Intro to intelligent systems & AI

• Module 1 introduces intelligent systems, exploring the relationship between human and artificial intelligence (AI).
• Key concepts covered are the definition of intelligence, the Four Schools of Thought on AI, and the Turing Test.
• John Searle's argument against AI, the history and success stories of AI, and debates over computers surpassing humans are reviewed.
• The ethical, philosophical, and technical dimensions of AI are assessed.

Intended Learning Outcomes

• Intelligence and AI are defined and differentiated.
• The Turing Test is described as a measure of machine intelligence.
• John Searle's argument and its impact regarding machine intelligence are summarized.
• AI history, with key milestones and developments are outlined.
• AI success stories and applications are identified and analyzed.

What is intelligence

• Intelligence is the capacity to learn and solve problems.
• It is an individual's or system's ability to understand, learn from experience, adapt, and use knowledge and reasoning to solve challenges.

Key aspects of intelligence

• Intelligence involves learning, storing, and effectively applying information, not limited to memorization.
• Tackling unfamiliar or complex problems highlights the ability to adapt and find solutions to unexpected events.
• A rational action aligns with logic and reason where decisions are based on available data and knowledge to maximize outcomes or minimize errors.
• Human intelligence is characterized by complex, social, creative, and adaptive behaviors; AI aspires to replicate human thinking.

Goals of Artificial Intelligence (AI)

• AI aims to understand human intelligence and develop practical systems for real-world problems with scientific & engineering objectives.

Scientific Objective

• Exploring cognitive processes for learning, reasoning, decision-making, perception, and problem-solving is an application of AI.
• Exploring human thought provides insights into the nature of intelligence itself, which informs the development of machines.

Engineering Objective

• Creating tools and systems to solve real-world problems is an engineering goal of AI systems.
• Intelligent agents are systems or machines that can perceive their environment, make decisions, and take actions to achieve specific goals.

Examples where engineering goals are applied

• Knowledge-Based Systems which use structured knowledge to solve specific problems by applying rules and reasoning.
• Natural Language Understanding Systems understanding, interpreting, and generating human language for applications like chatbots.
• To develop Intelligent Robots that are capable of performing complex tasks autonomously or semi-autonomously.
• Speech and Visual Recognition Systems recognize and interpret speech and visual data, which includes applications such as voice-controlled devices.
• Game Playing (e.g., IBM's Deep Blue) where AI can use strategy, prediction, and decision-making where machines outperform humans.

Turing Test Explanation

• The Turing Test assesses a machine's ability to exhibit intelligent behavior indistinguishable from a human.
• In Alan Turing's 1950 paper "Computing Machinery and Intelligence," Turing posed the question: "Can machines think?"
• He reframed the question to "Can machines behave intelligently?"

Key components of the Turing test

• the Imitation Game uses an operational test to determine if a machine can exhibit intelligent behavior.
• It involves a judge interacting via text with a human and a machine, where the machine aims to convince the judge it is human.
• The test focuses on simulating human-like responses, not actual consciousness or understanding.
• Components for AI include knowledge representation, reasoning (making deductions, solving problems), language/image understanding (processing human-like language or images) and learning (improving performance over time)

John Searle's Argument: The Chinese Room

• Searle argues against strong AI with the "Chinese Room" thought experiment, arguing that syntax alone cannot produce semantics.
  • an English speaker is locked in a room and cannot understand Chinese
  • the person manipulates Chinese symbols using English instructions to produce appropriate responses without understanding the symbols' meaning.
  • While they manipulate the symbols according to rules (syntax), not semantically (meaning) to the person from the outside it seems like Chinese is being understood
• argues that simulating understanding isn't actual understanding.
• Deep Blue vs. Kasparov highlights AI challenging and beating top human minds where strategic thinking is required

Strong AI vs Weak AI

• Strong AI is the belief that a machine can understand and have consciousness when processing information like human cognition.
• Weak AI is the belief that machines can simulate intelligent behavior but do not actually understand or have consciousness.

History of AI (Artificial Intelligence)

• McCulloch and Pitts introduced the Boolean circuit model of the brain in 1943 as groundwork for artificial neural networks..
• Alan Turing's 1950 paper set the stage for AI advancements by providing a framework for evaluating machine intelligence.
• The Dartmouth Conference in 1956 which is where the term "Artificial Intelligence" was coined by John McCarthy is marked as the formal birth of AI.

Early AI programs during the 1950's

• Arthur Samuel Checks Program was developed as an early AI that employed machine learning where the strategies improved over time.
• Allen Newell and Herbert Simon developed the Logic Theorist, one of the first AI programs capable of simulating human problem-solving.

1955-1965 "Great Enthusiasm" & Key Developments

• Newell and Simon developed General Problem Solver (GPS) to solve problems using a heuristic approach.
• In 1958, McCarthy Invented LISP which became of the most important AI research languages
• David Gelertner created Gelertner's Geometry Theorem Prover demonstrates the potential for understanding mathematical concepts.

History of AI (1966-1995) summary

• 1966-1973: Reality Dawns: AI problems were more complex, revealing limitations in existing AI methods.
• Intractable Problems are tasks that were computationally expensive
• Early neural network methods had limitations and failed to scale to real-world applications.
• 1969-1985: Adding Domain Knowledge: Domain-specific knowledge was incorporated to improve performance.
• Expert Systems were developed where reasoning was made with if-then rules to makes successful decision.
• 1986-1990: Rise of Machine Learning shifted Al research where data driven approaches was used.
• Backpropagation was revived where more efficient learning from data recognition, language processing, and game playing.
• Machine Learning Improvements where algorithms were adapted to new learning information, with more flexible and powerful tools.
• 1990: The Role of Uncertainty emerged where the focus was on handling uncertainty in aspects of real-world problems
• Bayesian models of probable networks were created
• 1995: AI as Science matured, saw AI techniques integrate it into a wide range of applications.

Deep Blue in 1997

• IBM's Deep Blue defeats Garry Kasparov, demonstrating AI's capacity to challenge and beat top human minds by using brute force

AI 1997 to Beyond

• Al combined with evaluation functions that assessed the positions on the chessboard to decide the best moves.
• In mathematical concepts, AL helped proved unsolved problems.

1991 gulf war

• Al-based logistics planning and scheduling system was utilized and military operations.
• Al helped improve efficiency and was able to handle masssive amounts of data and was time sensitive.

2001: A Space Odyssey (1969) in terms of AI

• HAL(Heuristically Programmed Algorithmic Computer) is an integral part of the spaceship where Stanley Kubrick's science film plays
• HAL a computer system, that is personified as an intelligent where sentinent entities with human-like capabilities
• HAL provides speech communication in a clear measurable voice who communicate calmly and clearly with team members
• The movie presents a vision of emotion, and automatic navigation, where Hal ensures saftey
• Hal is responsible for diagnosing problems, and also the more chilling where Hal makes life/death decisions.
• While complex, this shows there are limitations in AI with the lack of feelings.

AI vs Human Brain

• AI, modern Brains can perform billions of transistors
• Super computers typically use hundreds of CPUs at a time and are high speed

A.L.I.C.E

• Al program that simulates chatbots
• A.L.I.C.E. uses pattern-matching techniques
• Language is translated and responds to humans
• The project can be explored through it's website to download software or chatbots