At a glance. This is a senior-friendly dictionary of 30 common AI terms. Each entry has a one-sentence plain-English definition and a one-sentence example. Read end-to-end in 15 minutes; bookmark for daily reference.
A to Z of AI for Seniors
- AI (Artificial Intelligence)
- Computer software that can do tasks normally requiring human intelligence — reading, writing, recognising images, planning.
Example: ChatGPT reading a long PDF and giving you a one-paragraph summary.
- AGI (Artificial General Intelligence)
- A hypothetical future AI that can match or exceed human ability across any task — not yet built, much debated.
Example: An AI that could equally be a doctor, lawyer, novelist and engineer.
- Agent (AI Agent)
- An AI that doesn't just answer questions but takes actions on your behalf — booking, searching, sending.
Example: An agent that books your flight, hotel and cab for a Mumbai-to-Delhi trip in one prompt.
- AI4Seniors
- The India-based learning platform you're on right now — teaching retired professionals to use AI confidently.
Example: Umesh Prabhu's three-level workbook series for 60+ professionals.
- API (Application Programming Interface)
- A technical "door" through which other software can talk to an AI tool.
Example: A company plugging ChatGPT's API into its customer support system.
- Bias
- When AI gives unfair or skewed answers because its training data was unbalanced.
Example: An AI that assumes all doctors are male because most training photos were of male doctors.
- Chain-of-Thought
- When AI is asked to reason step-by-step before giving its final answer — gives more reliable results on hard problems.
Example: Adding "think step by step" to a maths question and watching AI work through it.
- ChatGPT
- The most popular AI chatbot, made by OpenAI, used by 200+ million people every week.
Example: Asking ChatGPT to draft a complaint letter to your housing society.
- Claude
- An AI assistant made by Anthropic, often praised for thoughtful long-form writing and safety.
Example: Using Claude to write a detailed 5-page consulting proposal.
- Context Window
- How much text AI can "remember" in one conversation — measured in tokens (roughly words).
Example: A 200,000-token context window can hold a 300-page book in memory.
- Copilot
- Microsoft's AI assistant, built into Word, Excel, Outlook and Windows.
Example: Asking Copilot in Excel to summarise a sales sheet into a chart.
- Embedding
- A way of turning words into numbers so AI can compare meanings — the "secret sauce" behind search and recommendations.
Example: Searching "scooter for daily commute" and finding a match for "Activa for office use".
- Fine-Tuning
- Further training a general AI model on specific data so it becomes expert in one area.
Example: Fine-tuning a model on Indian legal cases to make a law-specific assistant.
- Gemini
- Google's AI assistant family — strong at integration with Google Search, Maps, Docs and Gmail.
Example: Asking Gemini to summarise an email thread directly inside Gmail.
- GPT (Generative Pre-trained Transformer)
- The family of AI models behind ChatGPT — the "G", "P" and "T" describe how it was built.
Example: GPT-4 and GPT-5 are versions used by ChatGPT.
- Hallucination
- When AI confidently invents facts that aren't true.
Example: AI quoting a fake Supreme Court case that sounds real but doesn't exist.
- Inference
- The moment AI actually generates an answer for your prompt — different from training.
Example: You press Enter; the chatbot's reply appearing word-by-word is inference happening.
- Jailbreak
- A trick to make AI ignore its safety rules — generally a bad idea and against terms of service.
Example: Someone trying to get an AI to give bomb-making instructions through clever wording.
- LLM (Large Language Model)
- An AI trained on huge amounts of text to predict the next word — the brain behind ChatGPT.
Example: GPT-4, Gemini 1.5 Pro, Claude 3.5 and Llama 3 are all LLMs.
- Machine Learning (ML)
- Software that improves by learning from examples instead of being explicitly programmed.
Example: Gmail's spam filter learning what counts as junk by studying millions of emails.
- Model
- The trained "brain" of an AI — a giant file containing billions of learned numbers.
Example: GPT-4 is a model; ChatGPT is the chat app that talks to it.
- Multimodal
- An AI that handles more than just text — images, voice, video too.
Example: Showing ChatGPT a photo of your fridge and asking "what can I cook tonight?"
- Parameters
- The internal numbers an AI model learns during training — more parameters generally means more capability.
Example: GPT-4 reportedly has over a trillion parameters.
- Prompt
- What you type to an AI — your question, instruction or request.
Example: "Write a 200-word thank-you letter to my doctor" is a prompt.
- Prompt Engineering
- The skill of writing prompts that get the best results from AI.
Example: Adding "you are a senior chartered accountant" to a prompt to get more authoritative tax advice.
- RAG (Retrieval-Augmented Generation)
- A technique where AI looks up real documents before answering — reduces hallucinations.
Example: A company chatbot that searches its own product manuals before answering customer queries.
- RLHF (Reinforcement Learning from Human Feedback)
- A training step where humans rate AI answers, teaching the model to be more helpful and polite.
Example: Why ChatGPT feels friendlier than a raw GPT model — RLHF taught it manners.
- Token
- The small chunks of text AI reads and writes — roughly 3/4 of a word in English.
Example: "ChatGPT is helpful" is about 4 tokens.
- Training Data
- The huge collection of text, code and images used to teach an AI model.
Example: GPT-4 was trained on a large portion of the public internet plus licensed books.
- Vector Database
- A special database that stores embeddings — used for fast, meaning-based search.
Example: A vector database powering an AI that finds "all our reports about pollution control" even when the word "pollution" isn't used.
Try this now. Pick three terms from this glossary you didn't fully know. Open ChatGPT and ask: "Explain [term] to me as if I am a 70-year-old retired manager. Use one Indian real-world example." The combination of glossary + AI follow-up locks in your understanding.
Key Takeaways
- You need to know about 10 terms to be conversational about AI — prompt, model, LLM, ChatGPT, Gemini, hallucination, token, context window, agent, multimodal.
- The other 20 terms are useful when reading articles or talking to consultants.
- If a new term appears, ask AI itself to explain it in plain English.
Related Resources