What is Artificial Intelligence?

Jr Codex AI Notes

Level: Beginner Prerequisites: None Time to complete: ~20 minutes

Table of Contents

1. Defining Artificial Intelligence
2. Four Ways to Define AI
3. The Turing Test
4. What AI Is Not
5. The Goals of AI
6. Where AI Shows Up Today
7. Summary & Next Steps

1. Defining Artificial Intelligence

Artificial Intelligence (AI) is the field of building systems that perform tasks which, when done by humans, would be said to require intelligence — reasoning, learning, perceiving, and decision-making.

A Working Definition
─────────────────────────────────────────
  AI: the design and study of AGENTS that perceive their
  environment and take ACTIONS to achieve GOALS.
─────────────────────────────────────────

This definition is deliberately broad — it covers a chess program, a spam filter, a self-driving car, and a large language model, all of which are genuinely different technologies but share this common structure of perceive → decide → act. Chapter 4 formalizes this "agent" idea precisely.

2. Four Ways to Define AI

AI textbooks traditionally organize definitions along two axes: whether a system focuses on thinking or acting, and whether it's judged against human performance or an ideal standard of rationality.

                Human-like                    Rational
              ─────────────────────────────────────────
  Thinking    │ Systems that THINK like        │ Systems that THINK
              │ humans (cognitive modeling)     │ RATIONALLY (logic-based
              │                                   │ reasoning systems)
              ─────────────────────────────────────────
  Acting       │ Systems that ACT like            │ Systems that ACT
               │ humans (the Turing Test,           │ RATIONALLY (agents that
               │ Section 3)                           │ choose the BEST action
               │                                        │ to achieve their goals)
              ─────────────────────────────────────────

This curriculum leans toward "acting rationally" — it's the dominant framing in modern AI, since it doesn't require a system to mimic human quirks or limitations, only to make good decisions given what it knows.

3. The Turing Test

Proposed by Alan Turing in 1950 (further explored in Chapter 2's history), the Turing Test asks: can a machine's conversation convince a human judge that they're talking to another human, rather than a machine?

The Turing Test Setup
─────────────────────────────────────────
  Judge (human)
      │
      ├──── text conversation ────  Player A (human)
      │
      └──── text conversation ────  Player B (machine)

  The judge doesn't know which is which, and must decide.
  If the judge can't reliably tell them apart, the machine
  is said to have "passed" the test.
─────────────────────────────────────────

Why the Turing Test is influential, but limited as a real goal:

Criticisms of the Turing Test
─────────────────────────────────────────
  - Optimizes for FOOLING a human, not for genuine understanding
    or capability — a system could pass by cleverly evading
    hard questions rather than reasoning well
  - Says nothing about tasks that AREN'T conversational
    (driving a car, diagnosing an X-ray, playing chess)
  - Modern large language models complicate the picture further —
    convincing conversation is now achievable without most
    researchers considering it "true" general intelligence
─────────────────────────────────────────

The Turing Test remains historically important and a useful thought experiment, but modern AI research mostly evaluates systems on specific, measurable tasks (covered throughout this curriculum) rather than chasing "passing as human" directly.

4. What AI Is Not

A few common misconceptions worth clearing up immediately, since they shape expectations for the rest of this curriculum:

Misconception                          Reality
─────────────────────────────────────────
  "AI means robots"                       Most AI has no physical body at
                                            all — it's software (a
                                            recommendation engine, a
                                            search algorithm, a chatbot)

  "AI = Machine Learning"                    ML is ONE approach to building
                                              AI (learning patterns from
                                              data) — but AI also includes
                                              search (Module 2), logic
                                              (Module 3), and planning,
                                              none of which require
                                              "learning" from data at all

  "AI systems truly understand              Most current AI performs
   what they're doing"                       sophisticated PATTERN MATCHING
                                              and OPTIMIZATION — whether
                                              this constitutes genuine
                                              "understanding" is a deep,
                                              unresolved philosophical
                                              question, not a settled fact
─────────────────────────────────────────

The AI/ML distinction matters enormously for this curriculum: Modules 2-3 (search, knowledge representation, planning) are classical AI techniques that predate and don't require machine learning at all. Chapter 5 of this module maps out exactly how AI, ML, Deep Learning, and NLP/LLMs relate to each other.

5. The Goals of AI

What AI Research Aims to Build
─────────────────────────────────────────
  Reasoning        → drawing valid conclusions from known facts (Module 3)
  Knowledge          → representing what a system "knows" about the world (Module 3)
  Planning             → figuring out a sequence of actions to reach a goal (Module 3)
  Learning               → improving performance from experience/data (ML notes)
  Perception               → interpreting sensory input — vision, speech (DL/NLP notes)
  Natural Language           → understanding and generating human language (NLP/LLM notes)
  Acting/Manipulation           → taking physical or digital actions in an environment
─────────────────────────────────────────

No single AI system needs all of these — a chess engine needs planning and search but no language ability; a chatbot needs language processing but little physical manipulation. Real-world AI systems typically combine a handful of these goals for a specific purpose.

6. Where AI Shows Up Today

# A tiny taste of "AI" as a decision-making system — no machine learning needed
def spam_filter(email_text: str) -> bool:
    """A simple rule-based AI: acts on a percept (email text), returns a decision."""
    spam_keywords = ["win money", "click here now", "free prize"]
    return any(keyword in email_text.lower() for keyword in spam_keywords)
 
print(spam_filter("Click here now to win money!"))     # True — flagged as spam
print(spam_filter("Meeting moved to 3pm tomorrow"))       # False — not spam

This is intentionally a trivial example — but it satisfies the definition from Section 1: it perceives (reads the email), and acts (returns a decision) toward a goal (filtering spam). Real spam filters use much more sophisticated techniques (often machine learning), but the underlying agent structure is identical, and Chapter 4 builds this idea out fully.

7. Summary & Next Steps

Key Takeaways

• AI is the study of agents that perceive their environment and act to achieve goals — a definition broad enough to cover both a 1990s chess engine and a modern chatbot.
• AI definitions traditionally split along two axes: thinking vs. acting, and human-like vs. rational — this curriculum leans toward "acting rationally," the dominant modern framing.
• The Turing Test asks whether a machine's conversation can fool a human judge — historically important, but a limited and non-comprehensive measure of real AI capability.
• AI is broader than machine learning — classical techniques like search, logic, and planning (Modules 2-3) build genuine AI systems without any learning from data at all.

Concept Check

1. Why does this curriculum favor the "acting rationally" framing of AI over "acting humanly"?
2. What's a key criticism of using the Turing Test as a primary measure of AI progress?
3. Why is "AI = Machine Learning" a misconception, and what's a concrete counterexample?

Next Chapter

→ Chapter 2: History of AI

Jr Codex — 1-on-1 Personalized Coaching | Back to Module Index