How to Use This Plan #
- Tracking progress: The first column is blank. Fill it in when you finish a day so you can see completed and pending days at a glance.
- Schedule: 12 weeks, 6 study days per week, about 30 minutes per day. Day 01–72 are study days, not calendar dates. Use the seventh day of each week for rest or catching up.
- Goal: Explain AI concepts, summarize information, and discuss trade-offs in English at a B2 target level. These resources do not have official CEFR ratings.
- Daily routine: 3 minutes of recall, 10 minutes of video study, 8 minutes of reading, and 9 minutes of speaking or writing. Save three useful phrases, not a long list of isolated words.
- Manage the workload: Study a 3–6-minute video segment and roughly 300–500 words of the article. There is no need to finish a long video or article in one sitting. Focus labels below describe what to look for, not exact chapter titles or timestamps.
- Repeat deliberately: Each week reuses one video/article pair. Listen for the main idea, check details with English captions when available, imitate a short passage, explain the idea, then review without notes.
- Adapt the difficulty: If a segment is too difficult, shorten it to 1–2 minutes and use captions or a slower playback speed. Weeks 4 and 11 may require this extra support.
- Sources: Videos are from IBM Technology/IBM Think, except Week 4 from 3Blue1Brown. Articles are from IBM Think. They are useful language material but reflect a vendor perspective; separate educational explanations from product promotion. Video links lead to individual YouTube videos or IBM pages with embedded videos; playback and caption availability depend on your region and browser.
Week 1 — AI, Machine Learning, and Deep Learning #
Video:
AI, Machine Learning, Deep Learning and Generative AI Explained +
Machine Learning vs Deep Learning +
AI vs Machine Learning
Article:
What is artificial intelligence (AI)?
| Day | Topic | Video | Article | English Practice | |
|---|---|---|---|---|---|
| 01 | What is AI? | AI explained — listen for the main idea， AI vs ML, AI Simplified: 6 Concepts — overview of modern AI concepts | What is AI? — introduction | Record a 60-second baseline explanation without a script. | |
| 02 | AI vs. machine learning | AI vs ML +
AI explained — compare the definitions, Understanding AI Concepts — compare AI, ML, and GenAI | AI vs ML vs DL vs Neural Networks — compare the definitions | Write three comparison sentences using “whereas” or “while.” | |
| 03 | Machine learning vs. deep learning | ML vs DL +
AI vs ML — listen for the pizza analogy, Machine Learning Explained — ML, AI, and deep learning guide | Deep Learning — what is deep learning? | Explain the relationship using one everyday analogy. | |
| 04 | Where does generative AI fit? | AI explained +
AI vs ML — focus on generative AI, Evolution of AI — traditional AI vs generative AI | Generative AI — what is generative AI? | Draw a concept map and describe it aloud for 90 seconds. | |
| 05 | AI in everyday work | ML vs DL +
AI vs ML — replay an example, Rise of GenAI for Business — practical applications | AI Use Cases — valuable business applications | Write 80–100 words about one useful workplace application. | |
| 06 | Review: explain AI to a colleague | AI explained +
AI vs ML — replay without captions, Brief History of AI — review the full timeline | Types of AI — review the different types | Give a two-minute explanation; compare it with Day 01. |
Week 2 — Training and Inference #
Video:
AI Inference: The Secret to AI’s Superpowers
Article:
What is AI inference?
| Day | Topic | Video | Article | English Practice | |
|---|---|---|---|---|---|
| 07 | What happens when a model answers? | AI inference — first listen, How Deep Learning Works — how networks process input | What is AI inference? — definition | Describe inference in three plain-English sentences. | |
| 08 | Training vs. inference | AI inference — listen for differences, LLMs Explained — how models are trained and make predictions | Training Data — training comparison | Make a two-column comparison and explain it aloud. | |
| 09 | From input to prediction | AI inference — follow the process, Intro to LLMs — follow the prediction pipeline | Model Performance — how inference works | Use “first,” “next,” and “finally” to explain the process. | |
| 10 | Why response time matters | AI inference — replay a technical passage, Google Cloud AI Low-latency — why latency matters worldwide | Edge AI — latency and performance | Explain latency to a nontechnical colleague in 60 seconds. | |
| 11 | Quality, speed, and cost | AI inference — revisit the explanation, Model Providers Compared — speed, cost, and intelligence trade-offs | Model Deployment — efficiency and deployment | Write 100 words about a trade-off using “however” and “depends on.” | |
| 12 | Review: explain an AI request | AI inference — listen without captions, LLMs Explained — replay the full process | AI Infrastructure — verify your summary | Explain training and inference in two minutes without notes. |
Week 3 — Generative AI #
Video:
Generative models explained
Article:
What is generative AI?
| Day | Topic | Video | Article | English Practice | |
|---|---|---|---|---|---|
| 13 | What does generative AI generate? | Generative models — first listen, Intro to GenAI — what genAI creates | Generative AI — introduction | Give three examples of generated content. | |
| 14 | Generating vs. classifying | Generative models — compare tasks, GenAI vs Discriminative — generation vs classification | Generative Model — generative and discriminative models | Explain the contrast using “unlike” and “in contrast.” | |
| 15 | What is a foundation model? | Generative models — listen for model terminology, Intro to LLMs — LLMs as foundation models | Foundation Models — foundation models | Define three key terms in your own words. | |
| 16 | How prompts shape an answer | Generative models — replay a model example, GenAI Studio — prototype models with prompts | Prompt Engineering — prompts and model outputs | Draft two versions of an instruction: vague and specific. | |
| 17 | Useful applications and limitations | Generative models — revisit the applications, Responsible AI — limitations and responsible use | GenAI Use Cases — benefits and challenges | Write 100 words with one benefit, one limitation, and one example. | |
| 18 | Review: propose a realistic use case | Generative models — replay without captions, GenAI vs Traditional AI — when to choose generative AI | AI Model — fact-check your proposal | Give a two-minute proposal and answer one skeptical question. |
Week 4 — Large Language Models #
Video:
Large Language Models explained briefly — 3Blue1Brown
Article:
What are large language models (LLMs)?
| Day | Topic | Video | Article | English Practice | |
|---|---|---|---|---|---|
| 19 | What is an LLM? | LLMs explained briefly — opening explanation, Introduction to LLMs — listen for the definition | Large language models — definition | Explain an LLM without using the word “intelligent.” | |
| 20 | Tokens and next-token prediction | LLMs explained briefly — prediction example, Transformers, the tech behind LLMs — how tokens drive next-word prediction | LLM Inference — how LLMs work | Explain “token” and “prediction” using a short sentence as an example. | |
| 21 | Learning from text | LLMs explained briefly — training explanation, How LLMs Are Trained — how models learn from text data | Fine-Tuning — training | Write five sentences using “is trained,” “is used,” and “is generated.” | |
| 22 | Context and attention | LLMs explained briefly — transformer explanation, Attention in transformers — visual walk-through of attention | Attention Mechanism — transformers and attention | Give a simple explanation; identify one detail you still do not understand. | |
| 23 | Capabilities do not guarantee reliability | LLMs explained briefly — replay the overview, Why Does AI Hallucinate? — why outputs can be wrong | LLM Benchmarks — uses and limitations | Write 100–120 words using “can,” “may,” and “does not necessarily.” | |
| 24 | Review: how does a chatbot work? | LLMs explained briefly — selected segment without captions, How ChatGPT Works — end-to-end recap | Chatbots — verify terminology | Record a two-minute explanation and reuse five phrases from Weeks 1–4. |
Week 5 — Hallucinations and Fact-Checking #
Video:
Tuning Your AI Model to Reduce Hallucinations
Article:
What are AI hallucinations?
| Day | Topic | Video | Article | English Practice | |
|---|---|---|---|---|---|
| 25 | What is an AI hallucination? | Reducing hallucinations — first listen, LLMs Don’t Hallucinate — what a confident wrong answer is | AI hallucinations — definition | Describe a hypothetical incorrect AI answer and why it matters. | |
| 26 | Why can an answer sound convincing? | Reducing hallucinations — listen for explanations, Solving AI Hallucinations — why models make things up | LLM Temperature — causes | Explain a cause and effect using “because” and “as a result.” | |
| 27 | Confident language vs. reliable evidence | Reducing hallucinations — replay a key claim, Why AI Makes Things Up — why confident answers mislead | Explainable AI — examples and risks | Rewrite three overconfident claims using cautious language. | |
| 28 | Reducing errors | Reducing hallucinations — focus on mitigation, 5 Anti-Hallucination Techniques — concrete mitigation techniques | AI Guardrails — prevention approaches | Explain two ways to reduce errors without promising to eliminate them. | |
| 29 | Checking an AI-generated answer | Reducing hallucinations — revisit the advice, Solving AI’s Biggest Problem — grounding, search, and NotebookLM checks | Human-in-the-Loop — detection and verification | Write a five-step fact-checking checklist in English. | |
| 30 | Review: when should we trust AI? | Reducing hallucinations — replay without captions, Never Trust An LLM — a trust-but-verify mindset | Trustworthy AI — verify your conclusion | Speak for two minutes using “The evidence suggests…” and “I would verify…”. |
Week 6 — Retrieval-Augmented Generation #
Video:
What is Retrieval-Augmented Generation (RAG)?
Article:
What is retrieval augmented generation (RAG)?
| Day | Topic | Video | Article | English Practice | |
|---|---|---|---|---|---|
| 31 | Why retrieve information before answering? | RAG explained — first listen, What is RAG? — listen for why retrieval beats retraining | Retrieval-augmented generation — introduction | Explain the problem RAG is intended to address. | |
| 32 | Retrieve, add context, generate | RAG explained — follow the workflow, End-to-End RAG Pipeline — follow the retrieve–augment–generate pipeline | RAG and Vector Databases — how RAG works | Describe the workflow in four ordered sentences. | |
| 33 | External knowledge and freshness | RAG explained — external-information example, How Vector Embeddings Work — how external text becomes searchable knowledge | Knowledge Graph — knowledge sources | Explain why a company might connect internal documentation. | |
| 34 | Citations and verification | RAG explained — source-related explanation, RAG and Hallucinations — how grounded answers support checking | Vector Database — benefits | Explain why a citation helps checking but does not guarantee correctness. | |
| 35 | RAG vs. fine-tuning | RAG explained — revisit the RAG workflow, RAG vs Fine-Tuning vs Long Context — when retrieval beats retraining | RAG vs. Fine-tuning — fine-tuning comparison | Write a 120-word comparison using “whereas” and “depending on.” | |
| 36 | Review: design a documentation assistant | RAG explained — replay without captions, RAG Tutorial for Beginners — watch a documentation assistant being built | Enterprise Search — check your design | Present the data source, workflow, and one limitation in two minutes. |
Week 7 — AI Agents #
Video:
What are AI Agents?
Article:
What are AI agents?
| Day | Topic | Video | Article | English Practice | |
|---|---|---|---|---|---|
| 37 | What makes a system an agent? | AI agents — first listen, AI Agents, Clearly Explained — spot the agent in real examples | AI agents — definition | Explain an agent with one practical example. | |
| 38 | Goals, planning, and actions | AI agents — follow an example, AI Agents Explained Simply — how agents plan and act | Agentic Workflows — how agents work | Describe a task using “in order to” and “so that.” | |
| 39 | Tools and external systems | AI agents — listen for tool use, MCP Explained Simply — how agents connect to tools | Tool Calling — tool integration | Explain what a calendar or search tool adds to an agent. | |
| 40 | Feedback and iteration | AI agents — replay the process, What Agents Really Are — the agent loop: act, observe, repeat | Agentic Reasoning — reasoning and iteration | Describe an unsuccessful action and the next step using conditionals. | |
| 41 | Autonomy and human oversight | AI agents — revisit the example critically, Agentic AI vs RAG — when autonomy helps and when it does not | AI Safety — risks and challenges | Write three rules about actions that require human approval. | |
| 42 | Review: agent or chatbot? | AI agents — replay without captions, Build Your First AI Agent — build a mini agent to see the difference | AI Agents vs. AI Assistants — check the distinctions | Compare two hypothetical systems in a two-minute explanation. |
Week 8 — AI-Assisted Coding #
Video:
What is an AI Code Generator? LLM Coding, Productivity, & Risk
Article:
What is AI code generation?
| Day | Topic | Video | Article | English Practice | |
|---|---|---|---|---|---|
| 43 | How can AI assist a developer? | AI code generators — first listen, AI Has Changed How We Build Software — how AI changed day-to-day development | AI code generation — definition and uses | Describe three development tasks AI can assist with. | |
| 44 | Giving clear coding instructions | AI code generators — replay an example, How to Make Vibe Coding Not Suck — turning vague vibes into clear instructions | Vibe Coding — natural-language inputs | Draft a prompt with a goal, constraints, and acceptance criteria. | |
| 45 | Reading and explaining generated code | AI code generators — focus on the workflow, 5 Claude Code Skills — reviewing what the AI wrote | AI Coding Assistants — code-assistance capabilities | Explain a familiar function in five plain-English sentences. | |
| 46 | Testing before trusting | AI code generators — listen for risks, AI Is Lying to Developers — why claims need verification | AI Unit Testing — limitations | Explain why a plausible solution still needs tests and review. | |
| 47 | Security and maintainability | AI code generators — revisit the risk discussion, Your Codebase Is Not Ready for AI — maintainability before automation | Secure Coding — challenges | Write a 120-word recommendation for responsible team use. | |
| 48 | Review: should a team adopt an AI coding tool? | AI code generators — selected segment without captions, We Studied 150 Developers Using AI — evidence for the adoption decision | AI in Software Development — fact-check your recommendation | Give a two-minute recommendation with a benefit, risk, and safeguard. |
Week 9 — Productivity and Automation #
Video:
Reimagine business productivity with AI agents and assistants
Article:
What is intelligent automation?
| Day | Topic | Video | Article | English Practice | |
|---|---|---|---|---|---|
| 49 | What work can AI help with? | AI and productivity — first listen, Make Yourself AI-Native — which kinds of work AI can take on | Intelligent automation — introduction | Describe a repetitive task from your own work. | |
| 50 | Assistants, agents, and automation | AI and productivity — compare agents and assistants, What Is Agentic AI? — how assistants, agents, and automation differ | Digital Worker — automation components | Explain how these concepts overlap without treating them as identical. | |
| 51 | Mapping a workflow | AI and productivity — replay a use case, Build No-Code AI Workflows — map a workflow step by step | Workflow Automation — applications | Describe a five-step workflow and identify one possible AI-assisted step. | |
| 52 | Measuring useful improvements | AI and productivity — evaluate productivity claims, Does AI Boost Productivity? — what a 100k-dev study measured | AI in the Workplace — benefits | Propose two measures, such as task time and error rate. | |
| 53 | What should remain under human control? | AI and productivity — revisit the use case, Skills AI Can’t Replace — which tasks should stay human | AI Ethics — challenges | Write 120 words about a task you would not fully automate. | |
| 54 | Review: propose a small pilot | AI and productivity — replay a segment without captions, The Only AI Tools You Need — pick a minimal toolset for your pilot | Enterprise Automation — verify your proposal | Present a workflow, success measure, and human checkpoint in two minutes. |
Week 10 — Privacy and AI Governance #
Video:
What is AI governance?
Article:
Exploring privacy issues in the age of AI
| Day | Topic | Video | Article | English Practice | |
|---|---|---|---|---|---|
| 55 | Why do AI systems need rules? | AI governance — first listen, AI Governance Simplified — how rules turn principles into practice | AI privacy — introduction | Explain the purpose of governance and give one privacy example. | |
| 56 | Personal and confidential information | AI governance — replay the safety analogy, STOP Using ChatGPT — why personal data does not belong in public chatbots | Data Privacy — data collection and privacy risks | Describe three types of information you would avoid uploading. | |
| 57 | Consent and transparency | AI governance — listen for responsibility, What is Data Privacy? — listen for consent and disclosure | AI Transparency — consent and data use | Draft three questions to ask an AI service provider. | |
| 58 | Reducing unnecessary exposure | AI governance — revisit the control idea, 5 ChatGPT Privacy Settings — settings that reduce everyday exposure | Data Security — privacy protections | Write five practical rules using “must,” “should,” and “must not.” | |
| 59 | Accountability when something goes wrong | AI governance — replay the explanation, The EU’s AI Act, Explained — who is accountable under the rules | AI Governance — risks and safeguards | Explain who should investigate, communicate, and approve corrective action. | |
| 60 | Review: a responsible-use policy | AI governance — replay without captions, Introduction to Responsible AI — principles for your policy review | Responsible AI — check your policy | Present a two-minute policy covering data, oversight, and reporting. |
Week 11 — Evaluating AI Systems #
Video:
LLM as a Judge: Scaling AI Evaluation Strategies
Article:
LLM evaluation: Why testing AI models matters
| Day | Topic | Video | Article | English Practice | |
|---|---|---|---|---|---|
| 61 | What does “good output” mean? | LLM as a judge — opening explanation, How to Setup LLM Evaluations — what an evaluation actually measures | LLM evaluation — introduction | Define quality for one use case with three criteria. | |
| 62 | Human and automated evaluation | LLM as a judge — compare approaches, Master LLM Evaluations — human judgment vs automated metrics | Model Evaluation — evaluation methods | Explain one strength and one weakness of each approach. | |
| 63 | Asking a model to judge answers | LLM as a judge — follow the judging process, LLM-as-a-Judge 101 — how a model grades another model’s answers | Ground Truth — reference answers for judging | Describe the process using “according to the criteria.” | |
| 64 | Rubrics and test cases | LLM as a judge — replay a practical passage, How to Systematically Setup LLM Evals — metrics, unit tests, and rubrics in practice | Synthetic Data — test data for evaluation | Write a three-point rubric for judging an English summary. | |
| 65 | Bias and unreliable scores | LLM as a judge — listen for limitations, 6 Types of Machine Learning Bias — six bias types that skew results | AI Bias — biased outputs and fairness | Write 120–150 words explaining why one score is not enough. | |
| 66 | Review: evaluate two sample answers | LLM as a judge — selected segment without captions, LLM as a Judge 102 — can the judge itself be trusted? | Model Risk Management — oversight and controls | Draft two short answers yourself and justify their scores aloud. |
Week 12 — Open Models and Choosing an AI Approach #
Video:
Should you use open source Large Language Models?
Article:
What is open-source AI?
| Day | Topic | Video | Article | English Practice | |
|---|---|---|---|---|---|
| 67 | Open and proprietary models | Open-source LLMs — first listen, Open Source or Proprietary LLMs? — the same question from a practitioner’s view | Open-source AI — introduction | Explain the contrast without assuming that “open” means unrestricted. | |
| 68 | Open source, open weights, and licenses | Open-source LLMs — listen for openness claims, Open Source vs Open Weight — what “open” actually licenses | Open-Source LLMs — definitions and licensing | Explain why a model’s license and available artifacts both matter. | |
| 69 | Control, customization, and deployment | Open-source LLMs — focus on benefits, Learn Ollama in 15 Minutes — running models locally on your own machine | Model Training — customizing with your own data | Describe a scenario where local deployment could be useful. | |
| 70 | Costs and shared risks | Open-source LLMs — replay the trade-offs, Complete Open Source AI Explained — the full cost and risk picture | AI Security — risks to manage | Compare two approaches using quality, privacy, cost, and maintenance. | |
| 71 | Preparing a recommendation | Open-source LLMs — revisit the main argument, How To Choose The Best LLM — a framework for your recommendation | Model Selection — evidence for your decision | Write a 150-word recommendation using ideas from at least three weeks. | |
| 72 | Final presentation and reflection | Open-source LLMs — replay a segment without captions, You’re Not Behind — reflect on how far your clarity has come | Machine Learning — revisit the fundamentals | Give a three-minute presentation; compare your clarity with Day 01. |
Weekly Self-Check #
After each sixth study day, record:
- Listening: Can I identify the main point and three supporting details without captions?
- Vocabulary: Can I use five useful phrases from this week in new sentences?
- Speaking: Can I explain the topic for two minutes using only a few keywords?
- Writing: Can I produce a short summary with a main point, an example, and a limitation?
- Next step: Which one difficulty should I focus on next week? Repeat a difficult week if needed; finishing on schedule is not a language-proficiency test.