The Sensible Woman’s guide to AI has covered a lot of ground. Here, I make a list of Gen AI basic concepts…
This is the twenty-third edition of the Sensible Woman’s Guide to AI and Content Creation.
I’ve covered a lot of ground since I started this in December 2024 – there’ve been posts about the impact of AI on jobs, about how to use specific tools or how to make a virtual coach, or about “hallucinations”. And a lot more (you can see them all here).
But I’ve been wondering if it’s time to take a step back and go back to basics. There are new subscribers to the newsletter version of this blog who might have missed some of the foundational concepts. And a training course that my colleague Anne Taylor and I have been running has reminded me that there are people who are curious about AI and who don’t know where to start.
So I thought I’d make a list of basic concepts (I do love making a list.)
How I made the list (with help from AI)
I have all the Sensible Woman posts in a notebook in Google’s NotebookLM. I started by asking it to do this: “Please list all the foundational concepts covered in these posts that would be helpful to someone learning about Gen AI for the first time.”
It came up with a pretty good list, grouped in categories (you’ll note I didn’t ask it to do that, but hey – AI sometimes over-delivers!).
I’ve been through that list, edited and refined and rewritten it, and you can see the result below. Note that much of the information was researched at the time of the original post, where you will find the sources. I haven’t reproduced them all here for fear of inducing link fatigue in my readers.
Core definitions and technology
Artificial intelligence (AI): This is the ability of a computer or computer-controlled robot to perform tasks commonly associated with intelligent beings. Think of C-3PO, the golden droid in Star Wars who is fluent in millions of forms of communication (he’s the golden dude in the picture). One of my personal crusades is to spread the understanding that “AI” is a broad term for a wide range of tools and functions (from Netflix recommendations to Google Maps). It’s nowadays often used as shorthand for generative AI – and that shorthand means that people often don’t really know what AI is.
Generative AI (Gen AI): These are new tools, such as ChatGPT, that have boomed recently due to their ability to “talk” to people and assist with everyday tasks (like generating a recipe based on available ingredients). The underlying code is a “model”.
So what is an AI model? Tools and algorithms (see below) designed to train computers to process and analyse data similarly to the way that humans think. Models enable machines to learn from data and recognise patterns.
An important kind of model to know about: Large Language Models (LLMs) are the foundation for Gen AI tools like ChatGPT. Three things to take on board about LLMs:
LLMs function by analysing massive datasets of language (the Internet!), which allows them to recognise and interpret human language or other complex data.
LLMs are essentially prediction engines that operate by looking at preceding words and selecting the most probable next word based on their training data. They are not humans and have no knowledge of the external world.
Crucially, LLMs are trained to produce plausible text, not necessarily factual or true statements.
Algorithm: A structured set of instructions designed to perform a specific task or solve a particular problem (in the same way that a recipe gives you the steps needed to bake something).
Token: The smallest unit into which text data is broken down for an AI model to process (like breaking sentences into words or characters). The cost of using a generative AI model often depends on the number of tokens processed, and LLMs have a fixed limit on how many tokens they can handle at once.
Prompt: This is jargon. It’s just a request. When you say: “ChatGPT, give me a recipe for a tin of beans and a tin of tomatoes” you are “prompting” the tool to give you an answer. (It might feel the same as what you do when you put a query into a search engine, but the result is different. When you ask Google for a recipe, you get a series of links and you decide what you want to look at. When you ask ChatGPT, it synthesises a recipe for you from the data it has seen.)
Artificial General Intelligence (AGI): A buzzword you’ll see mentioned a lot in articles about AI. AGI is a theoretical type of AI capable of performing cognitively demanding tasks as well as humans can (sometimes called human-level intelligence AI). There is much speculation as to when AGI will be achieved (as of October 2025, we ain’t there yet). It’s not necessarily the thing that means machines will take over the world. That’s Artificial Superintelligence (ASI), which would outperform humans across every domain.
Types of AI tools
AI agent: A software program (bot, if you like) designed to understand its environment, process information, and take independent actions to achieve specific goals (e.g., an automated scheduling assistant, or an AI tool that can find you a hotel for your holiday and then make the booking). An agent can do what it’s name implies: take action.
Agentic AI: A system based on AI agents that can learn, reason, make independent decisions, and take goal-driven actions. Multiple agents can work together to handle complex workflows, such as planning a dinner party (including sourcing recipes, shopping, emailing guests and managing dietary constraints).
Custom GPT/Gem: Customised versions of foundational models (like ChatGPT or Google Gemini) that are created for a specific purpose by inputting unique instructions, extra knowledge, and skills. These tools use “system instructions” that remain consistent to define the AI’s role, context, and behavioural guidelines. That means you can assume the AI “knows” what you want and you don’t have to make detailed requests every time you talk to it.
Essential usage skills
Prompting vs. searching: A repeat of something higher up the list. AI tools like LLMs are not search engines; they process data they have “seen” to produce a synthesis, doing the filtering for the user.
Prompting mindset: Give the AI tool a clear set of instructions – think about it in the same way you would as talking to a new staff member or intern.
Management: Effective prompting aligns closely with management skills: clearly understanding the task and the information needed, explaining the task to the AI, giving useful feedback to improve outputs, and applying lessons learned into a process.
Elements of a good prompt: A prompt should ideally contain six elements, though not all are required: Task (action verb + goal), Context (background info), Examples (to guide the AI), Persona (who you want the AI to be), Format (desired structure), and Tone. Task and context are the most important elements of a good prompt. (You can also tell an AI what you don’t want it to do. Say: check this text for grammar and spelling but don’t rewrite it.)
Iterative prompting: Continually working with the AI, refining prompts, and asking follow-up questions to clarify thinking is encouraged.
Key risks and a critical mindset
Bias and discrimination: AI models learn from data, and if this data is non-diverse or unrepresentative, the models can perpetuate existing inequalities or replicate biases and nastinesses from the real world. It’s key to be aware of your own biases so that you can critically evaluate the answers you get from AI.
AI hallucination/bullshitting: This is the AI getting things wrong. Since LLMs lack an inherent understanding of truth, their output should be approached with sceptical curiosity or even distrust, as they are capable of generating convincing falsehoods. I think the term bullshitting is more accurate than hallucination – we’re looking here at speech intended to persuade without regard for truth.
Critical evaluation (checking facts): If a Gen AI tool provides a reference, fact, quote, time, or place, you are responsible for checking it against your own knowledge or against reliable sources. (It follows that you should stay in your lane when working with Gen AI. Don’t ask it to tell you how to perform heart surgery unless you are in fact a doctor.)
Privacy and data usage: Uploading information to an AI tool is risky because Gen AI systems are opaque (“black boxes”). You don’t know what data is collected, or how it is processed, and shared, or how your own input is being used to train models. Once data is entered, it is difficult or impossible to remove. Never share passwords, highly sensitive secrets, or proprietary company information with chatbots.
The tool-user mindset: Think of yourself (in relation to Gen AI) as a tool-user, maker, or craftsperson. Bring human skills and thought processes to bear.
Cognitive offload and laziness: Humans are “cognitive misers” who naturally try to expend as little mental effort as possible (cognitive offload). While AI can help manage information overload, reliance on it for basic tasks can lead to cognitive debt (underdeveloped mental faculties) and cognitive laziness, diminishing critical thinking and creativity. You make the choice when to use AI (for speed) and when to do the hard work yourself (for skill development).
Main picture: Christian Maass, Unsplash
Other things I have written
How to build yourself a virtual coach with AI (for free) – So you know how to use ChatGPT (or Claude, or Gemini). But you’re sure there’s more you could be doing. Here’s how to build yourself a virtual coach with AI (for free).
What about the jobs? Artificial intelligence and social responsibility – I use AI tools every day – and am starting to think about paying for one of them. Which leads me to wonder about artificial intelligence companies and social responsibility.
What you need to know about AI hallucinations (hint: they aren’t hallucinations) – When AI makes mistakes, it’s called hallucinating. There’s a better, if ruder, word for it. That means this post contains what the dictionary calls vulgar slang. You have been warned.
The AI tool everybody should be using – NotebookLM – Research assistant and note-taker rolled into one – allow me to introduce you to NotebookLM.
How to make a list that lasts – Making lists is one of life’s great pleasures. But when you have to make the same list again and again, not so much. Here’s a trick for that…
How can I help you make order from chaos?
Join the Safe Hands AI Explorer learning circle!
Sign up for my Sensible Woman’s Guide to AI and Content Creation, which comes out fortnightly.
Or sign up for my personal fortnightly email newsletter In Safe Hands (it’s a place to pause and think about things).
Book a free half hour of my time here. Find out how I can help with writing, editing, project management, general clear thinking, technical hand holding, or an introduction to AI.
Contact me via email.