AI Fundamentals: How to Define Artificial Intelligence

One of the biggest obstacles to understanding, agreeing upon & harnessing the potential of AI is in understanding it plainly. How to understand it one term at a time.

Apr 24, 2024

If we can’t speak plainly, clearly and understand one another, the task of using AI, or regulating its development will be rendered impossible.

Neuroscientist/policy researcher Joseph Keller wrote plainly about this issue recently. “It is imperative that we not only know what we are talking about regarding AI, but agree on how we talk about it,” he says.

It’s an issue that comes up a lot when I speak with families about AI and the impact it will have on their lives and those of their children: with no clear consensus on the lexicon, how do we move forward productively?

So instead of nodding our heads with a “note to self” to figure it all out “later” let’s dig into what we do know today…

Understanding is Doing

The short glossary below is an effort to help us keep focused on the basics. The ability to understand comes, of course, from doing. And from talking to our neighbors, families, co-workers, and friends.

This is a work-in-progress that I’ll update as needed. I add some of my definitions and editorializing of these ideas, but all words are hyperlinked to a more sophisticated definition as well. If you have any questions, ideas, or words you want to dig into, please let me know.

Artificial Intelligence (AI)

Artificial Intelligence (AI) is the “big tent” concept here. It is defined as technology that enables computers and machines to simulate/replicate human intelligence. AI is both complicated (i.e., in the sheer breadth of what is possible) and incredibly simple (i.e., we can all envision ways to “replicate” our thoughts and behavior).

I find it easiest to think of AI as a simple formula: data (mostly our data) + “math” (the computer algorithms used to process the data) + the computing power (chips, storage, energy) = “innovation” that can do what our brains do…solve problems, answer questions, analyze data and do rote activities.

If you think of the concept as “anything humans can do” in the broadest terms, you can see the breadth and scope of possibility with AI innovation. Here are some specific terms you might have come across already.

Artificial General Intelligence » This is used to refer to AI that is, essentially, smarter than us. We’re not there yet, but it’s a very important concept as it’s both ill-defined, but also considered the “holy grail” of where companies are off racing to.
Algorithm » This is the set of rules that a machine follows to learn how to do a task. They are specific to the company/organization/person developing AI innovation.
Deep Learning » This field is part of machine learning and focuses on the learning of complex patterns.
Generative AI » Exactly as it sounds, generative AI allows users to “create” something, image AI is a great example.
Machine Learning » This term is more intuitive than most: this is the field within AI where machines, well, learn. So, it’s the math behind AI.

Large Language Models (LLM)
Our day-to-day understanding of AI right now is generally LLMs. Again, it’s exactly as it seems: large amounts of sequential data, mostly text based, that is processed and turned into a response prompted by a user’s question.

Chatbot (also Conversational AI) » A computer program designed to simulate human conversation through text or voice interactions.
ChatGPT » While there are now several competitors, the fervor of the past year + came after the launch of ChatGPT.
Hallucinate/Hallucination » In a nutshell, this is when the answer a user receives is inaccurate despite the confidence in its response.

Affective Computing

Affective computing is also known as emotion AI (or emotional AI), is a technology that seeks to identify, process, and simulate our feelings. It’s a hybrid of computer science, psychology, and cognitive science. And, as mentioned above, while all the terms here are inter-related with AI, I’m isolating this area of computing for its interrelatedness to other fields and its implications in practice.

Sentiment analysis » The ability to process written data and “interpret” for a user’s “sentiment” toward an issue, topic, brand, etc. is not new per se. But now as AI advances we are talking about the overlay of voice (intonation), facial movements and other visual aspects to determine sentiment.

Data

Data, of course, underpins all of this. It is the “fuel” of AI. But we should be very specific and clear about how we talk about and understand data because of the profound impact it has (and will continue to have in larger measure) on our privacy.

Data Science » Data science is a field of study that uses statistics, computer science, and information science to analyze data and solve problems.

Big Data » These are large sets of data used for projects such as creating AI innovation.

Specific to AI, there are terms that often belie the greater impact on all of us in their simplicity. Because to create AI innovation means to “train” AI tools and systems. And that data comes from all of us.

Training Data » The dataset used to train a machine learning model.
Pre-Trained Data » Think of this like building blocks to that go into the AI model/tool that a developer wants to create. You’ll see this as an offering typically from enterprise businesses, such as Amazon Web Services.

How Others Define AI

For the past couple of years, I’ve been collecting links to pages where corporations, universities, and governments define AI. I find it a helpful resource and a great way to look at the bigger picture in how we approach understanding this space. Here is a list that I continue to update. It’s interesting context for the different ways in which AI is defined and communicated.