A Review of Artificial Intelligence (AI), related technologies and business use cases

What is Artificial Intelligence or AI? What are the related technologies of AI? And what are some of its basic use cases?

In this episode, we will review AI or Artificial Intelligence and how it’s being used across industries by organizations to boost their performance by doing things smarter, intelligently, better, faster, and cheaper. As this is a large topic, one episode won’t be able to do justice. So, we will begin this episode with the basics of AI, it’s underlying technologies, and their use cases. In later episodes on this show, we will slowly build up to cover more specialized topics such as developing an AI strategy within the organization, types of applications and overall capabilities that organizations can start building and other such topics.

So, let’s get started with the basics of the topic.

In the past few years, especially since about 2016 the interest in AI has picked up enormously as organizations for the first time have started witnessing and observing AI’s real business benefits and its direct impact on an organization’s overall performance. If we look at some of the use cases, we can easily see that a lot of organizations now have successfully deployed AI in various forms and fashions within their organizations. In fact, many industry pundits are saying that organizations, which won’t be using AI fairly soon will be left far behind resulting in the erosion of their competitive positioning. The National Science Foundation clearly states on their website and other channels that “To stay competitive, all companies will, to some extent, have to become AI companies.” So, what that essentially means is that all organizations will have to build certain AI capabilities to be able to compete effectively. For CIOs, this also means that not only will they have to start building AI capabilities within their organizations and to show its applications in real business situations but they should also have a longer-term strategy for making this happen.

So, to see what AI is able to do, consider Xiaoice (spells as xxx), a chatbot released by Microsoft, which can talk and interact with its users. In this chatbot, Microsoft has brought in their best research of all of its AI technologies related to speech recognition, natural language processing, machine and deep learning and more. This Chinese celebrity chatbot is learning so fast from its interactions that young Chinese men and women feel comfortable turning to Xiaoice to talk about their issues, heartbreaks, and daily stresses. Many send it love letters and some have even invited it to dinner in the hopes that it (or she) would show up. According to Microsoft, part of the chatbot’s popularity stems from the way she exhibits high emotional quotient (EQ) by remembering parts of a conversation and following up in later conversations. The chatbot is also a poet and has published a book of poetry and helps Chinese people write poems. To date, it has penned hundreds of millions of poems and is currently hosting a TV morning news show with viewership approaching ONE BILLION viewers. Yes, that’s a billion with a B.

In this context, we should also make it clear that AI does not refer to one system or one type of technology – such as a talking or a working robot. Rather, depending on the AI capability that your organization needs, AI refers to a number of systems and technologies that must be assembled and integrated depending on your organization’s business case. So, if you are trying to process large amounts of data and looking for specific patterns and insights in that data or to derive certain types of sentiments to gauge public opinions, you will need a different set of technologies and systems than let’s say if you were putting together a shop floor automation system involving computer vision and robotics or if you were trying to build an intelligent chatbot such as Xiaoice. Therefore, it’s important to always develop the right AI strategy before jumping in. In other episodes, I will accordingly delve deeper into the need to have an enterprise wide AI strategy that must be driven by your organization’s overall business goals, strategies, and objectives.

• Accelerates automation
• Enables derivation of insights not possible through traditional technologies
• Systems are self-learning – keep the context of previous insights and keep improving
• Able to handle and process large amounts of data
• Accelerates decision making due to the availability of the right insights at the right time
• Ability to process complex datasets integrated from various enterprise datasets

How big is AI?

To get an idea of how big AI is going to be in the future and its impact on the global economy and organizations, consider the following:

• A PwC Global Artificial Intelligence Study forecasts that AI will add $15.7 trillion to the global economy by 2030.
• The US government’s National Science Foundation invests more than $100M annually in AI research.
• The number of enterprise customers and government agencies deploying AI to dramatically transform their businesses is increasing manifold. Both organizations are governments have started to take notice of AI and how it can be used for their transformation. AWS public platform, which offers readymade AI based services is boasting more than ten thousand customers on their platform alone.
• At a government level, UAE, for example, sees the adoption of AI so urgent and important that they have created a top level position of that of a Minister of State for Artificial Intelligence. Not to get too detailed in how the UAE government is structured but a Minister position is one that is part of the Council of Ministers reporting to the Prime Minister of the UAE and that would be like a Cabinet Secretary level position in the US. We also see many other countries create high-level positions in the government overseeing adoption of AI both in the government and public sector.
• At a defense level, we see governments of all advanced countries such as US, China, Russia, and others investing a lot in AI.
• And the list goes on.

So, we can see that organizations of all types are seeing many benefits in AI and it should be part of every organization’s strategy and CIOs and technology executives should have a clear roadmap for its adoption and to bring relevant benefits to their businesses. Although as we just heard, interest in AI has picked up dramatically, we are still at a stage where organizations have the opportunity to catapult themselves to great successes and use their data to bring out major transformations to their operations and overall performance. However, if they were to wait longer, it may be too late.

In this Episode

So, in this episode, we will answer 3 questions.

1. First, is what is AI. This may be quite basic for some but before we get into more advanced topics in subsequent episodes, it’s important that we spend perhaps a few minutes getting back to the basics.
2. We will then review some of the technologies that are associated with AI or mentioned in the context of AI. These terms include Machine Learning, Deep Learning and some others. We will look at those and what they have to offer.
3. After that we will get into the specific use cases that apply to each of those technologies and see how they are benefitting organizations. This is not a small topic. So, although we will start it in this episode, we will cover more of this in subsequent episodes where we will cover the various strategies that organizations can pursue to bring AI in their organizations.

After we review the answers to these questions, we will be ready in the next episode to delve into the type of overall AI strategy that organizations need to start adopting AI related technologies in their enterprises and to boost their performance.

So, let’s get started and address the first question, which is ‘What is Artificial Intelligence?’

What is Artificial Intelligence?

There are lots of definitions of AI or Artificial Intelligence, with each providing a unique dimension of what AI can do. To put it simply, an AI program is able to learn and do more things with time without being explicitly programmed. And that’s where the term ML came from. It’s the ability of machines to learn and do more things and help machines achieve Intelligence or as we would like to call it ‘Artificial Intelligence’. That’s what distinguishes AI software from regular software.

AI also refers to a set of technologies in the computer science domain that make computers and machines perform human like activities such as those related to visual perception, recognizing patterns, speech recognition, continuous learning from experience, problem solving, reasoning, making predictions, and others. In some cases, these machines end up doing those tasks even better than humans. For example, using these technologies computers can identify patterns in vast stores of knowledge and information very quickly something that a human mind isn’t able to do. But in other cases, these technologies are still quite far behind, especially when it comes to visual perception, reasoning, learning, and so on.

But regardless, when applied and used adequately, AI and related technologies can help organizations achieve a manifold improvement in their business processes and overall performance.

AI technologies work by a using combination of AI algorithms, lots of data, programming models, and hardware acceleration. Since AI works with a lot of data and the fact that it needs these sophisticated neural network algorithms working in the background executing various models, it increases the computation and processing needs, thus necessitating advanced hardware as well. So, for example, if the given AI problem involves a lot of image processing then GPUs can be used to help accelerate the processing of tasks. As we know, GPUs stand for Graphics Processing Unit and are dedicated hardware units to perform graphics only functions.

So, with the development of these human like functions in software and hardware, we see full-fledged intelligent solutions and machines surface over the past few years, which are helping organizations solve complex problems. For example, we see the implementation of virtual chatbots, which automatically and seamlessly respond to customer and user inquiries thus freeing up agents to focus on answering more difficult types of questions. We also see manufacturing plants collect a lot of data related to their ongoing operations and then using AI technologies like ML and DL to look for specific patterns in that data and to derive useful business insights that are then being used to improve the overall operations by reducing wasteful steps, optimizing processes, reducing costs, and so on.

A quick note about chatbots as their use is constantly increasing. Chatbots are software help agents that allow users to have intelligent conversation with the system, which help the users with answering questions, filling forms, taking orders, ordering food, and so on. In many cases, the person may not even know that they are talking to a machine or software. A number of organizations have installed chatbots on their websites including UPS, Macys, and others helping users with their queries and helping them carry out more advanced tasks. Chatbots can interact intelligently with the users who may ask questions such as “where can I order dinner, or where can I drop my package, where can I buy some xyz perfume, etc.

What are the different technologies of AI?

Now, we will move on to answering the second topic, which has to do with the different technologies of AI. AI is usually associated with a number of key technologies. They include but are not limited to Machine Learning (ML), Natural Language Processing (NLP), Deep Learning, machine perception and even robotics. Again, although these technologies have been available for some time, they have become more accessible and practical for use now due to the world’s new found capabilities to process large data and availability of massive computing power through the cloud platform.

Machine Learning – Let’s look at ML first. ML is one of the most important and foundational technologies underlying the field of Artificial Intelligence, and refers to the ability of a machine to behave not through specific programmed instructions (the way computers and software usually work) but through its own learning. Normally, the program starts with a generic algorithm but then builds on the data that it’s provided to build its logic. The basic algorithm is programmed based on a mathematical representation of the problem that one is trying to solve and then uses available data to train the overall logic of the program. ML usually works on structured data. Essentially, this means feeding the algorithm the data, which is structured and labeled and then it uses that data to learn and improve on its performance.

Typical applications of ML include image recognition where a program becomes better at detecting certain objects such as faces or others. It can be used in existing applications such as E-mails. AI enabled plugins exist today, for example, that allow for E-mail filtering based on certain criteria. ML is also being widely used in medical diagnosis where a program scans numerous X-rays or CT-scans and helps the physicians in the diagnosis of a number of ailments. And there are numerous such examples.

Deep Learning is another technology related to AI. DL can be considered a specialized form of ML. DL algorithms are usually used for more sophisticated and complex cases in AI. DL algorithms work by mimicking the human brain, which is made of neurons and where thinking and reasoning works across various layers of neural networks. Similar to those concepts, in DL, a set of multiple algorithms working at different layers, act as Artificial Neural Networks to interpret the data that they feed on. With DL, a software program improves itself through ANN (Artificial Neural Networks). Data is processed through each layer of artificial neurons and progresses through processing to discover more features and patterns from the data. So, in summary, DL is inspired by neural networks of the brain where the brain’s network works to represent concepts, relationships, see and understand the context, and represent the vast amount of data thrown to it.

To illustrate an example of Deep Learning, some of us may have heard how Google’s AlphaGo system, which was able to beat humans at the game called Go. Go is an abstract strategy game that is played by two players in which the idea is to capture more territory on a board. By using DL techniques, the machine was able to learn by first playing many games with many players and similar to humans kept learning from its experiences eventually beating many game masters.

Natural Language Processing is another form of AI related technology. This technology take human generated text and renders it in machine form. For example, it takes human generated text, detects nouns such as people, places, things, relationships between those entities, detects sentiments and emotions in that text, extracts certain keywords from the text, categorizes information, and so on. One can build such capabilities by either programming all the rules in software or using machine learning.

These are only some technologies related to AI. So, depending on your specific use cases, your applications may use ML, DL, along with other technologies such as image recognition, Natural Language Processing, Robotics, and others to give you a full-fledged AI system.

 

What are the specific use cases for AI?

 

Next, we will address the third and final topic of this episode, which is to review the specific use cases of AI to give us an indication of how this technology can benefit organizations. So, here are some of the ways that AI can benefit organizations.

• AI is used in automation of mundane and processing intensive tasks – Although software has helped us to automate a number of tasks already, AI technologies help us automate more complex tasks and ones that are processing intensive. So, we see AI helping organizations and the society at large in anything that is data intensive or time and processing intensive.
• AI is used to perform advanced analytics to analyze lots of data, detecting patterns and trends, and extraction of predictive insights. Using ML and DL, for example, AI driven analytics provides insights related to improving a business’s operations, analyzing problem root causes, and much more. This technology, for example, is being used in the medical imaging field where AI trained systems can scan the data of millions of CT scans and provide diagnosis.
• AI programs are also being used to understand natural language and to extract meaning from that language. For example, they are being used in image recognition where AI programs mimic human capabilities in terms of recognizing images, informed decision-making, deductive reasoning, and inferences. And so on.
• One of the most common examples of AI use cases is that in customer service. We see numerous organizations started to employ AI assisted systems to improve an organization’s customer service operations. A business dealing with hundreds of thousands inquiries can be quite overwhelming for any business. Not only an organization needs more agents to handle these inquiries, the agents have to be trained on the various types of calls that flow in through online channels or call centers. So, as a business grows, this problem gradually gets even worse. This is therefore an ideal problem for AI systems to solve. An organization, for example, can train AI systems with various customer inquiries, be they in Email format, chat, or other. As the system learns, they can gradually get better to handle a large percentage of those calls, especially those that have simple and straightforward answers thus freeing up agents from being busy with repetitive rote tasks and instead to focus on handling more complex customer inquiries and problems. These AI systems can also be used by customer service agents to research better by demanding answers based on the large historical inquiries and other types of data that these systems may have access to. So, the process to respond to customer’s even more complex inquiries can get a major boost. When customers get quick and relevant answers to their inquiries, this can lead to increased customer satisfaction.
• As another use case, manufacturing organizations use AI systems to help them improve their manufacturing operations. As more data fuels the use and success of AI systems, the manufacturing organizations have plenty of it. They can use data from their production histories, and couple it with ongoing data feeds and then use various AI technologies to try to anticipate production problems and to prevent them to ensure a smooth flow of production.
• We also see AI use cases being used in financial institutions and banks. For example, these organizations are making use of speech recognition and working with Amazon’s Alexa tool to allow customers to talk to Alexa and conduct various banking transactions. So, Alexa understands a customer’s voice and requests and thus, a natural dialog between a human and a machine allows a customer to carry out various banking transactions. A bank can also provide personalized financial services to all of its customers based on their individual profiles, transaction history, and other relevant details.
• Amazon Go from Amazon is another great example of AI. As we know, Amazon Go are physical stores recently opened by Amazon. Amazon has equipped these stores with state of the art AI technologies, which enables customers to shop and checkout without having an in-store cashier or self-checkout station. Customers shopping at Amazon Go stores download an app on their mobile phones and this app along with computer vision, sensors, and deep learning algorithms and software enables customers to navigate the store, shop the items, and essentially walk out of the store without checking out.
• We are seeing AI being used in the area of new product design as well. For example, when an AI enabled system is provided enough inputs and models, it can generate enough simulations and recommend multiple product designs from which engineers can choose from. Watson, an AI system from IBM provides inspiration for new songs and even works of art when it’s fed by millions of pieces of music and art. So, just with this we can see that product innovation can accelerate in organizations, which employ AI systems.

These were only some of the use cases related to AI but the point was to provide an idea on how AI can dramatically help an organization advance its goals. Deciding where to start applying is a major strategic decision. We will review those strategic decisions in another episode on CIOtechCentral’s podcast.

Summary and conclusion

So, here are a key points that I want you to take away from today’s episode.

• First, as we discussed, we are already at a point where organizations are experiencing manifold increases in productivity and cost savings in their business operations using AI and its technologies. Many are also using AI in the creation of new products and services giving them a major boost in increasing their revenues. So, if your organization is not actively pursuing an AI strategy, it should get on it right away.
• Second, as we have seen, with enough data, processing power and sophistication of AI algorithms, AI will be able to completely transform the way business is conducted. So, organizations need to start looking at organizing their data and build scalable digital platforms that can make all of this work together. Every organization, large or small, therefore must have a sound data and technology strategy. One thing we should realize is that a sound data strategy is also needed because the digital world with sensors, cameras, and other IoT devices, APIs interfacing with other computing ecosystems, and more is creating lots and lots of data. And all that data eventually must be harnessed, processed, and used to advance the business and its goals.
• Finally, as AI can have different types of uses for different organizations, it’s important to have some type of an AI strategy before jumping in blindly. While it may be tempting to jump in to grab some low hanging fruit, it’s essential to develop a longer-term strategy. We will cover that in one of our next episodes in the days to come.