Welcome to certified information securities presentation on artificial intelligence as used in business. Now, while this presentation isn't actually part of the tested content of our training and certification program for the NIST CSF risk Management Framework 1.0 architect credential. It is, however, meant to provide a context for the use of AI in business and therefore the risk to business that it presents. After all, that's what we need to manage in order to ensure that we have effective use and benefit from AI. But also manage the corresponding risk that comes along with it. I'm Alan keel. Thanks for joining me. Early on, technology visionary Bill Gates once said that the power of artificial intelligence is so incredible that it will change society in some very deep ways, and how we've come to learn that this is true. We'll begin this introduction with a short presentation on the general foundational concepts of artificial intelligence. We'll then look through a brief history of AI development and where it's come to. Today we'll then talk about AI as used in an organization in a business context, as well as other areas of AI application. And then finally, we'll segue this presentation into AI risk management, which is the core content of our. Actual training for NIST AI risk Management Framework, 1.0, architect training and certification. Now we'll look at defining artificial intelligence. Artificial intelligence is described in many different ways. However, one approach might be that artificial intelligence is a machines ability to mimic human action and imitate logical thinking, learning, planning and creativity. AI systems can perceive and analyze their environment to solve problems independently of human assistance. So a unique feature of AI systems is that they can learn from past situations, and they use this information to adapt their actions in the future. So how does AI work? Well, artificial intelligence again enables computers to perform complex tasks by using human like faculties, such as data processing, pattern recognition and decision making. Machine learning is a key aspect of AI that uses algorithms and learning processes. To adapt to new data and improve performance over. Time. And then finally, we have computing power that is crucial for the functioning of AI systems. It involves memory capacity and rapid data processing to enable AI to perform complex tasks efficiently. So artificial intelligence is a broad term that can be broken down to get a better understanding of what part of AI we're looking at. So we'll compare and contrast artificial intelligence with machine learning and deep learning. So technologies that allow machines to mimic human intelligence. This is an example of artificial intelligence. An example of this would be perhaps a self driving car. We also have machine learning, which is a machines ability to actually think and independently solve tasks using large amounts of data. An example of this perhaps would be facial recognition to unlock a smart. And then finally, we also recognize that deep learning is a machines ability to train itself through neural networks and big data. And an example of this in recent times would be ChatGPT. So going macro to micro, let's dig a little deeper into machine learning and deep learning where we look at algorithms and neural networks. So again, machine learning is an umbrella term for algorithms that learn from data to recognize patterns and to make predict. Traditional neural network, then, is machine learning now. This then leads to more complex neural networks known as deep learning and deep learning is a specialized sub field of machine learning. It uses multi layer artificial neural networks. To learn complex patterns in large data sets and to mimic human like thought processes. So as we've learned, artificial intelligence is progressing from simple automation of tasks to actual cognitive thought and decision making. So as we look at the evolution of artificial intelligence, we started with reactive machines. These are considered classic weak artificial intelligence. That can only perform a task for which it was specifically programmed. In other words, it's an automation. Packs. Then we then look at limited memory. Most common types of weak AI use this today. It collects and analyzes data to apply it to current events. We then have theory of mind. This is where theoretical and powerful AI is used to perceive, understand, and respond to human emotion. It's. And finally, we're looking for that self-awareness that we all fear when AI actually becomes sentient. This is where an AI artificial intelligence system that in theory could actually reach or even exceed human level self consciousness. So how do we know when we have AI versus just good complex computing? Well, it turns out that Alan Turing was a pioneer in artificial intelligence, and he had developed a test to determine if artificial intelligence can actually be determined to be present. So developed in 1950, this Turing test was considered a significant milestone in the assessment of artificial intelligence in the test, it began with a human examiner who engaged in a text based conversation with an AI system and another human without knowing the identity of either one. And AI or artificial intelligence system passes the tests when it is able to successfully simulate human like communication and its responses are indistinguishable from those of the human in the experiment. So what can artificial intelligence actually do? Where do we see it in use today? Well, it's not just ChatGPT. It's not just open AI. We use artificial intelligence to facilitate speech to text conversion. To provide deep learning and predictive analytics to be able to use generative AI. To where we can actually provide an input that is then used for AI to independently calculate and output that we may use for our own business purposes. We also have AI use and planning and optimization. We have neural language processing which is classification translation. Data extraction. We also can use AI for image recognition and machine vision. We also use AI for robotics as well as expert systems to help us better determine decisions considering large volumes of facts. So this is where we start to consider the risk of AI that needs to be then weighed against the benefits of AI. Because while AI can bring just an incredible amount of opportunities and advancements, unfortunately, along with AI comes also the potential. Risk of AI. Potentially breaching privacy or even being used for malicious purposes, so the benefits of AI include automation of routine tasks, improved decision making, personalization and customization of the services that we purchase and enjoy, and also can help us in progressing medical research. It can provide increased efficiency in product production and logistics. It can also add to and can bring value in creativity and innovation through recognizing patterns and problems. But again along with this we recognize that risk are incumbent to AI. As well, we have potential loss of data protection and privacy. We also have to be concerned that as AI uses data that it accesses independently across large volumes of data sets to then determine its own opinions and ideas on applicability and and. Usage we have to worry that will AI actually have an understanding of ethics and accountability as we understand it? And what about potential job displacement or potential fairness and discrimination issues as we use AI and then don't forget that while AI can help us? To accomplish good things, it can also help people to accomplish bad. Things. With more efficiency and with better speed, we also have to be worried about the potential transparency issues and the explain ability of AI systems. After all, how can we trust in an AI system if we don't know how it derived its information that it provides? How it derives its recommendations? How do we know what's behind the AI that gives us these fantastic, well thoughts of artificial intelligence? And now we'll look at a brief timeline of AI development to see where it's come from and get an idea of where it's going. Our first example of artificial intelligence might be considered the touring machine, which was our first universal computing machine. Of course, Alan Turing, as I said, brought about the Turing tests in 1950, where we can then use this test to determine if AI is actually present. We then have our first. Functioning AI program in 1956, followed by the imitation of human problem solving behavior in 1957 through 60. Five and then towards the end of the 1960s, we have our first chat bot named Eliza, developed by an MIT professor. We then have in 1975 through 85 AI research that becomes widely known through expert system technologies. And then in 1997, we were excited to see IBM's. AI chess machine deep blue becoming the world chess champion and then in 2011 IBM developed Watson, which went on to of course win game shows. Then we have in 2016 Google developing Alpha go following it with Bert, a framework for pre training NPL models in 2018. Then in 2022 we see the acceleration of AI where open AI releases, chat bot, GPT and image generation AI. With Dally too, this was where AI really became recognized throughout the world as true artificial intelligence capability accessible to the. And then of course, here we have in 2024 already taking advantage of Microsoft Copilot that is now built in as an additional feature to its Office 365 Program suite. As AI development has accelerated, and as AI is being adopted for its benefits to business and life. So has AIS impact on the global economy matter of fact, referencing a recent study by McKinsey, we can see that 165 billion was invested in artificial intelligence in 2022 alone. 52% of the companies surveyed had invested more than 5% of their digital budget. And artificial intelligence in 2022. Two. We also had a 16% increase in enterprise value of private AI companies in the single year between 2021 and 2022. And finally, we have 292 AI companies that hit valuations of over a billion dollars and it's. Only grown faster with more acceleration today. So where do we see AI being used today? Across the world? Well, it turns out everywhere. Now there are some areas of the world that are adopting it with greater ferocity than others. So for example, as we look at the various countries represented in this chart, that again was drawn from a study. And by IBM in 2022, we see that actually we have stronger adoption of AI in China, India, Singapore, United Arab Emirates. Whereas we see equal amounts of exploration in AI across all of the countries. So let's look at how AI is actually changing the way that we do business. An early pioneer in the automotive manufacturing sector, Henry Ford once said. If I had asked people what they wanted, they would have said faster horse. But of course, he didn't go with what people had wanted in the past. He saw the future as a visionary, and while he wasn't the inventor of the automobile, he was good at making the production of automobiles efficient and was able to capitalize on the fact that for transportation. Automobiles and motorized transport would eventually replace our reliance on. Horse and buggy.
But.
So again, referencing IBM study, we see that 35% of the companies they surveyed were already using artificial intelligence in their business in 2022. In fact, chances are there is a lot more artificial intelligence in use at your organization than you realize. 66% of the company surveyed are either using or planning to use artificial intelligence. To address their own sustainability goals, 54% of the companies surveyed are already saving time and money by using artificial intelligence. And 48% of the companies surveyed report better customer experience through artificial intelligence. As I said, you're probably not aware of all of the artificial intelligence that is being used to facilitate the many things that your organization does. We now have AI in our customer service and our accounting, our sales, our marketing and our HR management, we use AI. In our manufacturing processes and quality management in our supply chain management and in everyday business processes. In 2022, Mckenzie's report gave us further insights as to actual applications of AI in companies. Organizations are using it to optimize their service operations. Organizations are now actually creating AI products, not just using AI. For more effective production of existing products, we're using AI for customer service analytics for customer segmentation and understanding. We now are enhancing almost all of our products and services with AI. Matter of fact the joke is we'll now soon see AI enabled toothbrushes. Maybe we already have. We also are looking to use AI for better customer acquisition and lead generation for call center automation for product feature optimization, risk modeling, analytics as well as predictive service and intervention. So we naturally want to run to the advantages of AI and how it's going to make, well, our organizations more effective, more profitable. Helping them to XLS well as how we can use AI to improve our very lives. Unfortunately, however, we have some impediments to using AI things that we need to overcome. For example lack of skills. AI has recently come upon us. And with that, we have insufficient skills and understanding of how to use and how to develop AI, so this lack of AI skills limited knowledge will actually naturally impede successful adoption of AI well. And also we have to consider that AI can have high costs. The costs associated with AI and the preparation it requires, such as development of skills, may simply be cost prohibitive. We also recognize that today, in the early stages of AI adoption and. Women, we have inadequate tools and platform availability. Using AI is challenging without the necessary tools or a platform with which to develop our AI models, and our AI is by nature complex and when projects are too complicated. It can be challenging to successfully integrate and scale AI systems. We also have high data complexity that goes along with our complex projects and when data complexity is high, AI deployment requires. Special new expertise that we haven't yet widely developed. We'll need specialist and significant computing power to be able to manage the high data complexity that is inherent to artificial intelligence. And then finally, we have an overall concern about our confidence in AI. How can we trust that AI will will be accurate? How can we trust that AI will be utilized and be be at our service according to our own? Ethical expectations and understandings. Companies are struggling with how they can use AI but use it responsibly so as not to actually bring more risk than benefit. So this means that organizations tend to be at different levels of development and adoption of AI within business. This goes for countries and society as well. So we can take a look example at level zero. Level zero is where we would see that we don't really have any presence of AI or even automation. We then progress to weak AI, which is where we have semi automated companies progressing on to then level 2 with strong AI where we have automated companies and then finally we actually have AI being used to cognitively determine new information and new solutions. And really, new data and information with super intelligence at Level 3. We're now ready to take another look at other areas of application of AI and business. So AI is revolutionizing the way that businesses operate. It can provide valuable insights and improved decision making, and AI is transforming industries across the board. AI helps us to make Better Business decisions. It helps us to better analyze data and identify patterns. It can provide valuable insights into customer behavior as well as help our businesses to identify new opportunities. And as I said, AI does help our organizations to improve efficiency and productivity, which is great so long as yours is the organization to actually take advantage of it. So the organizations that use AI. Why? To save time and money to help them identify areas for improvement that use AI to help reduce operational costs naturally become more competitive. Organizations that fail to adopt AI naturally become less competitive, which is an AI risk. Other real world examples of benefits of AI include those in government where AI can enhance public safety and security where it can streamline government services and reduce costs, and where it can even potentially improve public health and prevent disease. We are already seeing rapid adoption of AI in healthcare where we use AI to improve disease diagnosis and treatment. We use AI to help personalize our medical care as well as potentially reducing medical errors. Being able to understand how to address achievement of benefit and value with AI, as well as managing the corresponding risk. Requires getting a sense of feel and understanding for again how AI is actually used in business. So again, more examples within the healthcare sector would include how we can use a machine learning program to remotely analyze patients health via a mobile device, comparing it to. Medical records and warning of potential diseases. We can also use AI to perform image recognition during an examination to compare the collected data with specific disease patterns and to flag conspicuous areas on X-rays or ICT scans. We can use AI for autonomous diagnostic devices that can perform simple medical tests without human assistance, relieving doctors and nurses of routine activity. We can then use AI to improve our healthcare by predicting patient behavior and the likelihood of illness. We can use AI algorithms to optimize hospital processes, staff schedules, and inventory. AI tools can also be used to analyze environmental factors and patients medical backgrounds to recommend preventive care programs. We can also use AI for virtual consultants who will register patients, recommend suitable doctors and reduce waiting time. We can use AI to provide AI personalized treatment plans and to improve treatment efficiency by tailoring treatment to patients medical needs. And then finally, AI can also perform and compare large population health analysis for us as well. Shifting gears, let's go ahead and take a look at how we can use AI in financial services. AI can improve financial modeling and forecasting. It can help us to detect fraud and money laundering. It can also help us to automate routine financial tasks. In the transportation sector, AI can improve traffic flow and reduce traffic congestion. It can help us to enhance vehicle safety and reduce accidents, as well as potentially optimizing routes and reducing fuel consumption. Manufacturing AI can optimize production and increase efficiency, which of course then leads to greater profitability. AI reduces waste and help us to improve sustainability. AI can help monitor and predict machine maintenance requirements as well. And in the retail sector, we can use AI to enhance the customer experience while also enhancing our ability to sell more to the customer after all. Exposing the customer to what they want in a more practical and more efficient way helps them to buy so we can use machine learning and facial and speech recognition to enable virtual advisors to greet customers in person to predict their purchases and point them in the right direction. Now I've already seen. Where this came along with the. Risk a perceived lack of privacy. So for example, there's an example where a department store and a mall had actually used AI to monitor and watch customers as they were browsing and shopping to determine what they would most likely then want to buy. To complement what they had already purchased. So as you are leaving one part of a department store for another mysteriously. The TVBS are showing ads of products that relate to what you just put in your shopping cart, which of course gives people an uncomfortable sense of lack of privacy in their shopping experience. So again, this is an example of how we'll use AI to try to improve our business, but we also have to consider the risk that comes along with it. So when it goes to the customer experience, we can recognize that machine learning uses customer profiles to personalize advertising. And again, as I said, as customers walk through the store offers could then even be sent to their smartphones. We recognize that a computer vision system identifies the items packed by the customer and therefore could potentially automate the payment processing, allowing the customer to simply leave the store directly. Being charged for what they had picked up and taken. We also see that autonomous drones could use deep learning that can take over short haul deliveries, reducing wait times and to free up delivery crews. And we can also use AI to improve our processes themselves within the retail environment. Interactive screens can help identify items or recommend complementary products that matches a customer's profile. We can use again autonomous shopping carts that follows the customer in the store, automatically finding their way to the car. Or a drone if it's for home delivery. Stores could then optimize their prices in real time to maximize their profits. They are guided by competitors, prices, the weather and potentially inventory levels as well. And then finally, robots with artificial intelligence continuously track inventory and empty shelves to prevent stockouts. We can also use AI to maximize efficiency of energy supply sensors and machine learning enabled near real time response to wind conditions in order to maximize green energy production. Machine learning enables forecasting of peak demand and maximizes the use of short term renewable energy. Digital power systems detect the current situation and adjust real time power supply to a building's current grid load. And drones and robots can detect faults, predict failures and control equipment without interrupting production. And in the service sector, we can use AI to automatically log and forward documents, leaving technical staff more time to focusing their own problems. Field staff could receive real time updates to short response times and to reduce negative impact of outages. Virtual consultants could automate call centers and segment customers. They can resolve simple and common customer issues on their own or to refer them to an employee. And using smart meter data and AI service providers can offer new services based on usage patterns, weather and other factors. AI can also bring value to our business processes. A human instructor could guide the robot as it learns and corrects wrong actions and movements as they occur. Trained robots work together with the human worker and hand him objects. They continuously improve their algorithm in the process as. Well, once a robot has finished learning, it could then repeat the new routine of movements on its own and automatically complete repetitive procedures. And sensors and cameras can also provide information to computer vision algorithms that ensure reliable cooperation between workers and robots. I'm just hoping that we'll get better at it in the restaurants that we see using robots to serve tables. We can also use artificial intelligence to increase efficiency and yield in a manufacturing environment. We can use an AI powered root cause analysis to improve the quality and problem solving in our manufacturing processes. And AI based analysis of process information predicts yield disruptions and helps workers to better manage. We can also use AI to analyze data from past production runs to help ensure greater schedule adherence and future processes. And finally, data on production tools is automatically translated to AI to determine optimal process conditions. So we can now use AI to actually improve our quality management. AI can identify defects that are not typically caught by human inspection. AI can analyze large data sets to identify patterns of failure and to learn from that. AI can also improve product design and. Processes as well. So let's take a look at a few examples of how we can use AI to improve our quality management and efficiency in manufacturing. So for example, we could have AI process thousands of images to automatically to detect defects. On a volume that is otherwise simply infeasible for human inspection. We can ensure that workers are then automatically alerted to the specific location of the defect, so that way they can better and more effectively resolve the issue. We also have AI being able to produce a list of the most common errors that are then created by analyzing past errors in production. And finally, numerous cameras can take pictures, which it then sends to AI machines for further analysis and learning. Other examples include how we can have sensors detect sounds or vibrations, sending the data to the AI system for processing and analysis. We can have maintenance workers receiving automatic suggestions for anticipated maintenance and scheduling. We could then use machine learning algorithms to more accurately. Predict maintenance needs of machine parts, and even perhaps even providing guidance on what needs to be done.
Yes.
And then finally, predictive maintenance significantly reduces machine downtime caused by maintenance work. So we can use AI to more effectively and more efficiently manage our supply chain and reduce costs. So for example, AI can combine relevant internal and external data for highly accurate and improved demand forecasting. This then allows us to better manage our material flows and volumes. So they can be automatically adjusted based on real time data such as weather conditions. This means that artificial intelligence is transforming the traditional supply chain into a delivery network with a digital core. More accurate forecasting allows for lower inventory levels throughout the supply chain, which of course reduces costs, increases efficiencies. And because of its incredible analytical powers and its essential cognitive processing and thought. AI can really help us to better manage uncertainty and risk. AI is transforming the way that businesses identify and manage their risk. It helps us to better predict and prevent business disruptions as well as enabling our businesses to make more informed decisions, not only in avoiding downside risk, but in capitalizing on upside opportunities, which is also risk management. AI can also help organizations much more effectively manage potential disruptions. AI helps businesses by achieving greater resilience by automating and streamlining business continuity planning, as well as enabling businesses to better respond and work quickly respond to. Business disruptions. AI is completely transforming the way that we perform financial fraud detection. Machine learning algorithms are better able to identify anomalies and patterns to detect fraud more effectively than ever before. AI models are then used to learn and recognize new types of fraud and adapt to changing fraud patterns. AI can help financial institutions to process huge volumes of data quickly and accurately, enabling them to better identify and prioritize high risk transactions by automating fraud detection, AI can reduce the need for manual intervention, freeing up staff to focus on. Other areas of the business and with AI financial institutions can enhance their fraud detection capabilities, providing better protection for their customers and reducing the risk of financial loss. AI has become increasingly important in identifying and controlling cyber security risk. Cybersecurity is one of the biggest risks facing organizations today, and AI can help automate and streamline the process of identifying and responding to threats. So one of the key benefits of AI and cyber security is its ability to identify new and emerging threats. Because AI can analyze such large volumes of data quickly and accurately, it can detect patterns and anomalies that may indicate a new type of security threat has come about. This then helps organizations to respond more effectively to emerging threats before they become widespread. Additionally, AI is going to help automate the process of threat detection and response, which will reduce the need for human intervention, and it also enables security teams to respond. More quickly and more effectively to security incidents and by reducing the time that it takes to detect and respond to security incidents, AI helps minimize the impact of a security breach or other security incident on an organization's operations and reputation. So as I explained, it's just impossible to consider that your organization will go into the future without effectively using artificial intelligence. If it doesn't, it simply won't compete. However, how do we effectively seize and manage the opportunity? From a. I without. Incurring the downside of the risk that it brings, well, it turns out that this is a new and evolving. Fight that has come about on how to effectively manage AI risk. Towards that end, there are a few recently emerged recently released international frameworks for managing AI risk within business. We'll now proceed to a quick overview of the frameworks that are actually covered in depth in our. Nists AI risk Management Framework, 1.0, architect training and certification program. AI risk management can enable AI developers and users to better understand impacts and account for the inherent limitations and uncertainties in their AI models and systems. Which in turn can help improve overall system performance and trustworthiness, as well as the likelihood that AI technologies will be used in ways that are beneficial. And while some AI risk and benefits are well known, it remains a challenge to assess all of the negative impacts. Potential impacts and the degree of those impacts. And we need to understand that since AI is integrated into our organization, essentially facilitating and enhancing our business processes, we can't think of AI risk management as something that is done separately for managing risk. Throughout the organization and all of its processes, so AI risks should not be considered in isolation from other business and operational. Yes. Different AI actors, or the parties or people who work with AI systems throughout their life cycle, have different responsibilities and awareness depending upon their roles in that life cycle. For example, organizations developing an AI system often will not have information about how the system may actually be used by the end user of the system. So AI risk management should be integrated and incorporated into broader enterprise risk management strategies and processes. Treating AI risk along with other critical risks such as cyber security and privacy, will yield a more integrated outcome and will provide for greater organizational efficiencies. So how do we see this happening in an organization and how can we leverage some frameworks that have been published to do it better? Well, firstly, we have on an organizational level typical enterprise risk management. This is where an organization establishes consistent management of risk throughout the enterprise. To guide and harmonize niche types of risk, such as AI risk, cyber security risk, privacy risk compliance. Ask an example of a framework to help us establish and do this is for example ISO 31,000 or perhaps even code so's enterprise risk management framework. We then have for AI management in particular, not necessarily just AI risk, but basically an entire management system devoted to managing a AI as it's integrated into an organization. We now have a new ISO standard for this, ISO 42,001. This is entitled information technology, artificial intelligence management system. This can be used to establish a formal system strategy and policy for leveraging effective and responsible use of AI throughout the enterprise. After all, you're going to need to scope this. You'll need to assign roles, responsibilities, tasks. You'll need to fund the management of AI and your organization. Where can you get a plan for putting all of that together? That is ISO 42,001. That's his purpose in life. OK, but what about the risk of AI? Can we do a more specific job on managing the risk of AI within that overall AI management system? Well, indeed, that's where we have the AI risk management framework by NIST. The only version and the latest version. 1.0 just release at the end of 2023. It also complements and works with Isos, risk management approach that it established in Isos standard 23894. These can both be used in concert to better manage risk of potential undesired outcomes in AI development and usage. So again, this presentation really was just meant to give some context to how AI is used in business. So that way we can have a better understanding and feel for the risk that we need to manage in our full on nists AI risk management framework, 1.0. Architect training and certification. So our next stop from here would be the actual introduction within that course itself for establishing and using the NIST AI risk management framework. I'm Alan keel. Thanks for joining me.