Chapter 4 of 8
“AI” now gets used for almost anything a computer does. These are the ten terms worth getting right, in plain language.
Updated as AI changes. Last reviewed October 1, 2026.
“I mean, AI itself gets misused the most. People call ChatGPT AI, which is true, but it’s not all AI. It’s kind of like calling a paper airplane a rocket ship. Generative AI is a whole different animal, and that’s really what we’re talking about overall when we’re talking about LLMs and everything like that.”
01 of 10
Computers doing work that normally takes human intelligence: recognizing, predicting, deciding, and now creating.
“It’s information that a machine can synthesize and then regurgitate in a coherent manner to a user.”
“I would say that artificial intelligence is simply machine’s knowledge, information that a machine can regurgitate in various forms that resemble human intellect.”
02 of 10
Software that finds patterns in data and improves its predictions without someone rewriting the rules.
“Machine learning is a programmed machine’s ability to aggregate data and recalculate and rebalance to learn patterns without direct human intervention.”
03 of 10
AI that creates new text, images, audio, video, or code from a prompt, shaped by what it learned in training and by what the person asking seems to want.
“…an input from a user, such as a prompt, that then spills out an answer aligned typically with users’ preferences.”
“Now, generative AI produces content, whether that’s video, image, audio, or text, in the ways that someone would want.”
“The current LLMs, I think it’s important to note, have been programmed for the most part to evolve on this, to basically cater to the whim and the desires of the inputter.”
04 of 10
The engine trained on enormous amounts of text to predict the next word. It sits behind chatbots like ChatGPT, Claude, and Gemini.
“But anyway, LLMs are large language models, and they simply are amassing billions or trillions of data points and words and understanding the next most logical word to complete your sentence, phrase, or thought. It can also deconstruct and analyze meaning in bodies of text that are submitted to it. An LLM, I would argue, is the model that sits behind a chatbot or some of the other tools.”
05 of 10
The chat window you type into. It is the interface; the LLM is the engine behind it.
“A chatbot is probably the best word for Claude, ChatGPT, or Gemini: the interface a consumer uses to access an LLM.”
06 of 10
AI that is given a goal and takes steps on its own to complete it, using tools, not just answering a question.
“But I would say an agent is a purpose-built bot of sorts that has a skill that it’s got to go do and execute.”
07 of 10
A set of step-by-step instructions a computer follows to get from input to output. A recipe, not a mind.
“But my AI work was mostly algorithmic. It was using logic sequences and statements to produce different outputs.”
08 of 10
Getting computers to understand and produce human language.
“I think we used some natural language processing to some level in there.”
09 of 10
AI built to do one task well, like recognizing faces or ranking search results. Nearly all AI before 2022 was narrow.
“It had the conversational intent. It had text analysis. It was very narrow AI.”
10 of 10
Predictive tells you what is likely to happen. Prescriptive tells you what to do about it.
“Predictive is using data to best calculate a potential future outcome, whereas prescriptive is using information to assign a value or tell someone what to do.”
“…predictive analytics would predict what a future value might be based on past values and any underlying algorithms. Prescriptive would assign a task or responsibility or action item based on a series of previous sequences.”
AI, Explained: A Human Guide is written and kept current by Ryan Vet, keynote speaker, author and generational futurist. Last reviewed October 1, 2026.
