What AI terms really mean

Businesses have used artificial intelligence in their everyday workflows since the early 1990s with the creation of spam filters for email. But in recent years, generative AI – or AI that creates text, images or codes rather than simply sorting data – has exploded.

Change can be scary, admitted Rebecca Simons, communications manager at Cisco. She noted that conversations around AI often feel like they “need a degree in computer science to understand.”

One way to overcome any fears about AI is through education. Simons led a recent webinar during Ragan’s AI Virtual Conference for Communications, aiming to provide a clearer understanding of common AI concepts and their applications.

“The whole idea here is we’re going to work together to demystify AI jargon,” she said.

Here are some of the key points she covered:

• Deep learning: A form of machine learning system trained to make decisions like the human brain by recognizing patterns and learning from large data sets. YouTube and Spotify, for instance, use it to analyze user preferences to make personalized suggestions. “It’s basically analyzing what you are doing and then trying to provide you with recommendations based on your behavior,” Simons said.
• Large language models (LLMs): A deep learning model trained on massive data sets to understand and generate human language in a coherent way. For instance, Grammarly uses an LLM to analyze text for grammar, tone and clarity, providing advanced suggestions to improve writing quality.
• Natural language processing (NLP): Allows machines to understand, interpret and generate human language. For example, Google Translate uses NLP to analyze the structure and meaning of words and phrases to accurately translate text.
• Hallucinations: This occurs when an AI generates plausible sounding but factually incorrect information or content, like a hand with six fingers. She provided an example of an attorney who nearly got disbarred for using AI to produce case law citations. When it couldn’t find an example, it made one up. “Basically, it doesn’t have the data or information, so it’s trying to piece together a story to generate content,” Simons said. She believes AI tools are valuable for brainstorming and drafts, but the content will likely require diligent fact-checking.
• Agentic AI: A way of combining automation with the creative abilities of an LLM. Users can create a system that provides the LLM with access to external tools and algorithms that supply instructions for how the AI agents should use them. Simons believes agentic AI will significantly influence comms in the next wave by helping with “strategic brand reputation management driven by personalized user preferences.”

The best way to learn, Simons said, is by picking up the technology and trying it out. She recommended doing it at home and getting comfortable with it before moving toward any professional applications.

“I would say the most important part is really like getting in there, giving it a try,” Simons said. “It’s just a trial-and-error situation.”

Watch the full video below.