Generative AI has been around for over a year, disrupting the public relations industry and making communicators wonder about the future of their work. People are uncertain, especially with all the unknowns that the technology brings with it.
However, this fear is preventing people from understanding artificial intelligence’s capabilities, leading people to feel they can’t prepare for the future. Unfortunately, many communicators lack the knowledge to accurately describe what this technology is, how it works and what it’s capable of, both in terms of the organizations they represent and in terms of their own general knowledge.
Therefore, I’ve written a short glossary of commonly used AI terms, in plain English, to enable any communicator to understand what these buzzwords mean and explain what’s going on.
Related: There’s a Fine Line Between Using and Over-Relying on AI. Here’s the Role Leaders Need to Play.
AI
AI is a technology that enables computers and machines to simulate human thinking and intelligence as well as human-level problem-solving.
It encompasses everything from self-driving cars to weather forecasting models, machine learning, robotics and much more. Each one of these examples is a “subset” of AI, and entire articles can be written on each one. However, given that this article is about generative AI, we’ll dive deep into the lexicon surrounding this type of artificial intelligence.
And to do that, we need to look at the “machine learning” subset of AI.
Machine learning
The purpose of machine learning, or “ML,” is to use algorithms that can learn and generalize information. In essence, a machine learning algorithm is given information. It is then asked a question, and the algorithm thinks up an answer based on the information it’s been given.
There are dozens of subsets within machine learning. These include “decision trees” which are used in chatbots. There is “linear regression,” which is useful for predicting what will happen in the future based on previous data like weather models. There’s also “clustering,” which is how an adtech algorithm knows when and how to sell you a product or service.
All these subsets take information that was fed into it to make predictions about the future based on past events. They are all useful and impact our daily lives. However, there’s another subset of machine learning called “deep learning.” This is the subset in which we find generative AI.
Deep learning
Deep learning means there are more than three layers of neural networks. “Neural networks” are the brain of the algorithm, while “layers” are the depth of thought an algorithm can do.
In standard machine learning, there is an input layer (i.e. What will the weather be like today?); a “thinking” layer, like taking all the wind, rain and temperature data from past events and applying it to the current situation; and then the output layer (i.e. the weather forecast will be sunny). All these layers make up the neural network.
With deep learning, there are more than three layers to the neural network. This enables the algorithm to think deeper and with more nuance. In fact, this deep vs. shallow way of thinking is where the phrases “deep AI” and “shallow AI” come from.
In addition, to a difference in the amount of layers in the algorithm, the way the information is fed into these algorithms is also distinct. This is because a deep learning algorithm is based on foundational models.
Related: How to Leverage AI to Supercharge Your Business
Foundational models
“Foundational models” are giant stores of data, with each data point being called a “parameter.” The deep learning models are trained on these foundational models full of data, then “fine-tuned” to operate in a particular manner. Some foundational models have over 1 trillion parameters.
There are several kinds of foundational models, including “Large Language Models” or “LLMs.” They’re called this because they’re large — they can have over a trillion parameters — and are meant for processing and generating normal, human language. Other foundational models include vision models for generating video, sound models for generating different types of sounds and even biological models to predict how proteins will interact with each other.
Foundational models are important because they are huge repositories of data that any paying subscriber can use. Instead of spending millions of dollars and thousands of hours compiling all of this data, a company can subscribe to an already existing model (such as OpenAI’s model or Google’s model) and use this information to train their generative AI.
AI application
These foundational models provide the foundation for “AI applications.” The application itself can be anything from a piece of a platform to a full-blown application that fine-tunes a foundational model to be used in a certain way. A good analogy for an AI application is looking at how apps in general are built.
If you look at an app on the Apple Store or Google Play, that app was built to be able to work on the foundational tech infrastructure of that particular app store. AI applications work on the same idea — they are built to work with the foundational technological infrastructure of the AI model.
Related: How to Leverage Artificial Intelligence in Public Relations
So where does generative AI fit in?
“Generative AI” includes models that are specifically crafted to generate new content. It’s what is created using the knowledge base of the foundational models coupled with the fine-tuning coming from an AI application to get a desired outcome. That is how video generators such as Sora or language generators such as Perplexity or ChatGPT work.
In short, generative AI is used in AI applications that use deep learning neural networks trained on foundational models to generate a particular, never-before-seen piece of content.
It’s important for us as communicators to fully understand these AI terms so we can enable the public to understand how this world-changing tech works. Hopefully, PR professionals will be able to use this glossary to better communicate what AI is, as well as have a better understanding of how it can be implemented into their daily lives.
