Learning AI Series - Part I: Demystifying Artificial Intelligence (AI)

Introduction

Artificial intelligence (AI) is on the tip of everyone’s tongue these days. It seems like you can’t go anywhere without AI entering the conversation, and for good reason. When innovation becomes more accessible to the general population with the promise of making life easier and work more productive, we grab on to it hoping we can reap the benefits of that innovation (professionally and personally). And with vendors scrambling to inject AI into their offerings, it becomes more enticing to give it a go. So, before we get into all the excitement about AI, let's briefly describe this fascinating topic and provide a foundational understanding of what AI means.

The intent of this first article in the series is to help create a baseline understanding of AI for the masses (techies and non-techies alike), and attempt to uncover a shared understanding of what AI means. Then this AI series will build upon that foundation to explore common techniques, ethical considerations, and how to get started with AI. This article will not recommend specific tools to remain vendor neutral. There are far too many factors to consider when recommending tools, and it would be irresponsible to do so purely in the context of this introductory series.

So, here we go...

What is artificial intelligence?

An Internet search will quickly yield lots of definitions and information, potentially leaving you more confused than when you started. Here is my single-sentence definition in layman’s terms that is an amalgamation of sorts, including my experience building models:

“The art and science of using computers to mimic the actions or behaviors of human beings.” – Mark DeRosa

That’s it in its absolute simplest form. Essentially, AI is using automation to perform tasks/work that humans normally perform. And if you are looking for a longer definition from a more authoritative source, I believe Google’s definition of AI may fit the bill:

“Artificial intelligence is a broad field, which refers to the use of technologies to build machines and computers that have the ability to mimic cognitive functions associated with human intelligence, such as being able to see, understand, and respond to spoken or written language, analyze data, make recommendations, and more.” – Google

Please note that I included “art” in my definition because I firmly believe it’s far more than the bits and bytes of the tech and data; it’s the creative application of technology in ways that really embrace the cognitive functions (psychology) and mannerisms (actions/behaviors) of humans.

Okay, so what is machine learning then?

Machine learning (ML) is a branch of AI that means computers (or more specifically, algorithms) are learning and improving from patterns in the data. The more data you have, the better the model performs (i.e., generating output with greater accuracy). The easiest way to understand the distinction between AI and ML is that AI is the ability for machines to perform human-like tasks while ML is the continuous improvement of those abilities to become better and better at doing those tasks.

Cool, but what about the robotic stuff?

Robotic process automation (RPA) is essentially a rules-based approach to automating repetitive tasks. It is narrowly focused on performing a routine set of steps to complete a task (or series of tasks). They usually come in the form of bots. Bots can be attended or unattended meaning that they are either activated by humans (attended) or independently run based on certain conditions being met (unattended). You can think of an attended bot like a macro that you run and an unattended bot like a database trigger waiting to fire when something happens.

RPA bots are perfect for very prescriptive processes that are routine and predictable (or at least semi-predictable). For example, if a person manually downloads a file, copies that file, opens that file, adds/removes/updates data in that file, uploads that revised file, and does so in a very routine manner everyday … create a bot!

RPA is loosely considered a part of AI but generally separate because RPA does not rely on AI or ML. And that leads to another term called hyperautomation (a.k.a. intelligent automation).

Tell me more about this 'hyperautomation' term

Generally speaking, hyperautomation refers to the integrated capabilities of AI, ML, and RPA (and other tools) that orchestrate entire business processes. Beyond the individual benefits of AI, ML, and RPA, think of the inter-connections across those capabilities that make end-to-end automation possible. Leveraging AI and ML with RPA can bring a suite of solutions only achievable through an integration of complementary features.

AI automates 'thinking', ML automates 'learning', RPA automates 'repetition', and hyperautomation increases speed and efficiency by using all of them (and more). Ultimately, this level of sophistication frees up people to perform the really complex work (requiring human judgement/intervention) where people rely on hyperautomation as an enabler to achieve much higher levels of efficiency.

Closing Thoughts

Although I did not (and will not) get into the history and evolution of AI here, that is also an important part of understanding AI. There are two reasons for not doing that here: 1) there is a wealth of information already available on the history of AI and 2) it is beyond the scope of this AI series of short articles.

I do have an entire presentation dedicated to that very topic and had the honor of delivering that presentation as a guest speaker at the National Science Foundation (NSF). Feel free to contact me if you are interested and we may be able to arrange presenting a custom-tailored version of that presentation to your organization.

Stay tuned for the next article in this series – Part II: Common Techniques Used in AI.