Introduction to Logic in AI

TL;DR

Logic in AI is about making computers "reason" by representing knowledge as facts and rules. It helps AI systems understand situations and make decisions based on what they know. You'll learn how to break down complex problems into logical steps an AI can follow.

1. The Mental Model

Think of logic in AI like a detective's notebook. You gather clues (facts), write down general rules (if X happens, then Y is true), and then use those to figure out the truth of new situations or answer questions.

2. The Core Material

At its heart, logic in AI is about formalizing reasoning. We take human ideas and translate them into a precise language that computers can understand and manipulate. This lets AI systems derive new information from existing information.

We usually start with Propositional Logic, which deals with simple statements (propositions) that are either true or false.

2.1 Propositions

A proposition is a declarative sentence that is definitively true or false.
* "It is raining." (Could be true or false)
* "The sky is blue." (True)
* "2 + 2 = 5." (False)

2.2 Logical Connectives

Photo by Mike van Schoonderwalt on Pexels

We combine propositions using connectives:
* AND ($\land$): Both statements must be true.
* Example: "It is raining and it is cold."
* OR ($\lor$): At least one statement must be true.
* Example: "You can have coffee or tea."
* NOT ($\neg$): Reverses the truth value.
* Example: "Not (It is raining)."
* IMPLIES ($\implies$ or $\rightarrow$): If the first is true, then the second must be true.
* Example: "If it is raining, then the ground is wet."
* IF AND ONLY IF ($\iff$ or $\leftrightarrow$): Both statements have the same truth value.
* Example: "You will pass the test if and only if you study."

2.3 Truth Tables

Photo by Markus Winkler on Pexels

Truth tables show all possible truth values for a compound proposition based on the truth values of its simple propositions. They're how we check if a logical statement is true or false under different conditions.

Here's an example for P AND Q:

2.4 Inference

The real power of logic is inference – deriving new truths from existing ones. If we know "If it's raining, the ground is wet" and "It's raining," we can infer "The ground is wet." This is a fundamental rule called Modus Ponens.

Here's a simple flow of how logic helps an AI system reach a conclusion:

graph TD
    A["Known Facts/Observations"] --> B("Apply Logical Rules")
    B --> C{"Is a Conclusion Reached?"}
    C -- "Yes" --> D("New Fact/Decision")
    C -- "No" --> A

2.5 Limitations of Propositional Logic

Photo by Ann H on Pexels

While useful, propositional logic can't represent relationships or general statements about objects. For that, we need First-Order Logic (also known as Predicate Logic), which introduces predicates (properties or relationships), variables, and quantifiers (like "for all" or "there exists"). We'll cover that later, but it's important to know there are more powerful forms of logic.

3. Worked Example

Let's say we have an AI system managing a simple smart home.
Facts:
1. Is_Home = True (You are home)
2. Time_Is_Night = True (It's nighttime)

Rules:
* Rule 1: Is_Home AND Time_Is_Night AND NOT(Lights_On) IMPLIES Turn_On_Lights
* (If you're home AND it's night AND the lights aren't on, THEN turn on the lights.)
* Rule 2: Is_Home AND Time_Is_Night AND Has_Guests IMPLIES Play_Music
* (If you're home AND it's night AND you have guests, THEN play music.)

Question: Will the AI turn on the lights if Has_Guests = False?

Step-by-step reasoning:
1. Check Rule 1:
* Is_Home is True.
* Time_Is_Night is True.
* We don't know Lights_On, but let's assume NOT(Lights_On) is True for now (lights are off).
* The condition True AND True AND True is True.
* So, Turn_On_Lights is implied.

2. Check Rule 2:
   • Is_Home is True.
   • Time_Is_Night is True.
   • Has_Guests is False.
   • The condition True AND True AND False is False.
   • So, Play_Music is not implied by this rule.

Conclusion: Based on the rules and facts, the AI will turn on the lights. The Has_Guests fact doesn't affect the first rule.

4. Key Takeaways

• Logic in AI provides a formal way for machines to reason and make decisions.
• Propositions are statements that are strictly true or false.
• Logical connectives (AND, OR, NOT, IMPLIES, IFF) combine propositions into more complex statements.
• Truth tables systematically show the truth values of compound propositions.
• Inference rules, like Modus Ponens, allow AI to deduce new facts from existing knowledge.

Common Mistakes to Avoid:
- Don't confuse logical "OR" with everyday "either/or" (logical OR allows both).
- Be precise with your definitions of propositions; they must be unequivocally true or false.
- Misinterpreting "IMPLIES" – P IMPLIES Q is only false if P is true and Q is false.
- Overlooking the need for more expressive logic (like First-Order Logic) when simple propositions aren't enough.

5. Now Try It

You're building a simple AI for a gardening system.
Facts:
* Soil_Dry = True
* Temperature_High = True
* Sun_Shining = False

Rules:
1. IF Soil_Dry AND Temperature_High THEN Water_Plants
2. IF NOT(Sun_Shining) AND Temperature_High THEN Open_Shade

Based on these facts and rules, what actions will your gardening AI take? Write down your step-by-step reasoning for each rule.

Success looks like: You correctly identify which actions (if any) the AI will take and justify your answer by evaluating the truth of each rule's conditions based on the given facts.