Complete AI Mastery Course

{"contentFormat":"slides.v1","completion":{"requireAllSlides":true,"requireQuiz":true},"slides":[{"kind":"title","eyebrow":"Module 3: AI Productivity and Study Superpowers","title":"AI as a Research Assistant: Finding and Verifying Sources","body":"Research used to mean hours of typing keywords into a search engine, opening twenty tabs, and slowly piecing together an answer. Today, AI research assistants can find, summarize, and connect information in seconds. But this power comes with a serious responsibility: AI can present a confident, well-written answer that is partly or completely wrong. In research, being fast is worthless if you are not also being right.\n\nIn this lesson you will learn how to use AI to accelerate research while staying accurate. You will discover which tools are built for finding real, citable sources, how to read a citation critically, and — most importantly — how to verify what AI tells you before you rely on it. By the end, you will be able to run a research task from question to verified answer with confidence.","outcomes":["Distinguish between closed models and search-connected AI tools for research","Understand grounding, citation, and the risk of hallucination","Apply a four-step verification workflow: generate, locate, confirm, cross-check","Evaluate source credibility using publisher, date, and bias","Use AI to accelerate research while maintaining accuracy"],"narration":"Welcome to this lesson on using AI as a research assistant. We'll explore how to find and verify sources effectively, avoiding common pitfalls like hallucination. By the end, you'll have a practical workflow to ensure your research is both fast and reliable."},{"kind":"content","heading":"Two Kinds of AI Tools for Research","body":"Not all AI tools are equal when it comes to research. Understanding the difference helps you choose the right tool for the task.\n\nClosed models (e.g., default ChatGPT, Claude, Gemini without web access) answer from their training data. They are fluent but can be outdated and prone to hallucination — inventing plausible-sounding facts, quotes, or citations that do not exist.\n\nSearch-connected tools (e.g., Perplexity, ChatGPT with search, Gemini with Google, Claude with web search) retrieve live web pages and cite them. Their claims link to real sources you can open and verify.\n\n> Analogy: A closed model is like a well-read friend answering from memory — often right, sometimes confidently mistaken. A search-connected tool is like a librarian who walks you to the exact shelf and hands you the book.","callout":{"variant":"insight","title":"Key Insight","text":"For any research involving current facts, always prefer a search-connected tool. Closed models are best for brainstorming or general knowledge, not for sourcing verifiable information."},"narration":"Let's start by understanding the two main types of AI tools for research. Closed models answer from memory and can hallucinate. Search-connected tools retrieve live sources and cite them. Choose wisely based on your need for accuracy."},{"kind":"content","heading":"Grounding, Citation, and Hallucination","body":"Three concepts are central to using AI for research:\n\n- Grounding: An AI answer is tied to specific retrieved documents rather than general memory. Search-connected tools ground their answers in real web pages.\n- Citation: The AI shows where each claim came from. A clickable, working link is trustworthy; a citation without a link should be treated as unverified.\n- Hallucination: When a model states something false as if it were true. In research, this often looks credible: real-sounding author names, fake DOIs, invented statistics, or misquoted studies. Fluency is not evidence of accuracy.\n\nNotebookLM is a special tool that grounds answers only in sources you upload and cites the exact passage — ideal when you already have your documents.","callout":{"variant":"warning","title":"Warning","text":"Hallucination is the central danger in AI-assisted research. Always verify claims by opening the cited source. If a citation looks suspicious, treat it as unconfirmed."},"narration":"Now let's define grounding, citation, and hallucination. Grounding ties answers to real documents. Citation shows the source. Hallucination is when AI invents false information. Always verify by opening the cited source."},{"kind":"content","heading":"The Verification Workflow","body":"Treat AI as a starting point, not a source of truth. Follow this four-step workflow:\n\n1. Generate — Let AI find candidate sources and summarize.\n2. Locate — Open the actual cited source yourself.\n3. Confirm — Check the claim really appears there and is represented fairly.\n4. Cross-check — For anything important, confirm with a second independent source.\n\nEvaluating a source: Even a real source may be weak. Ask:\n- Who published it? (Peer-reviewed journal, government body, random blog?)\n- When? (Is it current enough for a 2026 question?)\n- Why? (Does the publisher have a bias or something to sell?)\n\nCitation formats: If you need a specific style (APA, MLA, Chicago), ask the AI to format citations accordingly. For example: \"Provide the citation in APA 7th edition format.\" Always double-check the generated citation against the original source.","callout":{"variant":"exercise","title":"Try It: Catch a Hallucination","text":"Ask a closed model (no web access) for \"three academic studies with authors and years about [your niche topic].\" Then try to find those exact studies on a search engine or scholarly database. You will often discover invented or garbled references — a powerful, memorable lesson."},"narration":"Here is the verification workflow: generate, locate, confirm, cross-check. Always evaluate the source's credibility. For academic work, ask AI to format citations in your required style, but verify them. Try the exercise to see how easily AI can hallucinate references."},{"kind":"content","heading":"Real-World Examples","body":"These examples show how professionals use AI for research while maintaining accuracy.\n\nA university student in Kenya is writing an essay on renewable energy in East Africa. She uses Perplexity to gather recent figures, then clicks each citation. One \"statistic\" links to a page that never mentions the number — a red flag. She discards it and keeps only claims she can confirm on the original site.\n\nA journalist in Germany investigating a company uses Claude with web search to gather background quickly, then independently opens each cited article and the company's official filings. AI saved her hours of searching, but every published fact traces back to a primary source she verified herself.\n\nA pharmacist in India wants the latest guidance on a drug interaction. She knows closed models can be outdated, so she uses a search-connected tool to find official health-authority publications, then confirms directly on the authority's website — never trusting the AI's summary alone for a clinical decision.","callout":{"variant":"tip","title":"Prompt Tip","text":"Add this to your prompt: \"Answer using current web sources. After each claim, include the source link. If you are not certain a source supports a claim, say so.\" This encourages transparency."},"narration":"Let's look at real-world examples from Kenya, Germany, and India. Each professional uses search-connected tools and verifies every source. Notice how they discard unverifiable claims and cross-check critical information."},{"kind":"content","heading":"Common Pitfalls & How to Avoid Them","body":"Avoid these common mistakes to keep your research reliable:\n\n| Pitfall | Fix |\n|---------|-----|\n| Trusting fluent answers | Never accept a claim without an openable, relevant source |\n| Accepting citations without clicking | Open every citation and confirm it actually supports the claim |\n| Using closed models for current facts | Use search-connected tools for anything time-sensitive |\n| Relying on a single source | Cross-check important facts with a second independent, credible source |\n| Ignoring source quality | Evaluate publisher, date, and motive before trusting |\n\nHands-on steps:\n1. Use a search-connected tool (Perplexity, ChatGPT with search, etc.).\n2. Demand citations in your prompt.\n3. Open every source.\n4. Catch a hallucination on purpose (see exercise).\n5. Cross-check key facts.","table":{"headers":["Pitfall","Fix"],"rows":[["Trusting fluent answers","Never accept a claim without an openable, relevant source"],["Accepting citations without clicking","Open every citation and confirm it supports the claim"],["Using closed models for current facts","Use search-connected tools for anything time-sensitive"],["Relying on a single source","Cross-check with a second independent, credible source"],["Ignoring source quality","Evaluate publisher, date, and motive before trusting"]]},"narration":"Here are common pitfalls and their fixes. The table summarizes them. Remember: always open sources, use search-connected tools for current facts, and cross-check. Practice the hands-on steps to build good habits."},{"kind":"quiz","heading":"Check Your Understanding","questions":[{"question":"Which type of AI tool is safest for finding current, verifiable information?","options":["A closed model like default ChatGPT without web access","A search-connected tool like Perplexity or ChatGPT with search","Any AI model, as long as it sounds confident","A model trained only on data before 2020"],"questionId":"cmrf73k5j000xpd27l7er2lha"},{"question":"What is the first step in the verification workflow?","options":["Cross-check with a second source","Generate candidate sources using AI","Confirm the claim in the original source","Locate the cited source"],"questionId":"cmrf73k5j000ypd27y81bd9u3"},{"question":"Why should you never trust a citation without clicking it?","options":["Because AI can fabricate real-looking references that do not exist","Because citations are always outdated","Because clicking is required by academic integrity rules","Because the AI might have used a paywalled source"],"questionId":"cmrf73k5j000zpd27uv32w59d"}],"narration":"Let's test your understanding with a short quiz. Choose the best answer for each question. After answering, read the explanation to reinforce your learning.","quizId":"qz_cmk7lgx4u000pg4p8z7kkq7so"},{"kind":"summary","heading":"Key Takeaways","takeaways":["AI is a research accelerator, not an authority — verify before you rely.","Prefer search-connected tools (Perplexity, ChatGPT with search, Gemini, Claude with web search) for anything current; they cite real sources.","Hallucination is the central danger: fluent does not mean factual.","Always open and read the cited source — never trust an uncheckable citation.","Follow the verification workflow: Generate, Locate, Confirm, Cross-check.","Evaluate source credibility by asking who, when, and why.","For academic work, ask AI to format citations in your required style (e.g., APA, MLA), but verify each one."],"narration":"To summarize: AI accelerates research but requires verification. Use search-connected tools, watch for hallucination, and always open sources. Follow the four-step workflow and evaluate credibility. With these practices, you can harness AI's power while maintaining accuracy."}]}