Introduction to Artificial Intelligence Concepts

3. Thinking Rationally

This approach is based on the Laws of Thought. It uses formal logic and notation to determine correct arguments and thought processes. Drawing from the traditions of Aristotle, this method seeks to define the rules of "correct" reasoning. However, a major challenge here is that not all intelligent behavior is mediated by logical deliberation, and calculating every possible logical thought can be computationally overwhelming.

4. Acting Rationally

This is the most common focus in modern AI engineering. It centers on Rational Agents—entities that act to achieve the best possible outcome. Acting rationally means "doing the right thing," which is defined as performing actions that maximize goal achievement given the available information.

Exam Tip: Remember that "acting rationally" does not always require "thinking." For example, a biological blinking reflex is a rational action to protect the eye, but it does not involve complex cognitive deliberation.

Understanding Rational Agents

In technical terms, an agent is an entity that perceives its environment and acts upon it. We can represent an agent mathematically as a function that maps a history of perceptions to a specific action:

$$\begin{aligned} f : P^* \rightarrow A \end{aligned}$$

Where $P^\*$ represents the history of percepts and $A$ represents the actions. An agent is considered truly intelligent if it meets four criteria:

1. Its actions are appropriate for its circumstances and goals.
2. It is flexible to changing environments and goals.
3. It learns from experience.
4. It makes appropriate choices despite its computational and perceptual limitations.

The AI Hierarchy: AI, ML, and DL

It is a common mistake to use "AI," "Machine Learning," and "Deep Learning" interchangeably. In reality, they exist in a hierarchical relationship:

• Artificial Intelligence (AI): The broad umbrella encompassing all methods used to develop intelligent machines.
• Machine Learning (ML): A subset of AI that focuses on systems that learn from data and algorithms rather than being explicitly programmed for every task.
• Deep Learning (DL): A specialized subset of ML that uses multi-layered neural networks to process large, complex datasets.

Foundations and Ethics

The development of AI is multidisciplinary, drawing from Philosophy (logic and the mind), Mathematics (algorithms and probability), Psychology (experimental techniques), Neuroscience (the physical substrate of mental activity), and Economics (rational decision theory).

As AI becomes more integrated into society—from facial recognition to autonomous vehicles—Machine Ethics becomes critical. We must consider how to design systems that can make moral decisions, even though machines lack human "feelings" or "consciousness". This involves navigating different ethical frameworks:

• Deontological Ethics: Judging actions based on whether they follow certain rules or duties.
• Consequentialist Ethics: Judging actions solely based on their outcomes or consequences.

Discussion Point: If an AI-powered car must choose between hitting a pedestrian or swerving and risking the passenger, should it follow a rule (Deontology) or calculate which outcome results in the least total harm (Consequentialism)?