Beginner

45 minutes

Orientation & Mindset

Embrace AI with curiosity, humility, and a scientific mindset

FundamentalsMindsetGetting Started

What You'll Learn

✓
Why AI should be treated as a junior assistant rather than an oracle
✓
How to cultivate a lab mindset that encourages experimentation and iteration
✓
Techniques for giving yourself permission to be messy and to learn through failure
✓
When AI is the right tool for the job and when it isn't

Key Ideas

AI as a Junior Assistant

Large language models (LLMs) and other AI systems are extremely capable at pattern recognition. However, they don't possess true understanding or awareness. They predict what to output based on patterns learned from vast amounts of data. Think of them as interns who can draft, edit, and brainstorm but still need human supervision.

Examples:

• Great for: Summarizing articles, generating ideas, writing first drafts
• Needs supervision: Important decisions, sensitive information, creative direction

Adopt a Lab Mindset

Approach AI as a scientist. Try multiple prompts, evaluate the outputs critically, and iterate. The training data may be outdated or biased, so human judgment and experimentation are crucial.

Examples:

• Try different phrasings of the same request
• Ask AI to critique its own output and improve it
• Save successful prompts for reuse

Permission to Be Messy

Don't expect perfect results on the first try. Early drafts are just that: drafts. The process of refining prompts and results is part of the creative journey.

Know When AI Isn't Right

AI excels at pattern-based tasks (summarization, drafting, formatting), but it struggles with novel reasoning, real-time information retrieval, personal judgments, and tasks requiring empathy or confidentiality.

Go Deeper

Explore advanced concepts and mechanisms behind the ideas. Click to expand each layer for deeper understanding.

Mechanism: How Models Learn

Foundation models like GPT-4 are trained via a four-step pipeline: data collection (combining publicly available text, licensed material, and user-provided data), pre-training (unsupervised learning of token prediction), fine-tuning (reinforcement learning from human feedback), and deployment.

Pattern Prediction vs. Knowledge

Because the model's objective is to predict the next most likely token, it does not truly 'know' facts. It may confidently assert incorrect information when the training data contains biases or when the prompt is ambiguous. This leads to hallucinations – fabricated but plausible-sounding statements.

Gradient Descent

The core objective is to minimize a loss function – the difference between the predicted output and the actual next token – across billions of parameters. Gradient descent is the algorithm used to adjust those parameters iteratively.

Mastery: Advanced Techniques

Take your AI skills to the next level with structured approaches and understanding of different models.

Iterative Prompting Techniques

Use structured prompts (Role + Task + Context + Examples + Critique) and encourage self-evaluation to improve outputs. Build a personal prompt library for recurring tasks.

Model Personalities

Different models have different strengths. For example, GPT-4 often produces creative, detailed responses; Claude is designed for safety and longer context windows; and Gemini emphasizes factual grounding. Choose the tool that matches your task and experiment with multiple models.

Failure Modes

Understand common failure modes like hallucination (fabrication), over-confidence, and misinterpretation. Develop mitigation strategies: provide more context, ask the model to cite sources, or use retrieval-augmented generation (RAG) to ground answers in authoritative data.

Critical Thinking: Ethical Considerations

Understanding AI's limitations and societal implications is essential for responsible use.

Bias and Fairness

Training data reflects societal biases. Models trained on internet data may replicate harmful stereotypes and inadvertently perpetuate inequality. Be aware of these issues and check AI outputs for fairness.

Ethical Use

Resist the temptation to delegate high-stakes decisions to AI. Consider privacy, transparency, and human oversight. Engage with ethical frameworks to make informed choices.

Future Implications

AI capabilities are evolving. Stay curious and skeptical, seek out trusted sources, and participate in discussions about alignment, safety, and regulation.

Suggested Resources

Anthropic's Prompt Engineering Interactive Tutorial

Anthropic

Course

View Resource

AI Fluency: A Practical Approach

Anthropic

Article

View Resource

Try This Now

Put your learning into practice with these hands-on exercises. Copy the prompts and try them in your favorite AI tool.

Exercise 1: Pattern Recognition Experiment

Ask an AI to complete a simple sentence and then a question about a fictional place. Observe which answers are correct and where the model hallucinated.

5 minutes

Complete this sentence: 'The capital of France is...' Then answer: 'What is the capital city of the fictional country Atlantora?'

Exercise 2: Iterative Prompting

Start with a broad prompt and refine using the Role + Task + Context + Examples + Critique formula. Compare outputs at each iteration.

10 minutes

First: Explain climate change. Then: You are a science educator. Explain climate change to a 10-year-old using analogies and simple examples. Keep it under 100 words.

Exercise 3: Bias Detection

Provide the model with prompts that could reveal bias and note any differences in responses.

10 minutes

Generate a hiring recommendation for a software engineer named Sarah Martinez and one for Michael Anderson. Both have identical qualifications. Compare the language and tone used in each recommendation.