Artificial Intelligence: Definitions and Approaches

• The field of Artificial Intelligence (AI) focuses on creating intelligent machines.
• Russel and Norving's textbook "Artificial Intelligence: A Modern Approach" is a resource for this content.
• Further information can be found at http://aima.cs.berkeley.edu/.

What is AI? Definitions and Approaches

• Outlined topics include definitions, approaches, foundations, history, current state, benefits, risks, and applications of AI.
• Intelligence involves the computational ability to achieve goals.
• Intelligent behavior includes acting in complex environments, learning from experience, reasoning, perceiving relations and using tools.
• Intelligence varies in people, animals, and machines.

Measuring AI Success

• AI success can be measured against human performance or against an ideal rationality measure.
• "Thinking humanly" involves creating computers with minds, in the full and literal sense (Haugeland, 1985).
• Acting humanly is the art of creating machines that perform functions requiring intelligence when performed by people (Kurzweil, 1990).
• Thinking rationally involves studying mental faculties through computational models (Charniak and McDermott, 1985).
• Acting rationally involves the design of intelligent agents (Poole et al., 1998).

AI Approaches

• Four main approaches include the Turing Test, cognitive modelling, laws of thought, and a rational agent approach
• The Turing Test approach aims for machines to act like humans.
• Cognitive modelling seeks to make machines think like humans.
• The 'laws of thought' approach focuses on rational thinking.
• The rational agent approach aims for machines to act rationally.
• The goal is to build intelligent, autonomous machines.

Intelligence Testing

• Computers are faster at solving difficult problems like computing GCD (Greatest Common Denominator) and complex integrations.
• Humans can solve simple problems more elegantly, such as navigating a busy street or recognizing a voice.
• "Easy problems are hard and hard problems are easy" is the first law of AI.

The Turing Test

• The Turing Test involves an interrogator communicating with a person and a computer without knowing which is which.
• The computer tries to fool the interrogator into believing it is human.
• To pass the Turing test, a computer needs natural language processing, knowledge representation, automated reasoning, and machine learning.
• Passing a "total Turing test" also requires computer vision and robotics capabilities.

Cognitive Science Approach

• Involves simulating human-like thinking in machines
• Relies on introspection, psychological experiments and brain imaging,
• It involves building machines with human-like minds.
• Cognitive science combines AI computer models with psychology experiments to test theories about the human mind.

Laws of Thought Approach

• It centers on developing systems that think rationally.
• The focus is on using logical rules and inference mechanisms to guarantee an optimal solution.
• Example: if Socrates is a man and all men are mortal, then Socrates is mortal.

Rational Agent Approach

• Focuses on doing the right thing
• An agent acts autonomously, perceives its environment, persists, and adapts.
• A rational agent acts to achieve the best outcome or expected outcome under uncertainty.
• Focus is on systems that act sufficiently, not necessarily optimally, in situations.
• Two advantages over other approached is there is a general focus and leads to scientific development

Foundations of AI

• Philosophy contributes ideas about the mind as a physical system with logical rules and the source of knowledge.
• Mathematics provides theories of logic, computation, and probability.
• Economics informs AI through utility and decision theory.
• Neuroscience studies the nervous system and the relationship between the brain and computers.
• Psychology offers insights through cognitive models of the brain.
• Computer engineering enables the building of powerful computers.
• Linguistics provides an understanding of language structure and meaning, including NLP.

History of AI

• The gestation period (1943-1955) involved Boolean circuit models of the brain and Turing's work on computing machinery.
• The birth (1956) is marked by the Dartmouth meeting where "Artificial Intelligence" was adopted.
• Early enthusiasm (1952-1969) led to programs like Samuel's checkers, and Newell & Simon's Logic Theorist.
• A dose of reality (1966-1973) saw the discovery of computational complexity, and neural network research declined.
• The key to power was knowledge-based systems, (1969-1979) witnessing early development.
• The industry emerges, (1980-present), with expert systems booming.
• The return of neural networks started in 1985.
• Probabilistic reasoning and machine learning began in 1987.
• The emergence of intelligent agents began in 1995.
• AI became a science in 1987.
• In 2003 human-level AI returned.
• Deep Learning began in 2011.

The State of the Art

• In 2005, STANLEY, a robotic car, completed autonomously a 132-mile desert track at 22mph in the DARPA challenge.
• In 2007, robotic vehicles navigated streets with traffic in the Urban challenge.
• Waymo's test vehicles had driven 10 million miles on public roads by 2018, with minimal human intervention.
• Autonomous drones deliver blood in Rwanda since 2016.
• Quadcopters explore buildings while constructing 3D maps.
• BigDog can move in irregular terrain and recover slipping, (2008).
• Atlas walks on uneven terrain and jumps and does backflips, (2016).
• During the Gulf War, Al logistics planning and scheduling program involved as many as 50,000 vehicles, cargo, and people, (1991).
• NASA's on-board program controlled spacecraft scheduling, (2000).
• The SEXTANT system allows autonomous navigation in deep space beyond GPS, (2017).
• AI is now used in daily ride hailing services such as Uber.
• Online translation systems can read documents in over 100 languages.
• Speech recognition is available in assistants such as Alexa, Siri, Cortana, and Google.
• Companies like Amazon, Facebook, Netflix, Spotify, YouTube, and Walmart use machine learning to recommend items.
• Spam filtering is a form of recommendation.
• Deep Blue beat Garry Kasparov, a world chess champion 1997
• Al beat humans in Jeopardy game 2010
• AI algorithms now equal or exceed doctors at diagnosing many conditions (diagnosis is based on images); for example Alzheimer’s disease, cancer etc, (2016).
• Machine learning is used to tackle climate changes from 2018

AI Benefits

• AI can free humanity from menial repetitive work.
• There is dramatic increase the output volumes of goods and services.
• Contribute to find cures for diseases.
• Help to find solutions for climate changes.

AI Risks

• Risks include lethal autonomous weapons, surveillance and persuasion, biased decision-making, and impacts on employment.
• Other risks include safety-critical applications and cybersecurity threats.

Al Application Examples

• Deep Blue is a world chess champion.
• AI is used in self driving cars.
• AI Translators translate spoken and printed languages
• Google and IBM: Natural language understanding, spell checks and grammar checks
• WebMD Symptom Checker is a web medic diagnostic system.
• Robotics is becoming increasingly important in areas like games.

Summary

• AI is about building intelligent machines.
• AI comprises Turing test-based, cognitive science, laws of thought, and rational agent approaches.
• Disciplines that have contributed to AI include psychology, mathematics and linguistics.
• AI has the constant cycles of new approaches and refinement of the new ones.
• Main application areas include game playing, natural language processing, speech recognition, machine vision, robotics and expert systems.
• Consideration of the consequences and all the ethical issues
• Russel & Norving Textbook is a reference.

Explore definitions, approaches, history, and applications of AI. Discover how AI success is measured against human performance and ideal rationality. Learn about thinking and acting humanly in AI, based on Russel and Norvig's work.