1.1 Definitions of Intelligence and AI (BT104CO)
1. What is Intelligence?
In the context of AI, intelligence is not defined as a single spark of consciousness, but as a computational process.
The Rational Definition
Intelligence is the study of Rational Agency. A system is intelligent if it perceives its environment and takes actions that maximize its chances of success, given its goals and the available information.
Intelligence = Perception + Reasoning + Action
2. Defining Artificial Intelligence (AI)
Russell and Norvig categorize AI definitions into four historical approaches, organized by whether they focus on thought processes vs. behavior, and human-like performance vs. rational performance.
A. Thinking Humanly (Cognitive Modeling)
Focus: Internal "mind" and decision-making.
Goal: To build machines that solve problems like humans.
Validation: Requires experimental evidence from psychology/neuroscience.
B. Acting Humanly (Turing Test)
Focus: External behavior and output.
Goal: To perform functions requiring intelligence when done by people.
Validation: The Turing Test benchmark.
C. Thinking Rationally (Laws of Thought)
Focus: Formal Logic (Syllogisms).
Goal: To use logical patterns to reach "right" conclusions.
Problem: Hard to turn "fuzzy" real-world knowledge into rigid logic.
D. Acting Rationally (Rational Agent)
Focus: Goal achievement and behavior.
Goal: Operating to achieve the best outcome (or best expected outcome).
Modern Standard: This is the most widely accepted definition in CS.
3. AI vs. Natural Intelligence (NI)
While both involve processing information to solve problems, their substrates and characteristics differ significantly.
| Feature | Artificial Intelligence (AI) | Natural Intelligence (NI) |
|---|---|---|
| Substrate | Silicon chips, digital circuits | Carbon-based, biological neurons |
| Speed | Extremely fast (nanoseconds) | Relatively slow (milliseconds) |
| Learning | Massive datasets; "copyable" | Continuous; individual-unique |
| Consistency | High; does not get "tired" | Subject to emotions and fatigue |
| Efficiency | High consumption (GPU clusters) | Extremely efficient (~20 Watts) |
| Durability | Permanent (Weights can be stored) | Perishable (Dies with organism) |
4. Key Exam Concepts
- Rationality vs. Omniscience: AI focuses on Rationality (doing the best with what you know), not Omniscience (knowing everything).
- The Goal of AI: To find the optimal or satisficing (sufficient) solution using computational means.
- Turing Test Requirements: Natural Language Processing, Knowledge Representation, Automated Reasoning, and Machine Learning.