Explainable Artificial Intelligence Definitions. Filtering the Science from Fiction.

Hungry to know more about AI? Me too. There’s a lot of buzz going on about it. But before I bite off more than I can chew, I decided first to prepare a glossary of some simple artificial intelligence terms. Formatted in plain language. Making it easier to digest for the tech-curious like myself.
People are talking about AI. But, what’s it all about though? Where to even start? I’ve discovered, the more I try to learn, the more complicated it gets. Once I start to get to grips with one definition, it branches out into a dozen other sub-topics.
Question is, how deep do you want to, even need to go? Best to answer this before going any further. Without a clear direction, it can get overwhelming real quick.
For me, I decided to take the external stakeholder perspective. I want to know more. But I’m no developer. I want to know what it is and what do I need to understand to be able to use it effectively.
First the basics. Filter the science from the fiction. There are plenty of misconceptions revolving around artificial intelligence. Pop-culture AI has captured our imaginations over the ages. As a result, many of us have this version of AI set in our minds.

What is Artificial Intelligence (AI)?
Might be a no-brainer to start with. But there still seems to be quite a bit of confusion about what it is and what it isn’t.
Production lines have been increasingly machine automated since the industrial revolution. Opening the doors to faster and more accurate outputs. More than anything a human worker could ever achieve. But automating with machines have always been limited to simple processes with predictable repeating patterns.
Essentially today, AI deals largely with expanding the capabilities of machine automation. Developing algorithms able to adapt itself as the scope of the function changes. Machine learning involves algorithms programmed to continuously add different variables to its scope of functions by learning or being taught.
AI is a tool. One that is becoming increasingly relevant. For example, in the digital age of big data. The daily volume of data the average professional deals with is increasing exponentially. Repetitively organizing endless streams of data is not a very productive use of the human mind. AI may help us solve this.
AI isn’t sentient, nor self-aware. It’s not here to take over the world. Ask Siri the meaning of life. The AI processes input and based on pre-programmed or ‘trained’ actions, an algorithm defines an appropriate response. Not as a result of deep philosophical contemplation. Merely because it was programmed as such.
At Talkwalker, AI is implemented for our customers so they can spend less time managing raw data and more time into maximizing analytics ROI. Reducing the strain of repetitive tasks. Focusing their energies more on formulating insights and optimizing strategies.
Here are a few outtakes. You can find more in my article at Talkwalker here: Artificial Intelligence Definitions, Upgrade your AI IQ.
Explainable AI definitions for starters
AI Science
Abductive reasoning
AI analyzes a statement or situation and finds the simplest and most likely explanation for it, based on what it has been taught.
Example: You notice that the roads are wet. The most likely hypothesis is… rain. It’s the obvious explanation.
Algorithm
A data set of sequences defining instructions or code of actions to be performed. Required actions can be rudimentary, to begin with, but increasingly complex over time.
Backpropagation
The process typically applied to train deep neural networks and refine the results of artificial neural networks. Information is processed by the system and results are sent back up the pipeline in reverse order for verification.
Deep learning
The furthest evolution of AI at present. It learns by example and uses multiple layers of nonlinear processing units to achieve phenomenal results. It requires a lot of computer processing power, and large amounts of labeled data, to comprehend the task at hand. But, it can achieve the highest levels of data accuracy.
Explainable AI (XAI)
Yes, there’s a term for that too. It deals with the theory that as AI learns and becomes more advanced, there should be steps taken to keep processes transparent. For example, so humans can still understand what leads to decisions the AI makes on its own.
The furthest evolution of AI at present. It learns by example and uses multiple layers of nonlinear processing units to achieve phenomenal results. It requires a lot of computer processing power, and large amounts of labeled data, to comprehend the task at hand. But, it can achieve the highest levels of data accuracy.
John McCarthy
A computer scientist at Stanford, credited with being one of the founders of AI as a field of study and for coining the term ‘artificial intelligence’. He developed the language commonly used in artificial intelligence programming, Lisp.
Reinforcement learning
Reinforcement learning is a type of machine learning with less specific goals. Instead, the goals are more abstract, such as “maximize brand mentions.” During training, the AI learns by acting towards the goal and evaluating its contribution after each effort.
Sentiment analysis
Combination of natural language processing (NLP), computational linguistics, and text analytics. Applied together to identify and extract subjective information from content. Its goal is to understand the attitude of a person.
Supervised/unsupervised learning
Supervised and unsupervised learning are two different AI training methods. Supervised training, includes labeled data sets. This allows the artificial intelligence to learn from the expected labels, and extrapolate that into wider data sets.
Unsupervised learning requires no labeling, and it is up to the AI to self-categorize its output. While unsupervised learning can perform more complex learning functions, it can also create unnecessary or complicated categories of data. Giving more clutter rather than less.
Turing test
Developed by Alan Turing in the 1950s. The Turing test was devised to see if people could distinguish between a machine or human interactions. The standard interpretation is to have an interviewer blind-question both a human and computer subject. Then see if the interviewer can accurately predict which subject is the AI based on their results.
AI Fiction
Bicentennial Man
Bicentennial Man is a 1999 science-fiction film based on an Isaac Asimov short story. Starring Robin Williams & Sam Neil, and directed by Christopher Columbus. The story of a robot butler that becomes self-aware, and its heartwarming journey to becoming more human.
Isaac Asimov
Influential science fiction author and biochemist from the United States. Asimov created the ‘Three Laws of Robotics’, a set of basic ethics dictating that a robot:
- May not injure a human being, or, through inaction, allow a human being to come to harm;
- Must obey the orders given it by human beings except where such orders would conflict with the First Law; and
- Must protect its own existence as long as such protection does not conflict with the First or Second Laws.
The laws remain a point of reference in modern science fiction. Often used in academic discussions related to the potential ramifications of robotics and artificial intelligence.
R2-D2
Affectionately referred to as Artoo Detoo. Iconic robot or more precisely ‘Astromech droid’ from the Star Wars series. Main functions included emergency starship repairs, emotional support, and comedic relief.
What’s next for artificial intelligence?
Well, that’s not all of it. Not by far. There are new ones being defined every day. Check out my article at Talkwalker for more definitions!
The field itself is constantly expanding and evolving. The potential applications and ramifications are up for debate.
Best bet is to get educated, at least on the basics, sooner rather than later.
Were my definitions helpful? Comments and feedback are always welcome and appreciated.
