AI on Tech Notes by ClaudeXu

AI 101

Sun, 18 Jan 2026 21:35:08 +0800

What is AI?

Some tasks are easy for computers to perform, such as:

Identifying the maximum number in a dataset
Calculating the shortest route from your office to the subway station

For these types of problems, we can provide clear, step-by-step instructions that computers execute reliably. However, other tasks are much harder to solve with traditional programming:

Is this person happy?
Is that a cat in this image?

While these questions might seem straightforward for humans to answer, they’re surprisingly difficult to program explicitly. Artificial Intelligence (AI) refers to systems that can perform tasks typically requiring human intelligence — such as recognizing patterns, understanding language, and making decisions based on complex data.

Rather than following rigid instructions, AI systems learn from examples and experience, allowing them to handle tasks that are difficult or impossible to solve with conventional programming alone.

What is Machine Learning?

Machine learning is a subset of AI where computers learn patterns from data rather than following explicit instructions. A machine learning system analyzes examples, identifies patterns, and uses what it has learned to make predictions or decisions about new, similar data.

The key difference from traditional programming:

Traditional Programming: You provide explicit rules → the computer follows them → produces output
- Analogy: Following a recipe step-by-step to cook a meal
Machine Learning: You provide examples (data) → the computer finds patterns → creates a model that can handle new situations
- Analogy: Tasting a dish at a restaurant and figuring out the ingredients and techniques (recipe) to recreate it at home

For instance, instead of writing rules like “if the email contains the word ‘prize’ and asks for personal information, it’s spam,” a machine learning system learns what spam looks like by analyzing thousands of examples of spam and legitimate emails, then identifies patterns we might never have thought to program explicitly.

AI, Machine Learning & Deep Learning

You’ve probably heard these terms used interchangeably, but they’re actually nested concepts:

Artificial Intelligence (AI) is the broadest concept — any technique that enables computers to mimic human intelligence
Machine Learning (ML) is a subset of AI—systems that learn from data without being explicitly programmed
Deep Learning (DL) is a subset of Machine Learning — uses neural networks with multiple layers to learn complex patterns

Example to illustrate the difference:

AI: A chess-playing computer (could use rule-based logic or learning)
Machine Learning: A spam filter that improves by analyzing thousands of emails
Deep Learning: Image recognition that identifies cats in photos by learning from millions of images through neural networks

AI vs. Data Science

While AI and Data Science both work with data and often use similar tools, they have different primary goals:

Aspect	AI	Data Science
Primary Goal	Enable machines to perform intelligent tasks (perception, reasoning, decision-making)	Extract insights and knowledge from data to inform decisions
Focus	Automating intelligent behavior	Understanding patterns and answering questions
Key Techniques	Machine learning, neural networks, computer vision, natural language processing	Statistics, data analysis, machine learning, visualization
Typical Output	A system that performs tasks (chatbot, recommendation engine, autonomous vehicle)	Insights, reports, predictions, dashboards
Example Applications	Virtual assistants (Siri, Alexa), self-driving cars, facial recognition	Customer churn analysis, sales forecasting, A/B testing, market research

The overlap: Data scientists often use AI/ML techniques, and AI systems require data science skills to prepare and analyze data. Many professionals work at the intersection of both fields.

Common AI Terms Explained

As AI evolves rapidly, new terminology emerges constantly. Here are some current buzzwords simplified:

LLM (Large Language Model) - AI systems trained on massive amounts of text that can understand and generate human - like language. Examples: ChatGPT, Claude, GPT-4. Think of them as extremely well-read assistants that can write, answer questions, and help with various language tasks.
Agent - An AI system that can take actions autonomously to achieve goals, not just respond to questions. Unlike a basic chatbot that only answers when asked, an agent can plan multiple steps, use tools, and complete complex tasks with minimal human supervision. Example: an AI that researches a topic, writes a report, and emails it to you.
MCP (Model Context Protocol) - A standard way for AI systems to connect with external tools and data sources. It’s like giving your AI assistant the ability to check your calendar, access your files, or pull data from various apps—all through a unified interface.
Skill - A specialized capability or function that an AI system can perform, often by connecting to specific tools or following particular workflows. Think of skills as apps for your AI—each one teaches it how to do something specific like analyzing spreadsheets, generating images, or searching databases.
RAG (Retrieval-Augmented Generation) - A technique where AI retrieves relevant information from a knowledge base before generating a response, making answers more accurate and up-to-date. It’s like giving the AI a library to reference instead of relying only on what it memorized during training.
Fine-tuning - Customizing a pre-trained AI model for specific tasks or domains by training it further on specialized data. Like teaching a general doctor to become a heart surgeon — the foundation is there, but you add expertise in a particular area.

MCP vs. Skill: What’s the Difference?

These terms are related but serve different purposes:

Aspect	MCP (Model Context Protocol)	Skill
What it is	The connection standard or “language” that lets AI communicate with external systems	The actual capability or know-how that AI uses to accomplish specific tasks
Analogy	The USB port on your computer	The USB device you plug in (printer, camera, keyboard)
Technical role	Infrastructure/protocol layer	Application/capability layer
Example	The protocol that allows Claude to connect to Google Drive	The “document editing skill” that knows how to read, analyze, and modify Word files
Focus	How AI connects to tools	What the AI can do with those tools

In practice: An AI system might use MCP to connect to various services (email, calendar, databases), and then use different skills to actually work with the data from those services — like a “meeting scheduler” skill or “data analysis” skill. MCP is the bridge; skills are what you accomplish after crossing it.

Machine Learning 101

Sun, 18 Jan 2026 21:35:08 +0800

What is Machine Learning?

Refer to this link

ML is NOT the solution to every problem. If simple rules work, use them instead.

Main types of ML

Supervised learning - you have data and labels. The algorithm learns by predicting labels and correcting mistakes.
- Classification - categorizing samples (binary for 2 options, multi-class for 3+)
- Regression - predicting numbers (e.g., “how many users will subscribe?”)
Unsupervised learning - you have data but no labels. The algorithm finds patterns, and you interpret them. Example: clustering customers into “summer buyers” and “winter buyers” for targeted promotions.
Transfer learning - reusing a trained model for a new task (e.g., adapting a car recognition model to identify dog breeds). Valuable because training models from scratch is expensive.
Reinforcement learning - the algorithm learns by trial and error within defined rules, earning rewards or penalties. Example: teaching an AI to play chess by updating its score based on moves.

Matching your problem

Supervised learning - you know inputs and outputs
Unsupervised learning - you have inputs but uncertain outputs
Transfer learning - your problem resembles an existing one

Typical Machine Learning Flow

Problem → Data → Evaluation → Features → Modeling → Experiments

Problem - What are we trying to solve?
Data - What data do we have?
Evaluation - What does success look like?
Features - Which variables should we feed into the model?
Modeling - Which model fits our problem best?
Experiments - What else can we try to improve results?

Evaluation

Evaluation defines what success looks like. Common metrics include:

Accuracy - How often is the model correct overall?
$Accuracy = \frac {TP + TN} {TP + TN + FP + FN}$
Precision - When the model predicts positive, how often is it right?
$Precision = \frac {TP} {TP + FP}$
Recall - Of all actual positives, how many did the model catch?
$Recall = \frac {TP} {TP + FN}$

When to prioritize each:

High precision - Use when false positives are costly. Example: spam filters (don’t want important emails marked as spam).
- The model acts more deliberately, so when it reports positive, it is very likely to be actual positive.
High recall - Use when false negatives are costly. Example: cancer detection (missing a case is worse than a false alarm).
- The modle becomes more sensitive. It may reports False Positive, but it tries its best to catch all ‘potential’ positives.

Modeling

Modeling is the core of the ML workflow, consisting of three stages:

Training - train the model on data
Validation - tune the model’s parameters
Test - verify the model’s performance

The dataset is split accordingly:

Set	Split
Training	70% – 80%
Validation	10% – 15%
Test	10% – 15%

A key goal is generalization — a good model performs well on data it has never seen before, producing similar results across all three sets.

Watch out for these two failure modes:

Data Set	Underfitting	Overfitting
Training	62%	95%
Test	50%	60%

Underfitting - poor accuracy across the board; the model is too simple for the problem.
Overfitting - high training accuracy but poor test accuracy; the model memorized the training data instead of learning the underlying pattern.

Commonly Used Tools

Core Python Libraries:

NumPy - numerical computing with arrays and matrices
Pandas - data manipulation and analysis
Matplotlib/Seaborn - data visualization

ML Frameworks:

Scikit-learn - classical ML algorithms (regression, classification, clustering)
TensorFlow - Google’s deep learning framework
PyTorch - Facebook’s deep learning framework, popular in research
Keras - high-level neural network API (now multi-backend with TensorFlow, PyTorch and JAX)

Development Tools:

Jupyter Notebook - interactive coding environment
Google Colab - free cloud-based Jupyter notebooks with GPU access
Anaconda/Miniconda - Python distribution with package management

Optional but Popular:

MLflow - experiment tracking and model management
Hugging Face - pre-trained models for NLP tasks