AI Primer: Key Terms & Concepts for LLM Beginners

Generative AI

As artificial intelligence {insert some AI-generated biz jargon here about how quick change is happening}.

Look, if you’re reading this, you are probably a leader on a team trying to bring the crew along on a race that feels like other orgs have a huge start on…
Good news: the tools are changing so quickly that actually stuff you would have learned last year has kinda changed.
Bad news: teams are already quietly using AI at work to the tune of 75% of employees across all ages according to Microsoft at Work research

TL;DR: Not having an AI policy and approach is increasingly a riskier move.

Quick Reference: Essential AI Terms

TermDefinitionWhy It Matters to Beginners
AI (Artificial Intelligence)Computer systems that simulate human intelligenceUnderstanding AI helps you identify opportunities to automate tasks and enhance your work
LLM (Large Language Model)AI trained on massive text data to understand/generate languageThese are the tools you’ll actually use (ChatGPT, Claude, etc.)
ModelThe “brain” of the AI system (e.g., GPT-4, Claude 4)Different models have different strengths and limitations
PromptThe instruction or question you give the AIYour primary way to communicate with AI; better prompts = better results
TokenChunks of text the AI processes (1,000 tokens ≈ 750 words)Helps you understand usage limits and costs
Context WindowAmount of text the AI can “remember” at onceDetermines how much information you can include in a conversation
HallucinationWhen AI generates incorrect or made-up informationCritical to know so you always fact-check important information
Prompt EngineeringCrafting instructions for better AI responsesThe skill that separates mediocre from excellent AI results
RAG (Retrieval Augmented Generation)AI that looks up external info before answeringEnables more accurate, current responses
AI BiasAI reflecting certain viewpoints from training dataEssential for ethical use and understanding limitations
(AI) GlazingTendency for AI to overly flatter the user to increase the scores it gets from satisfied usersSicophantic responses fail to challenge bad ideas

How AI Works: Core Terms & Concepts

Let’s start with the building blocks. Artificial Intelligence (AI) refers to computer systems that simulate human intelligence. Think of it as teaching computers to think and respond in ways that feel remarkably human.

At the heart of today’s AI revolution are Large Language Models (LLMs). These are AI systems trained on massive amounts of text data to understand and generate language. You’ve likely heard of ChatGPT, Claude, or Google’s Gemini. These are all LLMs, each with their own strengths and quirks.

The model is essentially the “brain” of the AI system. When someone mentions GPT-4, Claude 3, or Llama 3, they’re referring to specific models with different capabilities. These models learn through parameters, which are millions or billions of settings and variables adjusted during training, where the AI learns patterns from enormous datasets.

When you ask an AI a question, it performs inference, using its learned knowledge to generate an answer. The AI processes text in chunks called tokens (roughly 1,000 tokens equals 750 words), and its context window determines how much information it can “remember” during a conversation.

Here’s a crucial point: AI is a probabilistic model, meaning it predicts the most likely next word or token based on patterns it learned. This is why AI can sometimes experience hallucinations, generating plausible-sounding but incorrect information. Always fact-check critical information!

How to Use AI: Interaction & Prompting

Your prompt, the instruction or question you give the AI, is your primary tool for getting useful results. Prompt engineering is the art of crafting these instructions effectively. As we’ve learned at Whole Whale, the difference between a generic prompt and a well-crafted one can mean the difference between bland, “grey jacket” content and something truly valuable.

Remember the principle of “First, Not Final”: AI excels at generating first drafts and brainstorming ideas, but human judgment remains essential for polishing and finalizing work. This Human-in-the-Loop approach ensures accuracy and maintains your organization’s unique voice.

Pro tips for better prompting:

  • Use delimiters like <goal>...</goal> to organize complex instructions
  • Engage in multi-shot conversations for nuanced tasks
  • Prime the pump by providing relevant background information
  • Remember: Garbage In, Garbage Out (GiGo): the quality of your prompt directly impacts the quality of AI output

Advanced AI Concepts & Practical Risks

As you advance in your AI journey, you’ll encounter more sophisticated concepts. Agents are AI systems that can take actions or use tools and APIs autonomously. RAG (Retrieval Augmented Generation) allows AI to “look up” external information before answering, making responses more accurate and current.

Behind the scenes, vector databases store embeddings, which are mathematical representations of text that enable semantic search. Similarity scores measure how closely texts match, while chunk size and overlap determine how information is stored and retrieved in RAG systems.

Critical Considerations for Responsible AI Use

Bias remains a significant concern. AI systems can reflect and amplify biases present in their training data. For nonprofits committed to equity and justice, understanding and mitigating these biases is crucial.

Data privacy deserves special attention. Some AI tools train on user inputs, potentially exposing sensitive information. As highlighted in our “Don’t Train on Me” guidance, be cautious about what data you share with AI systems, especially when dealing with donor information or sensitive organizational data.

Finally, disclosure is non-negotiable. Always be transparent when AI has been used to generate or assist with public-facing content. Your stakeholders deserve to know when they’re interacting with AI-generated material.

Moving Forward with AI

Understanding these concepts is just the beginning. As AI continues to evolve, staying informed and experimenting responsibly will help your organization harness its potential while avoiding pitfalls. Remember, AI is a powerful tool, but it’s the human insight, creativity, and ethical judgment that transform that power into meaningful impact.

At Whole Whale, we’ve been integrating AI into our work since early 2022, learning what works and what doesn’t. The key is to start small, experiment thoughtfully, and always keep your mission at the center of how you use these tools.

Ready to dive deeper? Check out our AI Writing Prompt Formula for practical tips on crafting effective prompts, and remember: in the world of AI, staying curious and cautious in equal measure is your best strategy for success.