AI Mastery Plan Challenge Workbook

AI that’s worth using starts with the basics: how it evolved, the words people use, and where bias creeps in. You’ll build that foundation fast by creating a timeline, writing a personal glossary, and doing one bias case study you can reuse later in real client work.

Grab a notebook (or a doc) and keep it open. As you work, aim for “useful and clear,” not “perfect and academic.” If a term feels fuzzy, that’s a good sign-your glossary will fix it.

Module I - Foundations: History of AI, Symbolic to Modern Systems

AI history is often told as a shift in approach. Symbolic AI focused on rules and logic. Machine learning learned patterns from training data. Deep learning used neural networks with many layers to learn even more complex patterns. Modern AI combines these ideas and scales them with better compute and bigger datasets.

Your Turn (Timeline Challenge: Visual timeline of AI history)

Create a visual timeline with 5 milestones. Use these labels (in your own words if needed):

• Symbolic AI era
• Machine learning becomes practical
• Deep learning breakthrough
• Big modern AI systems emerge
• Today’s “modern AI” direction

Fill in the approximate time order (earlier → later).

Completion line: You have exactly 5 milestone boxes connected left-to-right, each with a 1-sentence description.

Module I - Foundations: AI vs. ML vs. DL (and Why People Mix Them Up)

AI is the broad goal (building systems that do tasks that seem “smart”). ML is a method for learning from training data. DL is a type of ML that uses neural networks.

Apply It (Fill-in lines)

AI means: ________

ML means: ________

DL means: ________

Module I - Foundations: Neural Networks, Training Data, Inference, Parameters

A neural network is a model made of layers that transforms inputs into outputs. Training data is what the model learns from. Inference is when the trained model is used to produce results. Parameters are the learned numbers inside the model.

Your Turn (Quick definitions)

Neural network: ______

Training data: _______

Inference: _______

Parameters: ______

Module I - Foundations: Tokens and Embeddings (LLM-Friendly Vocabulary)

Tokens are chunks of text the model processes. Embeddings are vector representations that capture meaning so the model can compare and generalize.

Your Turn (Glossary Building: Personal glossary of AI terms)

Define 10 key AI terms in your own words. Use these exact terms: AI, ML, DL, neural networks, training data, inference, parameters, tokens, embeddings, bias.

Completion line: All 10 terms have a definition you could explain to a teammate in under 2 sentences each.

Module I - Foundations: Bias in AI (Sources, Real-World Examples, Ethics)

Bias can come from skewed training data, missing groups, bad labeling, or feedback loops where the model’s outputs change what data gets collected next. The ethical risk is real: unfair outcomes, discrimination, and loss of trust-especially in systems that affect hiring, credit, healthcare, or policing.

Your Turn (Bias Case Study: Short essay or presentation on bias in AI)

Pick one real-world example you’ve seen (or describe a plausible one like “a screening tool that rates candidates lower because of biased historical data”). Then write:

• The likely source of bias: ______
• Who gets harmed (and how): ________
• One mitigation strategy (data, process, or evaluation): ______

Completion line: Your write-up is 180-300 words and includes all three bullets above.

Module I - Foundations: Mitigation Strategies You Can Actually Use

Mitigation usually means you change inputs (better training data), change the process (clearer labeling rules), and change evaluation (test across groups and edge cases). It’s not “fix once, done forever”-it’s “measure, improve, repeat.”

Reflect (Fill-in lines with a specific target)

One bias check I can run on a project is: ______

The metric or evidence I’d want is: ______

You’re building the language and the safety mindset that your later prompts and projects will depend on. Next, you’ll connect these foundations to how LLMs work in practice, then start shaping outputs with prompt engineering.