Table of Contents
AI is a branch of computer science and an endeavor to simulate or replicate human intelligence in machines. What makes things a bit complicated is the fact that there is no universally-accepted definition for the term AI. The problem with the simple definition of ‘intelligent machines’ is that it does not clearly explain intelligence and what makes machines intelligent.
Instead of focusing on definitions of AI, the focus of this chapter is explaining some basic concepts. Different types of AI, key differences between subsets of AI, and its applications have already been covered in the previous chapters. This chapter includes main areas where advances are being made, future prospects, current trends, and career prospects in the field of AI.
AI can be applied to almost any industry to improve efficiency and enable new possibilities. It helps humans tackle technical challenges by mimicking human capabilities, including understanding, planning, reasoning, perception, and communication[i]. Since its inception, research in the field of AI has remained focused on five key areas: knowledge, reasoning, planning, communication, and perception. These are also the areas involved in most business applications and processes, so advances in these areas can unlock new possibilities and capabilities.
Knowledge in the context of AI refers to the ability of machines to understand and represent information about the world around us. Knowledge includes facts, certain entities, situations, and relationships between different objects. AI-systems use many methods of learning, including trial-and-error in which the system keeps trying to solve a specific problem until it finds the right answer. The learning component is mainly responsible for ‘memorizing’ items and words, vocabulary, and possible solutions to problems.
Reasoning can be defined as the ability to apply logic to solve problems and derive beliefs, conclusions, and ideas from available information. Reasoning can be deductive, inductive, abductive, monotonic, or non-monotonic or be based on common sense. The art of reasoning is no longer limited to humans. Here are some brief explanations of the different types of reasoning:
Deductive: Deducting new information from information that is logically related. If premises are true, the conclusion of an argument must also be true. For example, all humans eat vegetables, John is a human, so John eats vegetables.
Inductive: Arriving to a conclusion through limited facts by generalizing. Specific facts are used to reach a general conclusion. For example, if all elephants in a zoo are grey, we can expect that all elephants are grey.
Abductive: Logical reasoning that starts with single/multiple observations based on which the most likely explanation is derived. For example, the computer needs power to start, there is no power, the computer won’t start.
Monotonic: Conclusions remain the same even if more information is added in the knowledge base. For example, planets revolve around the sun, the conclusion will remain the same even if we add more information using this reasoning method.
Non-monotonic reasoning: Used for uncertain or incomplete models in which conclusions may be invalidated if new information is added. For example, birds can fly, coco is a bird, fish cannot fly. If we add the statement ‘coco is a fish’, which will conclude that coco cannot fly and invalidate the information that coco is a bird.
Common sense: Refers to reasoning that is informal, gained through experiences, and stimulates how humans make presumptions about day-to-day events. It relies more on good judgement than logic. For example, a person can only be at one place at a time.
Intelligent machines need the ability to define and achieve desired goals, which involves clearly specifying the future or desired outcomes and the sequence of actions that will lead to those outcomes. Planning involves finding a course of action while optimizing overall performance to reach specified goals.
Communication requires understanding of written as well as spoken language, which AI systems must recognize, comprehend, and synthesize. Language is a set of well-defined signs that help differentiate between objects and ideas. AI-systems need understanding of commonly used words and expressions to be able to communicate in any language.
Perception refers to using sensory input to make logical inferences. AI-powered systems can be designed to identify, organize, and interpret input, including images, sounds, and other types of sensory inputs like touch and pressure. Perception enables AI systems to scan the environment, analyze objects and scenes to derive conclusions, and establish relationships between different objects. Perception is an essential component of many AI systems including autonomous vehicles and robots.
Capability-rich and agile programming languages are required to code AI applications[ii]. Each language has its own advantages and shortcomings, so programmers have to choose the ones that work best according to their project’s required functionalities.
C++ comes with powerful programming tools and library functions, which makes it a great candidate for resolving sophisticated AI problems and developing neural networks. Complete support for object-oriented principles allows developers to create powerful AI applications. C++ lags behind other languages when it comes to multitasking. One of the main reasons why new developers find it hard to code in C++ is because it favors the bottom-up approach. But the speed it delivers compensates for its shortcomings, making C++ faster and cost efficient.
Because it is relatively easy to understand, Java is one of the most popular and easy to implement programming languages. It is simple, multi-purpose, powerful, and makes debugging easier. Like Lisp, another language on this list, Java also comes with an auto memory manager, but it’s not as fast and execution is slower than other languages like C++. A wide range of open-source libraries makes Java a flexible language that can be used for a variety of AI projects.
Tree-based data structures give Prolog some distinct advantages, including quicker prototyping and higher efficiency. The language is still not fully standardized, while some features differ in implementation. This can increase the amount of coding and development work. The language is mostly used for building voice assistants, chatbots, and graphical user interfaces.
Python is a widely used programming language in NLP, ML, and neural networks. It comes with a wide range of libraries and tools, including nltk, SciPy, and Pandas. It supports object-oriented design and algorithm testing without implementation, which increases productivity. Developers like python because it requires less coding and has a simple syntax and can be integrated with languages like Java.
Python is considered to be faster than many other languages, including C++ and Java. Object-oriented programming, readable keywords, libraries and integration with programming languages make the development process faster. The python library PyBrain can be used for machine learning and Numpy for scientific computation.
Originally developed by John McCarthy in 1958 as a mathematical notation for programs, Lisp has a long history and was named after the LISt Processor. It’s fast and efficient, and supported by compilers. It allows developers to create new, dynamic objects and is popular for rapid prototyping. However, Lisp developers are a rarity these days and not many are acquainted with it to make a difference at scale. Lack of good libraries make it harder for developers to code everything from scratch.
Other Skills and Prerequisites
Programming isn’t the only skill required in making AI applications. AI is a multi-disciplinary field and also requires expertise from other fields, including mathematics, computer engineering, economics, philosophy, neuroscience, cybernetics and control theory, and linguistics. AI is a vast field and the prerequisites largely depend on individual roles.
Instead of covering prerequisites for each role, the focus will be on the most commonly used tools and required skills. Depending on the individual role and in addition to programming, other skills required in the field of AI include[iii]:
- Excellent grip on mathematics, including discrete mathematics, linear algebra, statistics, probability, and calculus
- Familiarity with Apache Spark (unified analytics engine) and other big data technologies like Cassandra, Hadoop, and MongoDB
- A deep understanding of neural network architectures, Bayesian networking and graphical modeling
- Understanding of cognitive science theory
- Understanding of algorithms, especially ML and deep learning algorithms
- Problem-solving, critical thinking, and communication skills
- Understanding of reinforcement learning (learning from the environment)
- And most importantly the patience and will to learn different languages and complex technologies, and keeping yourself updated with the latest trends
AI frameworks, tools, and templates can be great timesavers and allow developers to quickly get started and build prototypes in less time. Some of them are open-source and some are available for free including[iv]:
The open-source library was initially developed and used by researchers from Google’s Brain Team. The architecture allows developers to deploy computation in multiple devices including desktop PCs, servers and even mobile devices using a single API.
Made for Java programs, Neuroph is an open-source tool designed to help developers build and train artificial neural networks. It’s a GUI (Graphical User Interface) based tool, making it easier for developers to learn and start building components of neural networks.
The framework is designed to help AI engineers develop ML-capable systems using Big Data. Created by IBM, the framework is popular for its scalability and flexibility. It allows data and cluster-characteristics based optimization and algorithm customization and offers multiple execution nodes.
Backed by a large community of users, startups, and academic research, the framework was developed with a focus on speed and modularity. It is used for building deep learning systems. In technical terms it’s very expressive and focuses on computer vision networks. No hard coding is required to configure models using the tool, which accelerates the development process and deployment rates.
Based on LuaJIT coding language, the open-source ML library offers flexible tensors, a wide range of algorithms, and allows cloning, resizing, and sharing storage. Efficient GPU support, powerful interface, neural network models, and linear algebra routines make Torch a favorite framework for many tech giants, including Facebook, IBM, and Yandex. PyTorch, an open-source Python library and subset of Torch can be used for NLP.
Computer science specialists with the right skills are in high demand across a wide range of industries, thanks to the transformational reach of AI and machine learning. This global skill gap translates into highly paid jobs with bright future prospects. Industry trends show that public safety, fintech, banking, and healthcare sectors will offer growing opportunities in the near future. According to forecasts published by Forbes[v]:
- By 2023, spending on AI will reach up to $97.9 billion, while the machine learning market will reach $20.83 billion, which is a 44 percent increase on a year-on-year basis from 2017 to 2024
- Almost one in every 10 organizations is already using AI in one form or another, including fraud detection, chatbots, and process optimization
- 89 percent of IT leaders say AI and machine learning is transforming customer engagement, while 69 percent say these technologies are transforming their businesses as a whole
- The most popular use cases of machine learning include cutting costs (38 percent), generating intelligence and customer insights (37 percent) and improving customer experiences
- The biggest global owner of AI and ML patent families is IBM with 5,570 families
- World’s fastest growing AI platforms are Microsoft and SAS
- By 2022, the AI-related semiconductors market is expected to grow to over $30 billion, an 18 percent growth on a year-on-year basis, which is five times greater than semiconductors used in non-AI systems
- ML based systems will be the foundation of next-gen logistics and resource scheduling systems
- 71 percent businesses and organizations are spending more on ML for cybersecurity than before (2-3 years)
- Credit unions will automate routine activities using ML and free their human resources for delivering more personalized services
These statistics above are promising and underscore the need for AI and ML experts. Career opportunities in the field of AI and ML are increasing with the increase of possible applications across different sectors. Professionals with the right skills, education, and experience are hard to come by in this thriving industry. Although the AI’s job market is huge, not enough people are trained for it, creating a gap between demand and supply. The top five careers currently in huge demand are as follows.
AI research involves discovering new ways to advance AI and ML technologies. The term AI researchers generally refers to professionals who are pushing the edge of what is possible with this technology. With such a positive industry outlook, AI researchers are at the core of advances and even lower-level employees can expect to make something between $300k to $500K, while earnings of high-level researchers at the top can reach $1000K per year.
AI research scientists are mainly focused on applications of ML and machine intelligence, and are experts in statistics, applied mathematics, ML, and deep learning. They are expected to have at least a master’s degree and preferably a PhD degree in computer science or mathematics.
NLP is one of the fastest growing subfields of AI and powers some of the most popular applications ranging from virtual assistants to chatbots. These applications revolve around language and replicate human speech in other formats. The field requires professionals that are good at language and technology and have a deep understanding of both. Language is hard to deal with, so professionals with good NLP skills can expect above $100K per year.
Software engineers are needed in the field of AI to develop programs based on which AI tools work and are part of the overall development and design process. The key role of software engineers is to develop technical functionality of solutions that use AI and ML to carry out desired tasks. By 2029, the demand for software engineers is expected to grow[vi] at a rate of 22% with average salaries hovering above $110K per year. Software engineers who also have a specialty in AI can make more than that and are highly sought after.
Data Analysts/Big Data Engineer/BI Developer
Data is the heart of AI and ML, and resources are needed to perform data mining, interpretation, and cleaning. Qualified data analysts play an important role in AI processes and are responsible for sorting, analyzing, and managing data. They also have to communicate their findings to their managers using easy to comprehend visuals.
Average salaries of data analysts range from $60K to $80K per year. At least a bachelor’s degree in computer science of mathematics is usually required for this role with expertise in data management tools. Compared to data analysts, business intelligence developers are responsible for recognizing business trends using complex datasets and preparing, developing, and nurturing business intelligence (BI) solutions. Their main focus is on assisting in optimization of business workflows and processes.
User Experience (UX) Specialist
The role of a UX specialist revolves around working with products and making it easier for users to understand different functions. Although this role also exists outside the field of AI, it has gained more importance with the popularity of AI in different industries. UX specialists are responsible for understanding how customers use AI-enabled equipment and designing software that offer great functionality and user experience.
Apple is one of the most popular examples of how UX can influence technology. Compare iOS with Android and you’ll notice that it’s not just about coding and advanced technologies. User experience also matters a lot and customers are more likely to adopt user-friendly systems than more technologically advanced, but complex systems. With salaries ranging from $70K to 100K, UX specialists with the right experience and training have bright future prospects.
KPMG’s recent survey[vii] suggests that most of the Global 500 companies plan on increasing their investment on AI-related talent by 50 to 100 percent over the next three years. The top AI and ML trends include:
Increased Use of AI and ML
This trend should not come as a surprise considering the benefits AI and ML offer to businesses in key areas, including risk assessment, analysis, research and development, and cost cutting. More companies are embracing AI than ever before, including small and medium businesses and even individuals. New talent is needed to fill the gap between demand and availability of qualified resources.
Data Security and Regulations
Data is today’s new currency and a valuable resource. Businesses, government agencies and other organizations need to protect their data. Privacy violations have become a lot more expensive, thanks to regulations like California Consumer Privacy Act and GDPR. Demand for AI professionals will increase as global businesses come under more pressure to meet these regulations.
Despite all the advances, AI still suffers from issues related to trust and transparency. An increase in use of AI systems means businesses need to be more confident when making important decisions using machines, which they don’t usually understand very well. AI and machine learning professionals are needed to make these systems more transparent and explainable to businesses.
Augmented Intelligence combines the capabilities of humans and intelligent systems. AI-augmented automation requires specialized skills in AI and ML. Gartner[viii] predicts that over 40 percent of operations and infrastructure enterprise teams will use AI-augmented automation by 2023.
AI and IoT
The lines between AI and IoT (Internet of Things) are blurring and new terms like AIoT (Artificial Intelligence of Things) have started emerging. AI makes IoT devices more intelligent. Merger of these technologies is opening up unique opportunities for those who have the right skills.
AI has the potential of dramatically improving our lives by increasing efficiency and taking care of dangerous or repetitive tasks. This allows the human workforce to focus more on what they are better equipped for: creative tasks and empathy. AI comes with many challenges and learning experiences, but so does every other technology.
The impact of AI is visible in almost every industry, and our daily lives. Like most changes in life, AI will have both positive and negative impacts on the human race. What matters the most is how we create a balance between modern technologies and human values.
Knowing the future implications of AI accurately is difficult. But we should remain optimistic that the changes are going to be mostly for the good, and to our advantage. The future prospects of AI, machine learning, and deep learning are bright, and the same is true for those who are well equipped to enter an era of digital transformation.
[i] “The State of AI 2019: Divergence”. Retrieved from https://www.stateofai2019.com/chapter-2-why-is-ai-important/
[ii] “How Artificial Intelligence Works & How To Implement It”. Retrieved from https://www.digitalsilk.com/how-artificial-intelligence-works
[iii] “What Skills Do I Need to Get a Job in Artificial Intelligence?”. Retrieved from https://www.computersciencedegreehub.com/faq/skills-job-artificial-intelligence/
[iv] “Ten Popular Tools and Frameworks for Artificial Intelligence”. Retrieved from https://www.opensourceforu.com/2018/04/ten-popular-tools-frameworks-artificial-intelligence/
[v] “Roundup Of Machine Learning Forecasts And Market Estimates, 2020”. Retrieved from https://www.forbes.com/sites/louiscolumbus/2020/01/19/roundup-of-machine-learning-forecasts-and-market-estimates-2020/?sh=142cfdda5c02
[vi] “U.S Bureau of Labor Statistics – Occupational Outlook Handbook”. Retrieved from https://www.bls.gov/ooh/computer-and-information-technology/software-developers.htm
[vii] “AI Transforming the Enterprise – 8 Key AI Adoption Trends”. Retrieved from https://advisory.kpmg.us/content/dam/advisory/en/pdfs/2019/8-ai-trends-transforming-the-enterprise.pdf
[viii] “Gartner Predicts the Future of AI Technologies”. Retrieved from https://www.gartner.com/smarterwithgartner/gartner-predicts-the-future-of-ai-technologies/