It wasn't about, you know, getting more horses. It wasn't about getting more more steam into the, into to the rooms. It wasn't, it was getting all of these pieces and componentry to work together to automate a variety of things on a production line or, you know, to make the products bigger and faster. Artificial intelligence is that, but sort of on a, on an explosive scale that's greater than anything we thought. And Jensen Huang, who's the the co-founder of NVIDIA, where a lot of the processing from a graphical processing unit.

I'll throw in a couple of techie things for those that are really curious. A company called Nvidia, which processes, which creates most of the processing units for graphics cards that are used all over the world and mostly in A.I.. He really agrees with the same statement that it is going to be a national imperative.

Canada is very, very lucky to have had a significant involvement from the Canadian government and from the provincial governments. We've got three centers of excellence for artificial intelligence across the world, or across the nation rather. One in Montreal, one in Toronto, and one in Edmonton. But we have an ecosystem of startups that is expanding past, I think, 550 now different startups across Canada delivering artificial intelligence. So you may work in it, you may see it, you may have it, you may be part of that. So we're going to do a bit of rapid fire for the next 20 minutes or so, keeping myself on pace to have us into a question and answer round about half, half past the this, this, this webinar series.

And we'd love to have the dialogue as much questions as you want. Feel free to start putting them in and getting your questions in and we will answer them. For those that may be looking at this as a replay, hopefully you'll find my contact information and you'll have Michelle's, I'm sure.

And you can email us, call us. We're happy to chat about this. But here's the basic agenda. I'm going to give you more basics about what artificial intelligence is. I'm going to talk about the human problem or whether there is a human problem.

I'm going to talk about A.I. in our world today and some of the impacts and then we're going to get into that Q&A. So how can I help you understand further? So if you if you think about it, we want to give you some of the basics. We want to give you some of the things that we see and that I personally would believe are are going to be contributing factors to your ability to embrace A.I.

And then how is it actually affecting you? So the basics. What is and why are we talking about artificial intelligence? Well, a lot of people are quite surprised to know that artificial intelligence is actually almost a 70 year old concept. John McCarthy first coined the term at a Dartmouth College, Dartmouth University symposium that was bringing together mathematicians, statisticians, mechanical engineers and the like. And they came together and said, We've got to really start thinking about how we can automate things further. I imagine this is on the backs of the Industrial Revolution. It's a bunch of think tank, you know, members coming together, and they started talking about artificial intelligence.

They didn't really know what it was going to be, but they started to think about how to get there. And the first general purpose mobile robot was actually developed and deployed: Shakey. He was developed and deployed and... not sure why it has a male connotation.

But in 1969 so 14, almost 14 years later, through that same time there was actually a lot going on in artificial intelligence. The U.S. Department of Defense put together a program to translate English to Russian, Russian to English.

As you can imagine during this time was the Cold War. And so the two, if you will, the two superpowers were having a challenge. And one of the biggest challenges was how could they communicate and converse effectively without having a cast of thousands or hundreds of people being part of the information chain.

They wanted to have very much bidirectional conversation with each other. And so the Department of Defense put together a multimillion dollar project. They started translating English to Russian, Russian to English. This was an early concept of artificial intelligence, and it worked fairly well, except for the fact that it didn't understand colloquialisms or unique things around a given, you know, part of a person's dialect. So it would take things, you know, like "out of sight and out of mind" into something into Russian and then back into basically "blind idiot". So concepts like that didn't quite work as well.

And so we had all these stops and starts. In '97, for those of you that have been around as long as I have, might remember this: Deep Blue from IBM started playing chess earlier in the years preceding this, and there was finally a chess champion that was beat by a computer. 2002, we got our first robotic vacuum. I think most people will attest that don't like house chores like I do, robotic vacuums sounds really great. But, they were really basic. This is 20 years ago folks, right? Very basic.

Then we went through another couple of fits and starts, and starts and fits, and stops and starts. And we got to a point where we got the internet, we got wide scale adoption of search language and search engines. And we ended up to today where we've got Siri or Alexa or Google home, you know, sitting on our desk just ready to do something for us. And that's a pretty big 70 year journey. But when you really look at the advancements and you think about what happened, it's really the last almost, almost 15 years that have had the most impact to the world. And part of that is because of this.

I mentioned computational power. The amount of data that we're trying to process is quite significant when it comes to artificial intelligence, as Michelle said in her lead-in. Right? The ability for a financial professional to look at every single trade that's going on in a business, and every type of transaction that's going on in a business, and make sure that it's accurate and it's correct, and the right funds are going to the right people, to the right buyers, or getting from the right suppliers or customers, rather. It can be a nightmare.

And you can imagine that if you look at the very large machinery, 250 megabits of of storage was about 550 lbs. So that's more than double me, okay. And I'm about 5'10" and a little bit of a stocky build.

And it cost over $10,000 to deploy that much storage. You look at today and you can buy a 256 GB, you know, micro SD card, it's under two grams in weight and some sometimes you get them for less than $30. So you can imagine this computational power is a big piece of how we got here.

Right? Now on some of the basics, I don't want to go too technical because I think that it's the concepts that will make sense, but there's actually a microcosm or a set of envelopes that actual AI systems fit into. So an overarching artificial intelligence platform will have everything from natural language processing, statistical modeling, it will likely have machine learning algorithms, it will likely have some basic rules and and very much scripted things going into it. And then you get into this very specialized area. This this darker blue of deep learning.

And deep learning is really only the last ten years or thereabouts. And this is what really drove such a radical expansion of investment in AI because we were able to do things that were so unbelievable ten, even ten years prior, based on this confluence of computational power, smarts and engineering and the ability for us to develop new languages to do this. Now, what do I mean by this? And why is this interesting? So around the time I'm talking about, there's a gentleman named Demis Hassabis. Demi is co-founder of DeepMind. DeepMind is an organization which also has some Canadian roots.

Dr. Jeffrey Hinton, who's down at the University of Toronto, is one of their members, one of the founding members of the team as well. And they did something extraordinary. They took that little niche, deep learning neural network area and created something that had never been done before. It was called the Q-Learner. Now, let me explain this.

For those of you that are as aged, and hopefully as well as I am as well, who had an Atari 2600 back in the eighties to play video games. They took a video game, they basically put it on an emulator and they taught a piece of software how to play a variety of games. They then converted that into its own level of coding.

Think about, and this is why that picture of the brain with all synapses is there, think about all the decisions you make when you're playing a game, whatever that game might be. This is, I believe, Space Invaders up there on the screen. You know, you used to go across and you'd hit the red button to to explode what, whatever was in front of you.

They converted all of those mechanics that you would normally think of doing into computer code. And essentially got to the point where the system itself was able to take any game from the Atari 2600 and actually play it better and better and better than any human champion in less and less time. It started out with about 18 months, sorry, 18 hours to to successfully learn enough without a human intervention. Learn how to play the game without human intervention. It took 18 hours to win Space Invaders. Then they went on and on and on and it got to the point where it was working at a speed of about every 2 hours it could learn a new game and be extensively better than anything else that had happened.

So to me that's kind of crazy and wild and wacky and insane. Now Google, a couple, a company most of us know, or its parent company, now Alphabet, did something interesting with DeepMind. They bought it. They bought it for a ridiculous sum of money, and they turned it into Google's cat detector. You might be saying, What are you talking about? So the Google cat detector, it's it's a very funny story. They took this idea of the neural net and they pointed it at one of their assets, which is YouTube, and all of their cache of of search, searched websites.

And it went through the process to detect cats, dogs, other animals. And they essentially took that same code and made it better and better. And you can see this in your day to day life, right? When you go to a search platform like a Google or you're maybe using a streaming service like a Netflix, when you start typing in information, when you start asking for something, the responsiveness and the quality of the, of what you're getting back is significantly high. So Google bought DeepMind.

They created the Google cat detector, but as a proof of concept to show that we have really taken things to a new level. So we are now in this new A.I. Spring. It's been almost a decade that we've been in this spring or this resurgence.

And I love the Forbes quote from when this all started. Artificial intelligence is the broader concept of machines being able to carry out tasks in a way that we would consider smart. It it gives us an umbrella way to start thinking about it. But it is quite complex and it is quite, you know, integral into having to get to these levels of, of information. So I'm going to fast forward a little bit and speed up a little bit on to some of those concepts.

So why A.I.? Because it does it performs complex and laborious tasks. It doesn't need to sleep.

It doesn't have, traditionally it won't have the same level of bias. There's a whole bunch of reasons why we can pass huge amounts of data through it and provide the agility to act on the other side of it. And it basically takes all of this complex data, all this voluminous data, and it processes it in spite, speeds that you couldn't put enough human beings on, and comes out with very interesting insights extracted very, very quickly. So why A.I.? Because we're at a point where the computational processing power is available.

We have elements like storage costs going down further and further and further. We've got cloud computing. We've got all these areas. And we've got the Google cat detector, which has obviously broken a whole bunch of barriers into how do we create something that will do good things and identify the things that us humans want out of this complex data. So I'm going to transition into the human challenges with A.I..

And sometimes people don't like talking about this. And the human challenges are, some are based on our own DNA and our makeup and some of it is based on just the technology itself. So I'm actually going to go in reverse order here. I'm going to start with privacy. I'm going to talk about privacy.

I'm going to talk about bias. I'm going to talk about ethics. But we'll start with the black box of A.I. One of the biggest human challenges that we have with with A.I. is that everyone is nervous

about what that decision process looks like. And it doesn't matter whether it's in, you know, industries like mine, like the one that I work in, which is with financial institutions, with public accounting and financial recordkeeping, or whether it's in that business to consumer bot that's helping you pick your next cell phone or cell phone plan, or whether it's in, you know, self-driving cars. Everyone is very worried about what is this box actually doing and how do I get comfortable understanding what it's doing? And so we do try to make sure that the human side, right, that people side can get access to understand and be able to trust it. And that's one of the biggest challenges that we have with A.I. It is not a human problem. It is a problem that the the A.I. system vendors, the A.I.

solution specialists need to continue to to break apart. The second piece is bias. And what's really interesting is the last 20 years of of developing and deploying and designing software for a variety of organizations around the world.

You see human bias everywhere. When there's some analytics tools that people have seen. You know, just think of this as as, you know, charts and graphs out of the data. In enterprises it's very common that you will insert your bias into the question you're asking. I want to know how much our revenue grew. Period over period.

I'm going to go and find all the places that we grew. I may ignore all the places that we didn't because I'm already instituting a bias that I want us to grow. Therefore, I want to validate that hypothesis that we're growing. Okay, interesting.

Humans, as a, as individuals, we have a significant amount of bias. And I don't want to belabor this, and I'm not trying to make this in any way a statement of politics or policies or anything like that, but we have bias, and the bias doesn't come from the A.I. itself. It comes from the people implementing the A.I., or designing A.I. So there's a lot of time spent around how do we do this, and how do we do this ethically, and how do we work through this. And I'll give you an example of the challenges we have as it relates to our bias and then how it blends in with the ethical conversation.

Now, you can probably assume, based on this picture, what I'm about to talk about. So we have a self-driving car in Ottawa. We've got this wonderful program out in the Kanata Research Park that is going on to have self-driving vehicles.

There's a few other programs across the nation that have been deployed. I think Ottawa might have been the first city that that actually actively deployed something, and I think Toronto and a few others have followed suit. But we've got this issue of ethics with with AI and with self-driving cars or the neural network that's working behind it. So as you can imagine, we've got a car coming down the road and we may not have enough room for that car to go past things. In this case, I'll put people. Again, not trying to be biasing anyone in their thought process,

but just giving you the sense there's something in the way there's an A and a B. When we look at the biases and the, the way that people interpret the ethics of doing something, there's actually, this is a fairly significant study that has been done on this case where different parts of the world have different desires in terms of what they spare or what they protect. And so when we start thinking about building A.I. and when we start thinking about the

ethics, are we thinking about all these types of environments that we have to work in? When we look at these biases and these ethics, it can be based on age you know, it can be based on on on, you know, a level of gender. Right? Obviously, there's there's there's folks in the middle that are you know, that are nonbinary and identify differently. It doesn't matter, that, it's the the DNA based structure that we're talking about here. But there's different opinions, as you can see by the screen.

And finally, just even in terms of our level of education. Right? Who who would you know, sort of spare based on whether they've got one degree, two degrees, five degrees, no degrees. Right? And again, very different and differently applied. So when you bring all of these things together, you have this issue of, you know, starting with the black box, what did it actually do? What level of bias is in there and what type of ethical concerns have been driven into this? And you can imagine AI in things like medicine, right? We want to get some of those things right.

And so it comes to our last point about the human challenge which is the privacy of all of this. It takes a significant amount of data to run amazing A.I. systems. And so therefore, what level of data are we willing to use to train our systems, and to create inferences, and to ensure that there's a lack of leakage in the overall system? Right? How do we do this? Well, today, every time you sign up for a new service, right, you get a little terms and conditions. And I'm sure that many of us don't read all of them all the way. And that's okay.

But, you know, at the end of the day, we are selling a bit of our self into these ecosystems of data that we want to have. This personally identified information or PII, right, then goes into these programs. And I'm not picking on a single vendor here. It's just they're going into the programs, they're going into the programs that you're using. And those can, if not driven ethically, if not including bias, and when there's explainablity, you can get a very good sense of how that personal information and how that privacy is affecting or the outcome.

But at the end of the day, we had an issue a few years ago with a company called Cambridge Analytica who has been deemed to have influenced outcomes in certain political spectrums because of the amount of personal identifiable information or PII that they were able to use and leverage and build into their bots that were communicating in information flow through the media system. So obviously, by having that information and that level of depth, it created an ability for them to be very targeted. And people feel that that is uncomfortable. I get it. So let's talk about it in your world.

We set the stage, there's a bit of a basis. There's some really cool tech stuff that we're doing based on really amazing advances in the actual technology. But how is it affecting you? And so we'll spend the last sort of six or so minutes as we get ready for Q&A. Again, if you have questions and answers, you know, throw them into to the chat window down at the bottom or any of the other places that we have available to us, we're going to talk about how it affects you today and every day. Well, it starts with AI being everywhere. That was my very first slide, if you remember.

And the thing is, is that we do, in fact, use it every day. I'm sure that not everyone has a smart TV, and I'm sure that not everyone has a smart thermostat, and I'm sure that not everyone has a smart car. But the reality is since about 2016, maybe 2015, every single vehicle driven at a certain level of quote unquote 'trim' has been equipped with a variety of safety sensing componentry. These are all components that feed into an AI based system.

So Toyota safety sense, they've got this camera at the front, that's that's, and a LIDAR system that's actually pushing out and gauging how close you are to that next car. The vehicle I drive, which happens to be part of the GM family of cars, has a counter, forward collision countermeasure. You know, Tesla is always in the news talking about their full self-driving capabilities, which is really interesting because we don't have the right legal framework to actually enable all that where you can take your hands off the wheel and, you know, sleep.

That's not there yet, but you're even using it in some very basic things. We're about to hit tax season for most of you. You know, our RRSP deadline was the other day. You're probably getting ready to do your tax, whether it's TurboTax, Netfile, Ufile, you know, there's dozens of of these all of these programs and the people managing these programs. If you're going into an H&R Block or you've got an accountant, they are using A.I.

They're using it to, again, try to support and help you cull through all this data and help you make better decisions. And so I am very fortunate that I embrace this technology. A lot of people don't.

I find myself fortunate, though, that I know enough that helps me protect myself as much as possible from these things like privacy, ethics, and biases. I'm hoping that out of this you will come away with, okay, I'm going to spend a little bit more time on that terms and conditions type, type scenario. Now, I wanted to put a huge shout out to the government of Canada. They really have been a forward thinking leader around around artificial intelligence and on how businesses can thrive and how we can can move forward together.

They've actually got something called the Algorithmic Impact Assessment Program. When it's essentially a way for you to understand how much reliance you should put on a given type of artificial intelligence. That's fantastic. Back in 2018, I believe it was, a series of of businesses, Mindbridge being the first tech business, signed the Montreal Declaration which is all around ethical design, and development, and deployment of of artificial intelligence. And you know again I think that's a real good testament to us being safe. But it really does now lead us into the final sprint.

How does it affect you and your finances? That was the pull, right? The reality is it affects everything. And we already mentioned the whole tax and doing your tax returns and your filings. But there are so many other areas of your financial ecosystem that we need to talk about. So back in 2017 Toronto-Dominion Bank or TD bought an organization called Layer Six. Layer Six is an artificial intelligence team that was building amazing programs for financial services community.

And they have pointed all of those members into internally developed different tools and techniques. You may have seen if you're a TV user, you may have seen in their most current apps, they've got this thing called... what do they call it again? It's the spend alerts and the I think it's called 'My spend'. And it actually shows you how far, you know, above or below last month, and what your trends are.

There's elements of A.I. that are baked in there to try to help you figure out where you need to go. CIBC, I believe it is, has a program where they actually challenge you to save more. Right.

So they've actually, as part of their app system, when they go to pay a bill, it asks you if you want to push some to savings. Then you've got, you know, ClearBanc or Wealth Simple or some of these other great Canadian upstarts in the, in the banking and wealth management space where they're using A.I. to find the best product for you or the best investment for you. Even organizations like Caisse de dépôt et placement du Québec, or CDPQ.

A lot of their investment thesis is now being driven based on a variety of analytical programs that are steeped in A.I., as is CPP, which is our Canadian pension plan, obviously, or, you know, other private pension pension holders. So it really does impact you all the time. And what what's kind of interesting is the next stage of this, which is how A.I. and the white collar jobs transition.

Now, I come from a space in the last five years of working with corporations and working with public accounting firms who deliver on audits. Why is this important and why is audit, you know, why am I mentioning this about how it impacts you? There are some major failures that have happened over the years. For those that have been investors for maybe a couple of decades, might recall the Sarbanes-Oxley Act that was enacted in the United States that was as a direct result of big failures like Enron, WorldCom, and Tyco. These were big malfeasances composed of senior leadership and in those businesses actively, actively hiding money, moving money, doing very strange accounting things. And it made the companies look bigger than they were, more more performant than they were. People kept investing.

And then, lo and behold, big failures. In Canadian recent times, there was a big furor around the likes of Nortel Networks and even more recently in a few other parts of the world in Germany, with Wirecard, or in the United Kingdom or in England with Thomas Cook Travel, which actually affects all of us. I think all of us want to be back on a plane or at least having the ability to go and travel.

Thankfully, most restrictions in most provinces are are moving on. But Thomas Cook travel is a great example. It had a clean bill of health, an audit performed by a very large public accounting firm.

Six months later, filed for receivership and bankruptcy. So it affects you. It affects you and how you invest. It affects you and how you bank and how you, how you, you know, transform your ability to have wealth. And so as it relates to my specific world, and this is not a plug for, you know, for for Mindbridge being this great company, although I love it and I love being there.

The ability for us to transition, to have folks like auditors and financial professionals be able to use artificial intelligence to spot those errors and spot those challenges as quickly as possible is going to be a requirement for a more performant financial ecosystem. As you can see here, Klaus Schwab articulated that by 2025, the respondents expect to have seen that almost all of the corporate audits will have had some form of AI performed in them by 2025, which is fantastic to see. I'm sure it will actually take longer. Everyone that puts a stake in the ground, they don't think of all the other factors that go into this, but that's kind of where where we are. A.I. is all around you.

You're definitely working with it, are accepting it as different parts of your livelihood, and parts of your day to day life. And what I'm hopeful for is that people will start thinking about, Okay, how can I use that? Or How can I find products that are going to use that to make sure that my financial stability is there in the future? So last thing I'm going to say before we drop into question and answers, right, is: A.I. augments human capacity, it doesn't replace humans. Right? There may be a time for singularity. I'm not here to opine on that, but definitely part of the job of A.I. is to make it easier for humans to do more things either individually, or for their business, or for the hopefully, the world itself.

So that was my a little bit of opening up the curiosity. Michelle, maybe we can we can talk about where we go from here. Great. Thank you so, so much, John, for sharing some of the promises that are coming out of this this new kind of branch of technology, but also some of the challenges and some of the pitfalls and places where we could trip up. So I'm going to, I'm going to invite our audience...

there's already two questions that have come in. But I'm going to invite our audience to find the Q&A button at the bottom and type them in, and we'll try and get through as many as possible. But, John, I figured we'd throw you an easy question first because the questions coming in are kind of deep.

And so yes or no, and then you can explain further but... do you think we'll ever get to the point where finances, bookkeeping, auditing, etc. is ever going to be fully automated by A.I.? Yes. And I say that because we're already starting to see some of this happen.

And there's three technologies. Well, four. So there's three subcomponents of artificial intelligence plus an ancillary technology that will support this. So one is optical character recognition.

So this is the ability for computers to essentially translate, transfer a picture into text, right? So think about all of the the statements, the invoices we get in the mail. Imagine that actually all just being fully digital all the time. We're not even there yet. Right? But if you could have that, that's a stepping stone to having it fully automated with an AI system. There's already tons of technologies out there from companies like UiPath, Microsoft, Blue PRISM that do something called robotics process automation in this space. And so what they do is they actually use robotics process automation.

So essentially taking those those OCR elements and actually going through and posting entries in a business. So I went and I bought janitorial supplies or I have an invoice coming in from my marketing agency. It literally comes in to, think of it like a big file folder electronically, that RPA will look at it. It'll say, Oh, this one goes to janitorial expense, this one goes to marketing and advertising expense. Posts it. And then a bot will pick it up and say, Oh, it was net thirty on the janitorial and it was net 45 on the, on the marketing and I will now pay it and it will go in and create the, the banking to go and submit those funds to those vendors.

And then it will reconcile that at the end of the day that what I came in, whatever that amount was went out of my bank balance. Job done. So that's a really interesting place for us to be. If, apologies, that was not... meant to hopefully you didn't hear too much of that ringing. That was like I didn't even know where it was coming fro.

The joys of being at home. So that's that's the second piece is RPA. The third piece is, is actually the ecosystem of full and end AI players, right? Where like there's Bot Keeper, which is actually a process where you can submit anything from customer orders, invoices, etc. It'll plug it in. The last piece of the puzzle for me, though, is actually blockchain.

So totally different technology. We probably need to do a Curiosity on Stage on that at some stage. But blockchain is where you will have a level of transparency and a level of acceptance by all the parties. That will allow us to use AI to do the full spectrum.

The thing is, even if we get there, it's not going to replace a level of oversight that we need, whether that's in regulatory bodies, whether that's in human bodies at those individual enterprises and organizations. And for yourself, right, I don't think you you know, we already get direct deposit payments. We already get all of these things. I think there's a level of of human that always assists.

Right? It's but how much can we actually push down? I think it's the vast majority in terms of bookkeeping and presenting financial statements. So you're saying the AI is going to do a really good job of finding fraud in my credit card statement, but I'm still going to have to skim through and make sure that I bought all those things. Exactly. And there's there's a great Newfoundland based company, Verafin, which is protecting all of us.

Most, at least. Most of the people who are banking here in Canada they will be one of their customers. And they're already doing some of that for us.

But yes, you should always eyeball your bill and maybe eventually you start looking at other types of alerts. Right. And that will all be AI based. And this is what TD is trying to do with their, I think it's called my spend report, where every frequency that you set up, it'll actually go and look at the types of spends you have in the types of categories and say, hey, there's a blip over here, right? So that you're you're sort of drawn to it. And I think that's that's what we're trying to get to. We're trying to get you to the the thing that matters, not the "Yeah, every week I have a payment for this, and every month I get my mortgage payment".

And, you know, it's not that. It's "oh, that's a spike, that doesn't make any sense" or, "Wow, you're spending way more on shopping than you ever had". I mean, obviously, for credit card fraud, that's that's the place to look at. Look at those retail. Look at those. Unfortunately, oil and gas is a is a really big proponent of that.

But retail, travel, and things like your your gas for your car or places, for sure, you should be looking at that every every statement. Great. Thanks. So I thought, I think it was funny that you just had a little technological issue here because one of the questions in our chat is the... I mean, there's going to be a crash at some point.

Something's going to go down. I'm assuming that an AI crash at some point is kind of an inevitability. Could it fail? What would be the repercussions? Are there fail safes? You know, that's...

It's it's hard to assume that everyone's going to do the right things. So in a perfect world, yes. When it fails, it fails gracefully. Right. It will. There's a lot of redundancy in a lot of systems that exist today.

Although, you know, we see service outages all the time. Right? With with products that we use. So the question is, when you're building that A.I. system, what is the level of of really transparency in that element of failing gracefully? Right. What did happen? What do I need to look for? So when it happens, what I'm hopeful for is not like a... I was going to use a TV reference, but that's probably it won't translate to everyone who hasn't seen it.

But, you know, we don't want to have this situation where the world goes dark, right. All of a sudden just everything's turned off because the A.I. system failed. And we have to work really hard to to make sure that we don't have that situation happen.

And I think that's one of the reasons why, as much as the technology exists today to go way further, and I'll use the Tesla example that I mentioned. A Tesla today could literally drive in the city of Ottawa right? With the person sleeping from point to point and with a high degree of confidence in the nineties. It would get there without incident, without any issue.

We're not ready for it as humans, though. And therefore that sort of give and take of how much we're willing to adopt is going to slow down getting from nineties to 95 to... In the computer industry we look at not, you know, five nines or seven nines. 99 point and then you know, either three nines or five nines as being the level of stability we can provide.

Right. Most SAS vendors or subscription vendors will be looking for that for their uptime. And that's what we need AI to be and we're not there. Right.

That's just the reality. Can we get there? Yes. But there's got to be this almost two way dialog between well, three way, if you include the governments governing and regulators governing, you know, whether it gets used. But there has to be this this interaction between the AI provider, whatever that looks like, and the consumer to get to a point where we're happy at the end state. Yeah, definitely, great. Thank you.

There's a couple of questions coming in here about ethics. We're going to take parse this out a little bit. I'm a bit of a sci fi nerd. I like my sci fi movies. And we often see AI kind of portrayed as these villains.

If we want to think like: Terminator, Space Odyssey, Prometheus, Westworld, Blade Runner, like the plot's all very, very similar. So as we're building these A.I. programs, how do we go about building ethics into them? So you talked about some of the ethical choices that need to be made, or how do you actually put ethics into an A.I. program? Uh, so the best way I can describe or the best way that I would think about it is we need to have A.I.

systems that have built in checks and balances. Usually we talk about it in terms of resiliency. So resiliency in an A.I. context is that there are Fail-Safes that are constantly checking the things that we want them to do. I'm not going to disclose that the party that this happened to, but there's a very large technology firm who was using an AI bot to sift through resumes and decide who got selected for for things like interviewing.

Right. And so this is a very ethical challenge. Right. We we are all striving for a level of diversity and equality, for sure. Most tech firms, this is one of the things they think of all the time.

There's massive pushes into it with programs dedicated towards STEM or science, technology, engineering and math to increase the level of diversity. But there was a large firm that was using their historical data profiles of existing employees to then infer who they should give give the time to in an interview situation. And so in a, in a again, gender DNA perspective, male dominated environment in tech for the last 40 years, you can imagine that this bot did something that was not very good. Very biased. So they had to scrap it. Right? So they had to scrap it.

And so so the way to build these, you know, these things in is is give it more obfuscation of that PII that that sort of sensitive personal information. Things like gender, and look at core elements in this example. Right.

Strip away the name strip away anything that could relate it back to a gender, age. Right. Because ageism is a thing. There's something in the news of one of my former employers that, you know, talking about.

I'll just say dino babies and you can go and search what that that looks like. It's... we have to strip away elements of this. And so when we're building A.I., we need to be thinking of these things and it needs to

be resilient and not single fault tolerant. Right? It needs to be multiple, multiple fault tolerant and it needs to have dimensionality. And so using a big monolithic system is not going to be good for for pretty much anything.

And I would I would urge anyone thinking of A.I. and building A.I. to go as wide as you can with, if you will, dimensionality of of what you're looking for in order, in hopes that you will remove some of those ethical challenges because it will look at the dimensions as their, their natural state, versus looking at it in terms of what could be ethically compromising. But at the end of the day, A.I.

today is a tool. Right. And so the other problem we've got to solve is: do people using the tool subscribe to a level of ethics? So it's a bit of... there's no easy answer to that one, really. I think going to a resilient, multifaceted A.I. approach is going to be way better than trying to build a single system that looks at every information and treats every information as a sort of... yeah, I think you kind of get the point.

So following up from that, then, obviously the AI's using all this this data, this personal data of ours. Are there regulations either in existence or that you think should be put into existence by, say, government safety and government about personal data gathering and how it's used? So there there are quite a few, everything from the Canadian anti-spam legislation, which starts to safeguard what information you collect, to areas within the technology itself, where they have to identify the technology, the data they have on you. And if you request it, you can delete it. There's there's some things that have already gone down the path of stronger regulation, stronger awareness and transparency.

There's more to do. I think what was amazing about the Canadian government's foray into this is they actually built a fairly rigorous program to think about how we should, we should look at AI in businesses and specifically for the work they do. This is the AI impact assessment report, which essentially is a big piece of trying to get there. The other piece is the Montreal Declaration. So it's it's an idea that even without regulation, firms will sign up and do good. Right.

So based on, on, it's a commitment. And right now it's a, very much a... what do you call that? Sort of like an honor system. Right. But I think it will have to progress further and it will. The reality is that it will.

We will get more and more regulation and companies will have to reduce the amount of personal data they capture without you know, direct related consent. And you see this mostly - I'm not picking on a specific industry - but you do see it in what's happening with with our smartphones and our smart devices. Right. How much information is shared when you open up from one app to the other app? I don't know if you've noticed this behavior, but if you're shopping on Amazon and then you go to Facebook, you get some really interesting ads, right.

Typically directly related. And that's all around around tracking you individually. You can set certain things in your Google profile. You can set certain things in your Amazon profile. You can set things on your smartphone to limit that level of of sort of advertising tracking. It's it actually is - on an iPhone anyways - it's called limit ad tracking, and that will help separate it.

But the reality is they're going to find a different way to get to a similar programmatic answer, which is geofencing. Right. So where is the device? You know, where does that device normally go? Does it go to... So I live in Ottawa, you live in Ottawa, right. Are we at Bayshore Mall or are we at Rideau Center or are we at Saint Laurent shopping center? Right.

Very different profiles of stores in each one of those. Right. Did that device stop at X? And so now when it sees that device and it has a particular home, they're going to try to figure out. Can I can I get there? There's going to be no easy answer other than, I would... So I applaud what the federal government's done. I actually applaud what the provinces are doing right now as well.

But I would say there's more to come. And we need to be very laser focused on, you know, investing in businesses that are willing to make the step of being open, transparent and building ethical and responsible AI and not defunding any of these other ones, but making it more of a reporting regulation issue where people are aware of what they're buying, how they're buying it based on, you know, how that company performs. So, a follow up to this. Do you think the onus should be on the companies and the government like? Consumers, I mean, you talked about these terms and conditions, which I'm guilty of definitely not spending the couple of hours required to sift through all that information.

But people should kind of be aware of what they're sharing and how it's going to be used. Do you think there's a way we can better help people understand what they're sharing and how it's going to be used? So I've got a few friends in the legal community, so I'm going to offend them right now, maybe. I apologize if anyone is online and in the legal profession. I find that the most infuriating thing at the moment is how long it does take you to read those terms and conditions and how long it takes you to find the button or the key that says, No, thank you. So I think there's... and I get it right?

And why I call it the legal profession is they are writing it in a way, right, which covers the bases and limits the liability. Right. We don't want to become a litigious society where anyone and everyone is suing every company for everything.

So I get why they're so long. I get why they're so involved. But I think simplifying the language would be, go a very long way to people being more comfortable and confident, and making the choice to say yes or no in an opt in or opt out.

Then I'd follow that up with, you know, we need a better way to get at the information. So that's on the company. But definitely government has to play.

So it's really it's not a single industry and it's not a single group. It really has to come from educating consumers and individuals, having a layer that that abstracts the legalese and the minutia detail of protection of liability, right, and IP protection into a more simplified state that people can understand it. Companies signing up to do good things and then obviously government supporting that and enforcing especially in... I mean, we've already got massive regulation around banking, insurance companies, anyone that's, you know, that's touching your finances as well as, you know, even just things like CRTC and what's happening in communications.

Right. Rogers and Telus and Shaw and Bell, they have other levels of, you know, what they're allowed to keep and store and capture based on on your consumption of content right. But I think it's it's all three parties, right? It's the businesses for sure. Right.

I think, you know, if we gave transparency of whether you are using ethical, and you have you know, you signed up for it and you get an ethical you know, audit, non-biased audit whatever it might look like, that would be a step for the businesses. I think individuals need to get more in tune with it. And that means we actually have to force the businesses to simplify. Right. So that it can be for everyone. Right.

Versus just folks that have gone through and understood the legalese. And then government has to have the right and appropriate influence from a regulatory perspective or from a, from a direct sort of consequence perspective of of whether these businesses should be able to do the things that they're doing. Right.

We're coming close to the end. So I want to do a little bit of a speed round with you. A couple of these questions.

So the rule is you have 30 seconds to a minute to answer the next couple of questions. All right. Okay. Question one. Do you think AI tech will reduce financial inequality or exacerbate it? Currently exacerbating. The desire to have it reduce right.

I think that it should, it should get to the point where it reduces that inequality. People like Wealthsimple and a few others, getting better access to trading tools and information and smartly investing is huge for people of all walks of life. But right now, it is sided the other way. It needs to become more more equal.

Right. Number two. Humans use IQ to measure intelligence and we know there's issues associated with that. But is there such a scale or a rating system for A.I.? Not yet.

The algorithmic impact assessment tries to get there in terms of what the Canadian government believes to be rating it in terms of how much reliance you can you can have on the AI system. And so I think they've done a good job to start creating that level of awareness. There's nothing concrete, and this is a bit of a dilemma. It needs to be a concerted effort from the firms and from from the public coming together, trying to get to something that they can, they can agree on.

Right. And that's going to take a while. So there is no scoring system today, but the more diverse and the more resilient someone's built something, they will want to tell you about it, because that's, it's the way to go.

I do imagine it'll be tricky to come up with a system when all the AI programs are programed to do such different things and use different types of intelligences. Yeah. Are there any AI technologies that you think are a little scary or that we should be wary of? Um, I wasn't expecting that one. I don't... I don't think anyone should jump in to an area that they're not willing to... It's a risk reward system when dealing with AI, I guess is what I'm going to try to say.

I don't think... it's not a one size fits all. So my my tolerance for risk on on whether A.I. is good or bad because I happen to be in the space, I probably am a higher degree risk profile right? I'm willing to take more risk because I understand elements of those consequences. But there's no technology out there today that's that's really being used, that I think is doing something nefarious.

Yes, you get the phishing scams with the Prince of X-Y-Z Country asking for money or you won some lottery. Right. I think that type of A.I. trying to target you that way. Yeah, we got to we got to stay protected. But I think for general mainstream use, in things that you're probably touching today, I don't think there's anything that's that's super scary yet.

Yet. All right. Last question. Do you think we will achieve singularity? An A.I. that is self-aware, intelligent on all of the different levels that we consider intelligence, which passes for a living being.

I think there will be enough experimentation that based on our current definition of what it takes to be a human, the answer is yes. I think there is a point in time where we will do enough experimentation with robotics and all the accouterments to get to that point with all the coding systems that that is that is, in all likelihood, based on what we qualify it as today. I think we should maybe update that a little bit. Because I think of, I think of just you know, just simple tasks right? Driving the car. An A.I. system is going to absolutely be better at driving a car than I will. 100%.

And it should be there and it should... But is it sentient enough to know while it's driving: "Oh, hey, I forgot about this other thing that I was supposed to do, so I'm going to make a left here." Could it happen? Yeah, for sure. Should it happen? Not sure yet, but I think definitely based on our current, you know, definition, you know, Webster's Merriam Dictionary, whoever we want to use of what intelligence is, we will absolutely get to that point at some stage with with A.I.

systems for sure. The question is, is whether we move the needle or not on what we feel it is to be intelligent. Great. Well thank you. So we're at 4:00, so it is unfortunately time for us to wrap this party up. So I'd like to say a huge thank you, John, to you for speaking with us this afternoon.

Thank you for your time and your passion and kind of showing us, giving us a bit of an insight into A.I. that we might not have might not have had before. And showing us some of the ways we might not have noticed it being used before. I&

2022-07-28 07:52

Show Video

Other news