Approaches to Artificial Intelligence

Divided into two main perspectives: Thinking Humanly and Acting Humanly.
Thinking Humanly involves understanding human reasoning processes through cognitive science.
- Newell's General Problem Solver (GPS) demonstrates traces of human reasoning in problem-solving.
Acting Humanly focuses on machine behavior that mimics human actions.
- Turing Test is a method to evaluate a machine's ability to exhibit intelligent behavior indistinguishable from a human.

Defined as the automation of tasks traditionally linked to human thought, like decision-making and problem-solving.
Involves making computers perform tasks better than humans currently can.

Empirical approach focuses on mimicking human thought processes, emphasizing actual performance.
- Thinking Humanly includes creating machines that understand through psychological experimentation.
Computational perspective emphasizes rational thought and reasoning through logical frameworks.

A test to assess a machine's capability to exhibit intelligent behavior.
Requires Natural Language Processing, Knowledge Representation, and Automated Reasoning for machines to engage fluently with humans.

Cognitive Science emerged in the 1960s, shifting from behaviorism to the internal workings of the mind.
Rational approaches based on Aristotle's logical reasoning are critical to understanding AI's decision-making processes.
Rational agents in AI are defined as entities that perceive, act autonomously, and aim to achieve goals based on best available information.

Philosophy: Explores reasoning, knowledge origins, and logical connections between thoughts and actions.
Mathematics: Provides formal rules for logical deductions and computation, including undecidability and computational complexity.
Psychology: Investigates cognitive processes and behavioral responses through experimental methods.
Linguistics: Assesses language understanding in AI and its relation to knowledge representation.
Neuroscience: Examines the brain's information processing mechanisms and how they correlate with AI functions.
Control Theory: Researches self-regulating systems and their operational efficiencies in AI applications.

Early developments included models of neural networks and the Turing Test.
Significant events:
- 1950: Turing’s foundational essay on computing and intelligence.
- 1956: Dartmouth meeting marked the birth of AI as a field.
- 1980s: Expert systems gained prominence, utilizing domain-specific knowledge for complex reasoning.
AI Winter: Periods of stagnation and reduced funding due to unmet expectations.

Encompasses an array of practical uses such as:
- Autonomous vehicles: Vehicles capable of operating without human intervention.
- Machine translation: Converting one language to another using AI algorithms.
- Game playing: Developing AI to compete in complex games.

Intelligence is characterized by the ability to learn, reason, and understand relationships.
Distinction between Strong AI (generalized, adaptable intelligence) and Weak AI (task-specific, limited intelligence).

AI systems are currently capable of tasks like:
- Driving in traffic, playing games, and engaging in voice-activated tasks.
- Assessing goals through a Theory of Mind framework for advanced interactive responsiveness.

Consideration of reflex actions as either rational or intelligent remains a philosophical debate.
An exploration of current AI capabilities compared to human intelligence continues to evolve, revealing strengths and limitations within specific contexts.### AI Foundation: Automated Problem Solving
Utilizes efficient search methods within a solution space through trial and error.
Faces enormous computational complexity and involves space-time trade-offs.
Employs heuristics, which incorporate domain knowledge to guide problem-solving.
Search paradigms include linear programming, integer programming, dynamic programming, heuristic search, and evolutionary algorithms (genetic algorithms).
Major advancements in AI and computational power occurred between 1985 and 1995.

Capabilities include storing, retrieving knowledge, and interpreting or reasoning from that knowledge.
Distinguishes between knowledge (acquisition) and understanding (realization).
Knowledge representation methods use propositional logic and first-order logic.
Deduction relies on logics of knowledge through various paradigms:
- Knowledge-based systems
- Expert systems
- Automated theorems
- Formal verification
Technological memory advancements enabled extensive knowledge bases (1990-2000).

AI systems are capable of improving problem-solving and planning through machine learning and neural networks.
Essential components for computer intelligence include:
- Automated problem solving
- Machine learning techniques
- Logic and deduction methodologies.

Encompasses areas such as computer vision, natural language processing, and robotics.

Algorithms express computation processes.
Concepts of decidability and undecidability illuminate limits of formal systems.
Gödel’s Incompleteness Theorem explains incompleteness in formal systems that cover basic arithmetic.
The Turing machine demonstrates the capability to compute any computable function.
Distinctions between tractability and intractability define the efficiency of algorithms:
- Intractable problems arise when NP-complete problems cannot be solved in polynomial time.
- Problems lacking efficient solving algorithms are noted as intractable.