The questions raised in Beverly's post (or some modified versions of these same questions) are ones that I have heard many times throughout my career as a leader in OR and Analytics. They are not easy questions to answer and the question, "How do these things differ?" is even harder to explain in a few simple sentences. There is no perfect answer. I have answered them over the years along the following lines, based on my view of the different fields.
I will first add one more term to the discussion about Data Science, Machine Learning, and Artificial Intelligence: Analytics.
Analytics is a set of mathematical modeling methods and tools used to derive value and insight from data depending on the business problem at hand. They include different techniques that vary from the type of data that is being analyzed and the types of problems you wish to solve and are included in different disciplines and fields of study: Statistics, data mining, forecasting, optimization, simulation, quality control, multiple areas of operations research and others. These tools help you gain insight about the past and discover relationships (descriptive analytics); predict future outcomes and discover patterns based on available data (predictive analytics); and recommend optimal decisions (prescriptive analytics). As the volume and variety of data available continued to grow, several newer techniques continue to be added to this arsenal of tools to tackle different forms of data including text data, audio and video data, digital images, etc.
While Analytics typically refers to the tools for data-driven decision making, Data Science includes a deeper knowledge of the collection and management and storage of data as well as the complete life-cycle of deriving insight from data to solve a variety of business problems and continuously monitoring the incoming and changing data to adjust the models as needed. One can also think of Data Science as the science of data-driven decision making using the appropriate set of analytical tools from the complete arsenal of tools available to the modeler.
As data volumes continued to explode, computing power has also been growing at immense rates along with the business need for faster decisions for more complex problems requiring the automation of several of the analytical modeling tools used in data mining, predictive modeling, forecasting, text analytics, image recognition, etc. In simplest terms, Artificial Intelligence is the science of training machines or computers to perform human tasks automatically and at fast speeds. While there are multiple definitions of Artificial Intelligence, one that resonates with me is:
Artificial Intelligence is the science of "automated data-driven decision making at scale" which requires several enabling technologies and a framework for implementation. All the science and math that we do are just various arrows in our quiver, or tools in our toolbox: Forecasting, machine learning, optimization, predictive modeling techniques, natural language processing, image recognition, etc, etc. The distinguishing aspect of AI is the ability to do all the processing at super-fast speeds as well as tackle large volumes of data that was not possible before. This ability helps us make advances like automated decisioning for self-driving cars, classify thousands of images enabling quicker diagnoses of skin cancer and other diseases, create the best predictive models for hundreds of micro-segments and many other tasks that benefit from super-fast processing of large amounts of data. Note that even though we have heard the term artificial intelligence for several decades, AI has become a household word today because of the perfect trifecta: availability of massive amounts of data, advanced and innovative algorithms and tremendous advances in computing power and storage.
Machine learning (ML) includes a class of powerful analytical modeling tools that typically "learn" from data to determine patterns, predict future trends, extrapolate to other members of similar populations, etc. Most ML methods can do this with multiple types of data including images, audio and video files, text data, etc. Quite often ML techniques are also used to augment the insight / value derived from other analytical modeling approaches. These tools are an essential part of the "magic" that makes artificial intelligence systems work!
Do these definitions and descriptions seem to be very similar and are you wondering what is the difference? You will often see many of these terms used interchangeably. The reality is that there is a great deal of overlap in their definitions. I like to think of these four fields in the following way: Analytics and Machine Learning refer to different types of tools that derive insight from data; Data Science uses these tools in addition to a framework of implementation to make and deploy data-driven decisions; Artificial Intelligence is highly scalable automated data-driven decision making that uses an arsenal of appropriate modeling tools and large volumes of data to mimic human behavior.
For a good overview of these different areas and related examples, please see the following pages:
https://www.sas.com/en_us/insights/analytics/what-is-analytics.html
https://www.sas.com/en_us/insights/analytics/data-science.html
https://www.sas.com/en_us/insights/analytics/machine-learning.html
https://www.sas.com/en_us/insights/analytics/what-is-artificial-intelligence.html
------------------------------
Radhika Kulkarni
Vice President, Advanced Analytics R&D (retd.)
SAS Institute Inc. (retired)
Durham NC
------------------------------
Original Message:
Sent: 11-29-2022 15:29
From: Beverly Wright
Subject: Explaining our field to laypeople
Hi all, INFORMS asked me to help create content to help explain basic concepts of the data science field to a non-INFORMS audience. Questions like: What is data science? What is machine learning? What is artificial intelligence? (And how do these things differ?) I thought it would be helpful to crowdsource some ideas on this. What are the right questions to answer? How do you explain these things to someone who has no background in them? Think on the level of an undergraduate or maybe someone planning to go to college soon, who is considering potential paths, but hasn't necessarily taken advanced math or statistics courses yet. I appreciate your input on how you would describe the data science field!
------------------------------
bw
Beverly Wright, PhD, CAP® (Certified Analytics Professional)
Head - Data Science Solutions, Burtch Works
Chair, Analytics Society, INFORMS
Chair, Data Science & Analytics - FOR GOOD Initiative, Technology Association of Georgia
Academic Director - Data Science, University of Georgia
Podcast Host, TAG Data Talk
C: 770 823 2396
E: beverlywrightphd@gmail.com
------------------------------