INTRO TO ARTIFICIAL INTELLIGENCE
AlphaGo vs. (2016)
Copyright By PowCoder代写 加微信 powcoder
Scene from AlphaGo documentary,
Progress in self-driving cars
Testing of Cruise autonomous vehicle, by Dllu
https://en.wikipedia.org/wiki/Cruise_(autonomous_vehicle)#/media/File:Cruise_Automation_Bolt_ EV_third_generation_in_San_Francisco.jpg
Amazon Go: A “just walk out” store
Artificial Intelligence becoming mainstream
• Industrial Revolution → major increase in manufacturing productivity
• AI Revolution → promises major increase in knowledge work productivity
Currently AI research is receiving significant attention and investment, Estimated 133 million new jobs globally by 2022 (Dwivedi et al, 2019)
Artificial Intelligence becoming mainstream
• Industrial Revolution → major increase in manufacturing productivity
• AI Revolution → promises major increase in knowledge work productivity
Currently AI research is receiving significant attention and investment, Estimated 133 million new jobs globally by 2022 (Dwivedi et al, 2019) However, we have seen ‘AI winters’ before…
Goals of the field of AI research
• Understand what intelligence is, and • Build intelligent systems
What is intelligence?
Ingredients of an intelligent entity:
• Communication
• Internal knowledge, memory
• World knowledge, model of the world • Intentions, and plans to achieve them • Creativity
Definitions of AI
• “[The automation of] activities that we associate with human thinking, activities such as decision-making, problem solving, learning …” (Bellman, 1978)
• “The study of computations that make it possible to perceive, reason, and act.” (Winston, 1992)
• “The study of how to make computers do things at which, at the moment, people are better.” (Rich and Knight, 1991)
Nuances in definition of AI
• Human ideal:
Thinking humanly
Acting humanly • Rational ideal:
Thinking rationally Acting rationally
Acting humanly: Turing test (1950)
Judge C uses written responses to questions to decide whether A or B is the computer
Requires knowledge, reasoning, language, learning By definition an empirical test to judge success
Turing test, A read
https://plato.stanford.edu/entries/turing- test/
Acting rationally
Rational behaviour: always doing the best thing
Acting rationally
Rational behaviour: always doing the best thing
Where best means: that which maximizes the expected performance on the goal, given all available information.
Acting rationally
Rational behaviour: always doing the best thing
Where best means: that which maximizes the expected performance on the goal, given all available information.
Limits on rational behavior: CPU time, memory capacity
Bounded rationality: do the best possible under given resources
Acting rationally
Rational behaviour: always doing the best thing
Where best means: that which maximizes the expected performance on the goal, given all available information.
Limits on rational behavior: CPU time, memory capacity
Bounded rationality: do the best possible under given resources This unit covers algorithms that target rational behaviour
Nuances in the path to AI
• ‘Strong’ AI
Build a machine that is actually thinking
Top-down, philosophical perspective: does my program think? Create general strategy to intelligence
One program that can do any task. • ‘Weak’ AI
Build a machine that acts as if it were intelligent
Like a submarine acts as if it can swim
Bottom-up, goal-driven perspective: does my program work? Create domain-specific strategies:
a chess program,
a train scheduling program,
a language translation program, an image classification program
Nuances in the path to AI
• ‘Strong’ AI
Build a machine that is actually thinking
Top-down, philosophical perspective: does my program think? Create general strategy to intelligence
One program that can do any task. • ‘Weak’ AI
Build a machine that acts as if it were intelligent
Like a submarine acts as if it can swim
Bottom-up, goal-driven perspective: does my program work? Create domain-specific strategies:
a chess program,
a train scheduling program,
a language translation program, an image classification program
More AI Definitions
AI Definitions by Prof. from Stanford University Human-Centered AI. shorturl.at/cqrJZ
Scope of FIT3080
Problems and approaches in designing intelligent software
• Abstracting problems to allow solving them
Representing states of the world, properties of environments (PEAS)
Scope of FIT3080
Problems and approaches in designing intelligent software
• Abstracting problems to allow solving them
• Finding the best sequence of decisions to take
Search algorithms (A∗), planning algorithms (value iteration)
Scope of FIT3080
Problems and approaches in designing intelligent software
• Abstracting problems to allow solving them
• Finding the best sequence of decisions to take
• Learning how to perform well in an unknown environment
Reinforcement learning algorithms (Q-learning, Deep Q-Networks)
Scope of FIT3080
Problems and approaches in designing intelligent software
• Abstracting problems to allow solving them
• Finding the best sequence of decisions to take
• Learning how to perform well in an unknown environment
• Reasoning about information to derive new knowledge Logic and probabilistic (Bayesian) reasoning
Scope of FIT3080
Problems and approaches in designing intelligent software
• Abstracting problems to allow solving them
• Finding the best sequence of decisions to take
• Learning how to perform well in an unknown environment • Reasoning about information to derive new knowledge
• Generalizing to unseen data from historic examples Principles of machine learning
程序代写 CS代考 加微信: powcoder QQ: 1823890830 Email: powcoder@163.com