Data Science, AI And Machine Learning For C-Suite, Business Leaders, And Non-Technical Executives

Part 1: Fundamentals

I would like to provide a summarized go-to resource for business stakeholders and executives who do not have time for – or do not need to know - the technical details of Data Science (DS), Artificial Intelligence (AI) and Machine Learning (ML), but are (or will be) involved in projects that have elements of the mentioned concepts. These fields are growing extremely fast, and very soon all business professionals will need to have some level of understanding about them and know the important considerations about the related business cases, project lifecycle, and strategies.  

To understand the differences between the mentioned 3 terms (DS, AI & ML), we need to know the fundamentals of the concepts, their definitions, and identify what each includes.

Data science is about methodologies, tools (statistical, algebra, programming, etc.), processes, and techniques for understanding data to drive insights and values from data sets. It’s an interdisciplinary field that uses computer science, statistics, data analysis and visualization methods, and can be based on structured or unstructured data to create data products. Data Science has become an important field - because of the extreme rise to data - to provide predictable outcome and anticipate changes. Data science has an overlap with artificial intelligence, however, DS is not a subset of AI. In simple words, DS is extracting data from a big data source to answer a question, then processing, analyzing and visualizing it (understanding the data) to give meaning to data for business decision-makers.

Example: DS can gather, analyze, process and visualize data related to banking customers, and help the banks in customer segmentation, credit risk modelling or risk analytics.

Four main activities of data analysis are, stating the question and refining it, exploring the data, building statistical models, and interpreting and communicating the results.

Data Analysis Checklist:
 

•    Define the question
•    Set expectations
•    Read your data - Collect data
•    Compare the data to expectations, and if required, revise one of them, so they match
•    Validate with an external data source
•    Run a plot and visualize your data
•    First, try an easy solution
•    Follow-up (Do we have the right question and the right data?)

AI is about the development of computer systems that have a type/level of intelligence (mostly focused on problem-solving) as well as learning capabilities. It’s making machines and computers intelligent to function appropriately and apply the knowledge received from data in the environment. AI examines data points to find trends, predicts patterns, creates intelligence in computers through learning and self-correction, and performs tasks which normally require human intelligence.

Example: Siri (Apple’s friendly personal Assistant) is an AI system that interacts with its user and provide information and direction, or help us to send messages.

The word intelligence based on Merriam-Webster dictionary is defined as “the ability to learn or understand or to deal with new or trying situations”.
Intelligence can be categorized into systems that:

•    Respond to changing circumstances - example: aviation systems
•    Recognize various types of objects, and take actions - example: simple robots
•    Answer by “understanding” questions - example: robo advisor
•    Solve problems by gathering information – example: advanced robots
•    Focus on worker augmentation (AI is not always about worker “replacement”)

Machine Learning doesn’t equal artificial intelligence. It’s a subset of AI and a field of computer science that gives computers the ability to learn from data and environment (without being explicitly programmed) which improves accuracy, recommendations, and efficiencies. ML includes the development of enabling algorithms that enable computers to evolve human-like behaviours (based on the available data and knowledge base) to identify patterns and relationships in data and make data-driven predictions or decisions (based on previously unanalyzed data). It can also determine how to effectively group people. In traditional programming, the inputs are data and the program, to produce the output, however, in ML, the computer develops/improves the program, using the data and the output. To make it simplified, ML is making computers to program itself by analyzing feedback from the results. As humans/ environment interact with AI-enabled computers, they teach the computer. The more data from the interactions, the smarter the AI algorithms become. ML builds up over its past experiences and improves itself from experience.

Example: Recommendation systems help human find what he/she is looking for. A good example is Netflix. Netflix is suggesting (“recommending”) movies to its clients, based on their watch pattern.

Machine Learning and Traditional Statistical Analysis have some overlaps, however, let’s use an example to show the difference:

 



Machine learning has 3 categories: Supervised Learning, Unsupervised Learning, and Reinforcement Learning.

 

To understand Deep Learning, first we need to know what Artificial Neural Networks is! An Artificial Neural Network is a method in machine learning that includes a network of nodes or neurons, similar to the human brain and neural system. But unlike the brain, neural networks have layers that direct the data flow.

Deep Learning is a subset of the ML that drives by neural networks. It works by adding many layers of neural networks (having many steps/layers is the reason why it’s called “Deep”) and running data through the system to train the computer. The output of each step is the input for the next step continuously to get a final result. It needs very powerful computers. Example: Systems that learn from the human voice and enable the computer to speak like humans.

To read the other parts of this article, please click on

• Strategy
• The Hub
• Business Case
• Project Life Cycle
• Change Management and Communication