AI jargon shouldn’t slow you down. Here’s a plain‑English guide to 15 terms you’ll see in roadmaps, vendor decks, and product docs—distilled from the MIT Sloan EdTech AI Basics Glossary and tailored for operators.
The 15 terms that matter now
- Artificial Intelligence (AI): The broad field of building systems that perform tasks that typically require human intelligence. Why it matters: It’s the umbrella term; everything below lives inside it.
- Machine Learning (ML): Algorithms that learn patterns from data rather than being explicitly programmed. Why it matters: Most practical AI in products today is ML under the hood.
- Deep Learning: A subset of ML using multi‑layer neural networks to model complex patterns (e.g., vision, speech, language). Why it matters: Powers state‑of‑the‑art results but is compute‑hungry.
- Large Language Model (LLM): A deep learning model trained on text to predict the next token. Why it matters: LLMs enable chatbots, drafting, code assistants, and knowledge search.
- Token: A chunk of text (often ~4 characters or ~0.75 words in English). Why it matters: Costs and context limits are measured in tokens—budget and prompt length depend on them. Tip: Short, specific prompts save tokens.
- Transformer: The neural network architecture behind modern LLMs that uses “attention” to weigh context. Why it matters: It’s why LLMs scale and maintain coherence over longer inputs.
- Context Window: The maximum number of tokens an LLM can consider at once. Why it matters: Exceed it and the model “forgets.” Tip: Summarize or chunk long docs before prompting.
- Embeddings: Numeric vectors that represent the meaning of text. Why it matters: Used for semantic search, deduping, clustering, and RAG. Tip: Store them in a vector database for fast retrieval.
- Retrieval‑Augmented Generation (RAG): A pattern where you fetch relevant documents (via embeddings) and feed them to the LLM as context. Why it matters: Reduces hallucinations and lets models answer with your private data.
- Grounding / Attribution: Supplying sources or citations to support an answer. Why it matters: Enables trust. Tip: Ask the model to cite passages or URLs and keep links clickable.
- Prompt Engineering: Crafting instructions and constraints for the model (system, user, and example prompts). Why it matters: Better prompts improve accuracy without code changes. Tip: Specify role, format, and evaluation criteria.
- Zero‑Shot / Few‑Shot: Getting the model to perform a task with no examples (zero‑shot) or a handful (few‑shot). Why it matters: Few‑shot often boosts accuracy for niche tasks.
- Fine‑Tuning: Training a base model further on your labeled data. Why it matters: Useful for style, domain, or format consistency; heavier lift than prompt engineering but more reliable.
- Hallucination: When a model produces confident but incorrect output. Why it matters: A core risk in production. Mitigate: RAG, validation checks, citations, and clear refusal rules.
- Inference vs. Training: Training builds the model; inference runs it to produce outputs. Why it matters: Training is capex‑heavy; inference is your ongoing opex (latency, cost per request).
How these pieces fit together (cheat sheet)
- Ingest data → create embeddings → store in a vector DB.
- User asks a question → retrieve similar chunks → build a prompt within the context window.
- LLM generates an answer → add citations → apply guardrails/validation → return result.
Red flags in vendor pitches
- “No hallucinations.” All models can err; look for RAG, citations, and evaluation metrics.
- “Unlimited context.” There’s always a token limit; ask for exact window size and truncation strategy.
- “We fine‑tune everything.” Ask if prompt engineering or RAG would be cheaper and faster first.
Source & further reading
See the MIT Sloan EdTech AI Basics Glossary for authoritative definitions and examples.
Takeaway
Master these 15 terms and you’ll evaluate AI tools faster, design better prompts, and cut risk in production.
Like this? Get one practical AI nugget in your inbox each week—subscribe to our newsletter: theainuggets.com/newsletter
