ITE 153: Intro to Artificial Intelligence

• Introduction to Artificial Intelligence & Expert Systems is the subject of ITE 153 in the 2nd Semester of AY 2024-2025.
• The instructors are Dr. Lumer Jude Doce from the IT & Physics Department.

Key Questions About AI

• What defines artificial intelligence?
• How did the concepts of AI come into existence historically?
• What is the current state-of-the-art situation for AI technologies?
• What is the likelihood of robots dominating the world in the future?

Definitions of Artificial Intelligence

• AI definitions can be categorized into those focused on thinking humanly, thinking rationally, acting humanly, and acting rationally.
• "The exciting new effort to make computers think ... machines with minds, in the full and literal sense.” (Haugeland, 1985)
• "[The automation of] activities that we associate with human thinking, activities such as decision-making, problem solving, learning . . .” (Bellman, 1978)
• “The art of creating machines that perform functions that require intelligence when performed by people.” (Kurzweil, 1990)
• "The study of how to make computers do things at which, at the moment, people are better." (Rich and Knight, 1991)
• "The study of mental faculties through the use of computational models.” (Charniak and McDermott, 1985)
• "The study of the computations that make it possible to perceive, reason, and act.” (Winston, 1992)
• "Computational Intelligence is the study of the design of intelligent agents." (Poole et al., 1998)
• “AI ...is concerned with intelligent behavior in artifacts.” (Nilsson, 1998)

Core Concept of AI

• Artificial Intelligence (AI) is the science and engineering focused on creating intelligent machines, particularly intelligent computer programs.
• AI is considered a field within computer science, with an emphasis on automating intelligent behaviors and actions.

Thought Processes and Reasoning

• Some definitions of AI refer to the study of mental faculties through computational models.
• AI studies the computations to perceive, reason, and act.

Evaluating and Acting Humanly

• Measuring AI success can be based on how well it mirrors human performance.
• Definitions focus on creating machines that perform functions requiring intelligence when done by humans

Rationality in AI

• A system is considered rational if it performs the "right thing,” given its available knowledge.

Expert Systems

• These are computer programs designed to solve complex decision problems within specific domains by replicating the thought processes of human experts.
• Expert systems leverage AI technologies to simulate the judgment and actions of experts, often complementing human capabilities rather than replacing them.

Expert System Architecture

• Non-expert User interacts with a User Interface by creating a query.
• The User Interface sends this to an Inference Engine.
• With the data the Inference Engine uses a Knowledge Base to make an Advice.
• There is also an Explanation Module for explain-ability
• The Knowledge Base is continually added to via a Knowledge Acquisition Module
• The entire system serves as an Expert.

Computational Rationality in AI

• Humans are deemed intelligent based on how predictably their actions achieve set objectives.
• Machines are considered intelligent if their actions predictably meet their objectives, following principles such as minimizing cost, maximizing utility, or minimizing loss.
• AI aims to create computational rational agents, aligning with logically defined goals.

Designing Rational AI systems

• An agent can perceive and act within an environment

Brains vs. AI

• Human brains are good as making rationale decisions, but are not perfect
• Lessons learned from brains are memory, knowledge, feature learning, procedure formation, and simulation.

A Short History of AI

• Philosophy, Mathematics, Neuroscience, Economics, Control theory, Psychology, and Linguistics all form the prehistory of AI.
• Near Miss (1842): Babbage designs for universal machine and Lovelace proposes a 'thinking machine'
• AI had its official birth at Dartmouth in 1956.

Key Eras in AI History

• 1940-1950: Early days of AI included the McCulloch & Pitts Boolean circuit model of the brain (1943) and Turing's discourse on "Computing Machinery and Intelligence" (1950).
• 1950-70: AI saw excitement with early programs for chess and checkers, and the Dartmouth meeting in 1956 adopted "Artificial Intelligence."
• 1970-90: Knowledge-based approaches, expert system development, and industry booms.
• 1990-2012: Statistical approaches and subfield expertise mark a period with a resurgence of probability and technical depth, leading to an "AI Spring.".
• 2012-: the period of big data, computation, neural networks, and renewed excitement in AI's capabilities and industry applications.

AI Key Eras and Milestones

• 1950s-1960s saw the development of early problem-solving programs like the General Problem Solver (GPS).
• The 1970s-1980s marked the rise of expert systems, focusing on rule-based reasoning.
• Since the 1990s, AI has seen the emergence of machine learning, big data, and breakthroughs in neural networks.
• IBM Deep Blue defeating world chess champion Garry Kasparov (1997)
• Google DeepMind's AlphaGo defeating world Go champion Lee Sedol (2016)

Foundations of AI

• Philosophical foundations: Rationalism, dualism
• Mathematical foundations: Logic, probability theory, statistics
• Other Disciplines: Game theory, utility maximization, Neuroscience, Linguistics

Approaches to AI

• Symbolic AI (Good Old-Fashioned AI) focuses on explicitly encoding rules and knowledge.
• Expert systems like MYCIN for medical diagnosis serve as an example of symbolic AI.
• Statistical AI relies on probabilistic methods and machine learning, utilizing decision trees, support vector machines, and neural networks.
• Connectionist AI deploys artificial neural networks modeled on the brain for image recognition, natural language processing, & generative AI.
• Hybrid Approaches are the combination of multiple approaches.

AI in Action: Industry Applications

• Healthcare is improved by AI-driven diagnostics, improved surgeries, and patient treatments.
• Finance is improved by predictive analytics, fraud detection, algorithmic trading
• Transportation is improved by autonomous vehicles, and real-time traffic prediction.

AI in Action: Consumer Applications

• Virtual assistants (e.g., Alexa, Siri, Google Assistant).
• Recommendation systems (e.g., Netflix, Spotify, Amazon).

Scitific Contributions

• Protein structure prediction (e.g., AlphaFold).
• Climate modeling and renewable energy optimization.

Areas of Natural Language Processing

• Speech Recognition, and Text-to-speech processing.
• Question and answering systems

Vision Perception

• Enables face and object detection Semantic Scene Segmentation is possible via AI

Al Everywhere

• AI is used in Search engines, Route planning, Logistics, Medical diagnosis, help desks, Spam / fraud detection, cameras, Product recommendations, and smart homes

The Future of AI

• Create AI systems.
• To magnify the benefits from it
• Net present value of human-level AI ≥ $13,500T

Ethical Implications

• Algorithms may reflect and amplify societal biases due to biased training data.
• Transparency: Many Al models, especially neural networks, act as black boxes, making their decision-making processes difficult to interpret.
• Accontability: Is it the Developer, Companies, or the systems themselves?

Societal Implications

• Job displacement: Humans replaced by machines, robotics in industries like transportation, manufacturing, and retail.
• Large scale data collection and survelliance.

Algorithmic Bias

• Value misalignment: It is important for AI to align with societal values
• The machine's only objective is to maximize the realization of human preferences
• The robot is initially uncertain about what those preferences are
• Human behavior provides evidence about human preferences
• Still to human-level AI are understanding of language, integration of learning with knowledge, and cumulative discovery og concepts with theories

Al Landscape

• The rise of democratized and open-source Al tools.
• The use of explainable AI (XAI) to ensure trust.
• Generative AI Models have been revolutionizing content creation, graphic design, and software development.