AI: AI 是什么

AI (Artificial Intelligence)
├── Machine Learning (ML)
│   ├── Supervised Learning
│   ├── Unsupervised Learning
│   └── Reinforcement Learning
├── Deep Learning (DL)
│   ├── CNN (Computer Vision)
│   ├── RNN / Transformer (NLP)
│   └── GAN / Diffusion (Generation)
├── Natural Language Processing (NLP)
├── Computer Vision (CV)
└── Robotics

# Traditional approach: you write the rules
def is_spam_traditional(email):
    """Rule-based spam detection"""
    spam_keywords = ["free", "winner", "click here", "urgent"]
    for keyword in spam_keywords:
        if keyword in email.lower():
            return True
    return False

# AI approach: machine learns the rules from data
# from sklearn.naive_bayes import MultinomialNB
# model = MultinomialNB()
# model.fit(training_emails, training_labels)  # Machine learns rules
# prediction = model.predict(new_email)         # Machine applies learned rules

# Example: loading and inspecting a dataset
import pandas as pd

# Load a CSV dataset (e.g., house prices)
df = pd.DataFrame({
    'area_sqm': [50, 80, 120, 65, 200, 90],
    'rooms': [2, 3, 4, 2, 5, 3],
    'location_score': [7, 8, 6, 9, 5, 7],
    'price_usd': [150000, 280000, 350000, 220000, 500000, 260000]
})

print(df.describe())

        area_sqm    rooms  location_score     price_usd
count    6.000000  6.00000        6.000000  6.000000e+00
mean    84.166667  3.16667        7.000000  2.933333e+05
std     51.968585  1.16905        1.414214  1.196533e+05
min     50.000000  2.00000        5.000000  1.500000e+05
max    200.000000  5.00000        9.000000  5.000000e+05

Algorithm + Data → Training → Model

Example:
Linear Regression + House Price Data → Training → Price Prediction Model
(trained weights: price = 2500 * area + 15000 * rooms - 5000)

# Example: Adding AI to your app is just a few lines of code
# (This requires an OpenAI API key, covered in Lesson 10)

# from openai import OpenAI
# client = OpenAI(api_key="your-api-key")
# 
# response = client.chat.completions.create(
#     model="gpt-4o-mini",
#     messages=[
#         {"role": "system", "content": "You are a helpful coding assistant."},
#         {"role": "user", "content": "Explain recursion in Python in 2 sentences."}
#     ]
# )
# print(response.choices[0].message.content)

# ============================================
# Compare: Rule-based vs Machine Learning
# Task: Spam email detection
# ============================================

# --- Approach 1: Traditional rule-based program ---
def is_spam_rules(email_text):
    """Detect spam using hand-written rules"""
    spam_words = ["free", "winner", "click", "urgent", "congratulations"]
    spam_count = sum(1 for word in spam_words if word in email_text.lower())
    return spam_count >= 2  # Arbitrary threshold

# --- Approach 2: Machine Learning (simulated) ---
# In real code, we would train a model on thousands of labeled emails.
# Here we simulate the trained model's behavior.
def is_spam_ml(email_text):
    """Detect spam using a trained ML model (simulated)"""
    # A real model learns patterns like:
    # - ALL CAPS subject lines → likely spam
    # - Multiple exclamation marks → likely spam
    # - Specific sender domains → likely spam
    # - Word combinations humans wouldn't think of
    learned_spam_score = 0.85  # Simulated confidence
    return learned_spam_score > 0.5

# --- Test both approaches ---
test_emails = [
    "Congratulations! You are our lucky winner! Click here to claim your FREE prize!",
    "Hey, can we reschedule tomorrow's meeting to 3 PM?",
    "URGENT: Your account will be closed! Click here immediately!",
]

print("=== Rule-based vs ML Spam Detection ===")
for email in test_emails:
    rule_result = is_spam_rules(email)
    ml_result = is_spam_ml(email)
    print(f"Email: {email[:60]}...")
    print(f"  Rule-based: {'Spam' if rule_result else 'Ham'}")
    print(f"  ML model:   {'Spam' if ml_result else 'Ham'}")
    print()

=== Rule-based vs ML Spam Detection ===
Email: Congratulations! You are our lucky winner! Click here to cl...
  Rule-based: Spam
  ML model:   Spam

Email: Hey, can we reschedule tomorrow's meeting to 3 PM?...
  Rule-based: Ham
  ML model:   Spam

Email: URGENT: Your account will be closed! Click here immediately!...
  Rule-based: Spam
  ML model:   Spam