1.4 Goals and Challenges of AI (BT104CO)
1. Primary Goals of AI
The goals of AI can be divided into Short-term (Applied) and Long-term (Scientific) objectives.
A. Developing Rational Agents
The Engineering Goal: Systems that maximize performance by perceiving and acting on their environment.
- Problem Solving: Navigating complex state spaces.
- Perception: Understanding sensory input.
- Learning: Improving without explicit reprogramming.
B. Understanding Human Intelligence
The Scientific Goal: Building models to uncover the fundamental principles of human cognition.
- Cognitive Science: Testing theories about brain processes like memory and language.
C. Artificial General Intelligence (AGI)
A machine capable of performing any intellectual task a human can do, featuring human-level versatility across all domains.
2. Major Challenges of AI
Despite rapid progress, several fundamental challenges remain. These are the "Hard Problems" mentioned throughout the textbook.
The "Black Box" Problem
Explainability is critical in fields like medicine or law. We need to understand the why behind a decision to ensure safety and fairness.
Combinatorial Explosion
Real-world state spaces are vast. Searching through billions of possibilities requires efficient Heuristics to guess the best path.
Data Dependency & The Long Tail
AI requires massive datasets and often fails on "Edge Cases" (rare events) that weren't captured during training.
The Alignment Problem
Ensuring AI goals align with human values is difficult. A robot might pursue a task with unintended, destructive efficiency.
Commonsense Reasoning
Humans have deep "background" knowledge of the physical world that is almost never explicitly recorded in datasets.
3. Comparison of AI Goals
| Goal Type | Focus | Outcome |
|---|---|---|
| Narrow AI | Specific Task | Excellent at one thing (e.g., AlphaGo). |
| General AI (AGI) | Cross-Domain | Human-level versatility across all tasks. |
| Superintelligence | Mastery | Exceeding human performance in all areas. |