Artificial Intelligence: Explore How Machines Learn, Reason and Act

Artificial Intelligence is the field of computer science concerned with developing systems capable of performing tasks associated with intelligent behavior. In this handout, we will examine the main types of Artificial Intelligence, distinguishing Narrow AI, Artificial General Intelligence and Artificial Superintelligence. We will compare Symbolic AI, based on explicit knowledge and rules, with machine learning, where systems learn useful regularities from data. Finally, we will investigate AI and consciousness, exploring the Turing Test, weak and strong AI, the Chinese Room argument, and the distinction between regulation, cognition, understanding and subjective experience.

Introduction to Artificial Intelligence

Introduction to Artificial Intelligence Download

Key points

Artificial Intelligence (AI) = branch of computer science that develops systems capable of performing tasks normally associated with intelligent behavior.

Intelligence = ability to acquire and use knowledge, process information, solve problems, adapt behavior, and pursue goals in different contexts.

Automation ≠ autonomy ≠ adaptivity ≠ learning: automation executes procedures, autonomy reduces continuous human intervention, adaptivity changes behavior, and learning modifies performance through experience.

ANI (Artificial Narrow Intelligence) = AI specialized in specific tasks; AGI (Artificial General Intelligence) = hypothetical general-purpose intelligence; ASI (Artificial Superintelligence) = hypothetical intelligence exceeding human abilities.

ANI ≠ AGI ≠ ASI: the distinction concerns the breadth, transferability, and generality of cognitive abilities, not merely computational power or task-specific performance.

Symbolic AI = approach based on explicitly represented knowledge, logical rules, and inference procedures, offering interpretability but limited flexibility beyond predefined knowledge.

Data-driven AI = approach that learns statistical regularities from training data, enabling generalization to new cases without explicitly programming every rule.

Symbolic AI ≠ data-driven AI: the former applies explicitly defined rules, while the latter learns relationships from data and depends on training quality.

Neuro-symbolic AI → combines learning from data with explicit knowledge representation and symbolic reasoning to pursue flexibility, consistency, and interpretability.

AI conceptions → thinking humanly, thinking rationally, acting humanly, and acting rationally; each evaluates intelligence according to different processes or outcomes.

Thinking humanly ≠ acting humanly: the former models human cognitive processes, while the latter reproduces observable human behavior without necessarily sharing its mechanisms.

Thinking rationally ≠ acting rationally: the former derives logically valid conclusions, while the latter selects actions appropriate for achieving goals.

Rational behavior ≠ human behavior: a system can achieve goals effectively without imitating human reasoning, just as human-like behavior does not guarantee rational decisions.

Turing Test = behavioral evaluation of whether a machine can be distinguished from a human through conversation; passing it does not establish genuine understanding.

Weak AI hypothesis ≠ strong AI hypothesis: the former does not infer genuine mental states from intelligent behavior, while the latter allows that artificial systems might possess them.

Chinese Room argument → demonstrates Searle’s objection that correctly manipulating symbols according to rules does not, by itself, establish semantic understanding.

Syntax ≠ semantics: syntax concerns rules for manipulating symbols, while semantics concerns their meaning; correct symbolic processing does not necessarily imply understanding.

Structuralist hypothesis = interpretive framework proposed in the handout that examines internal organization, information integration, state maintenance, and environmental interaction when evaluating cognitive properties.

Regulation ≠ cognition ≠ consciousness: regulation maintains conditions, cognition supports representation and goal-directed processing, while phenomenal consciousness involves subjective experience.

Operational autonomy, cognitive competence, consciousness, and social impact are distinct properties; none automatically establishes the presence of the others.