CCAI 221: AI definitions and approaches

• CCAI 221 is about AI fundaments
• The notes have been assembled from various sources including a textbook by Russel and Norvig, a website and Inst. Ebtehal Alsaggaf

What is AI? Definitions and Approaches

• Intelligence includes the computational ability to achieve goals
• Intelligent behavior involves acting in complex environments, learning, thinking, reasoning, perceiving relations, and using tools
• Varying kinds and degrees of intelligence are observed in people, animals, and even machines
• Artificial Intelligence is the science and engineering of making intelligent machines, according to John McCarthy
• Patrick Henry defines AI as the study of ideas that make people intelligent and incorporating them into computers
• AI includes studying how computers can perform tasks better done by people, according to Elaine Rich
• Another definition of Al is getting computers to do tasks that require human intelligence

Four Main AI Approaches

• Turing Test approach seeks to make machines act like humans
• Cognitive modeling approach attempts to make machines think like humans
• The Laws of thought approach aims at making machines think rationally
• The rational agent approach focuses on building machines that act rationally
• The ultimate goal of these approaches is to construct autonomous intelligent machines

Intelligence Tests and AI Laws

• Computers excel at quickly solving difficult problems, like computing GCD or solving complex integration problems
• Humans elegantly solve simple problems, such as navigating a busy street or recognizing voices and images
• Easier an issue is for a human, the harder it is for AI to tackle and vice versa

The Turing Test

• Involves an interrogator who communicates with a person and a computer without knowing which is which
• The interrogator determines if the computer can fool them into thinking it is a human
• The human tries to convince the interrogator they’re human
• If the computer succeeds in fooling the interrogator, it passes the intelligence test

Necessary Computer Capabilities

• Natural language processing: To communicate effectively
• Knowledge representation: To store and retrieve information
• Automated reasoning: To use stored information to answer questions and draw new conclusions
• Machine learning: To adapt to new circumstances and extrapolate patterns
• Computer vision: To recognize actions and presented objects
• Robotics: To manipulate objects and move around

Thinking Humanly

• Simulate human-like thinking in machines
• Introspection includes catching one's own thoughts
• Observing a person in action constitutes psychoological experiments
• Observing the brain in action includes brain imaging
• The focus is on the human reasoning process
• Cognitive science combines computer AI models and experimental techniques from psychology

Thinking Rationally

• The objective is systems that think rationally
• Focus on logical rules and inference mechanisms
• Guarantee an optimal solution

Acting Rationally ("Rational Agent" Approach)

• Rational behavior includes doing the right thing
• Rational agents operate autonomously, perceive their environment, persist, and adapt
• Act to achieve the best outcome or the best-expected outcome under uncertainty
• Acts sufficiently, if not optimally, in all situations
• The rational-agent approach is more general than the "Laws of thought" approach
• The rational agent approach is more suitable for scientific development than approaches based on human behavior

Foundations of AI

• Philosophy: Minds operate as physical systems following logical rules
• Establish the source of knowledge
• The mind is the connection between knowledge and action
• Mathematics provides theories of logic, computation, and formal representation
• Economics contributes utility and decision theory
• Neuroscience studies the nervous system and neurons that lead to thought, action, and consciousness
• Psychology views the brain as an information-processing device
• Computer engineering builds powerful computers
• Linguistics focuses on language's meaning and structure; Computational linguistic and NLP

History of AI

• The beginning of Al: The first steps were characterized by Boolean circuits in 1943 and Turing's work on "Computing Machinery and Intelligence" in 1950
• The birth of Al took place in 1956 at the Dartmouth meeting, where the term "Artificial Intelligence" was adopted
• Early enthusiasm characterized by AI programs like Samuel's Checkers, Newell & Simon's Logic Theorist, and Gelernter's Geometry Engine
• A dose of reality (1966-1973) showed AI's computational complexity, causing neural network research to decline
• AI became an industry in 1980
• Expert systems industries boomed between 1980-88
• Neural networks returned to popularity in 1985
• AI adopted the scientific method in 1987
• The emergence of intelligent agents began in 1995
• Big data played an important part in 2001
• 2003 - HI back on agenda
• Deep Learning played an important part in 2011 due to speech and visual object recognition

The state of the art in AI Development

• In 2005, the robotic car named STANLEY completed a 132-mile desert track at 22 mph in the DARPA challenge
• In 2018, Waymo test vehicles traveled 10 million miles on public roads with human driver taking over control only every 6000 miles
• In the air, autonomous fixed-wing drones have been providing cross-country blood delivery in Rwanda since 2016
• Quadcopters are exploring buildings while constructing 3D maps
• BigDog can move in irregular terrain and recovers when slipping on an icy puddle
• Atlas walks, jumps, and backflips on uneven terrain
• During the Gulf War, and US forces deployed AI to schedule 50,000 vehicles, cargo, and people
• Deep space is navigated by the SEXTANT system
• Online machine translation systems translate documents in over 100 languages
• Virtual assistant AIs offer suggestions
• Systems help diagnose various conditions using AI algorithms
• Ride-hailing companies are using current and future predicted traffic conditions
• AI can handle voice recognition
• AI algorithms now equal or exceed expert doctors diagnosing conditions based on diagnostic images
• Machine learning can also be a climate science solution
• Deep Blue defeated Garry Kasparov in chess (1997)
• AI beat human champions in the Jeopardy game (2010)

Risks and Benefits of AI

• AI can free humanity from menial repetitive work
• AI can increase the productions of goods and services
• AI can help with finding cures for diseases
• AI can help with finding solutions for climate changes
• Risks include lethal autonomous weapons, surveillance, biased decision making, safety-critical applications, and cyber security

Examples of AI Applications

• AI translators translate spoken language and convert it to printing in foreign languages
• Natural language understanding (spell checkers, grammar checkers)
• AI can be seen in natural language processing, speech recognition, machine vision and robotics
• Robotics becoming increasingly important to do games and tedious jobs
• Other applications include self-driving cars, diagnostic systems, natural language understanding, and robotics

Explanation of AI foundations based on CCAI 221 course notes, including definitions of AI, intelligent behavior, and various AI approaches. Drawn from multiple sources, including Russel and Norvig's textbook.