Lecture 1-Intro to AI Flashcards
What are 3 recent breakthroughs in AI?
-Speech Recognition & Machine Translation –>2010+
-Image Recognition & Computer Vision–>2012+
-Natural Language Processing–>2014+
What are the two approaches to AI?
-Engineering approach
-Cognitive approach
What is the engineering approach?
-Tries to find optimal solutions
-No matter how–> not necessarily what human do
What is the cognitive approach?
-Tries to understand the process
-Tries to reproduce human behavior EVEN IF WRONG RESULT
What are the 4 points of view?
Cognitive:
-Behavior-> act like humans
-Reasoning->think like humans
Engineering:
-Behavior-> act intelligently
-Reasoning->think intelligently
What is considered weak AI(3)?
-System whose capabilities are not intended to match or exceed those of human beings.
-CAN demonstrate intelligence, BUT no need for mind or consciousness
-SMALL APPLICATION WITH SMALL PURPOSE
What is considered a strong AI(4)?
-used in science fiction
-System that matches or exceeds human intelligence
-SHOULD HAVE consciousness, sentiments, etc.
-GENERAL PURPOSE CAPABLE OF SEVERAL TASKS
What are 4 types of agents?
-Intelligent Agent Continuum
-A Simple Reflex Agent
-More Sophisticated Reflex Agent
-Learning Agent
Name 3 characteristics of the intelligent agent continuum?
-Agents receive percepts from environment
-Agent performs actions
-Each agent performs a function to map percept sequences to actions
Name 3 characteristics of the more sophisticated reflex agent?
-A bit smarter than simple-reflex version
-Has the ability to internally model the outside world
-Not at the mercy of what can at the moment be directly sensed
Name 3 characteristics of the learning agent?
-sense the environment
-criticize with a standard in mind
-provide feedback to update learning elements
What is the Turing test?
-It measures the intelligence of a computer vs a human
-If a human interrogator cannot tell the computer and human apart, then the computer is intelligent
What are 4 examples of capabilities required to pass the Turing test?
- TO COMMUNICATE: Natural Language processing(NLP)
-SHARE KNOWLEDGE: Knowledge representation
-STORE KNOWLEDGE: Automated reasoning
-Machine learning
What are arguments FOR the turing test(3)?
-Objective notion of intelligence
-Prevents us from arguments about the computer’s consciousness
-Eliminates bias in favor of humans
What are arguments AGAINST the turing test(2)?
-Not reproducible
-Not constructive
What does CAPTCHA stand for?
Completely Automated Public Turing test to tell Computers and Humans Apart
What does the CAPTCHA do?
-System asks a user to complete a test which the computer is able to generate and grade, but not able to solve.
-Because computers are unable to solve the CAPTCHA, any user entering a correct solution is presumed to be human.
Why is the CAPTCHA known as reverse turing test?
-its given by a machine and targeted to a human and not the opposite like the turing test
What are 3 concrete tasks examples of the Modern Bot Evaluation Frameworks?
1.Natural Language Understanding(NLU): understanding + processing human language
2..Creative Tasks: AI’s ability to generate creative content
3.Decision-Making Skills: AI’s effectiveness in scenarios requiring complex decision-making
What are 3 metrics examples of the Modern Bot Evaluation Frameworks?
1.Accuracy and precision: measure of correctness in outputs
2.Response time: speed at which AI provides responses
3.Robustness and generalization: ability to handle unexpected inputs or scenarios
What do we do in AI(8)?
-Knowledge representation (formal logic)
-Search (especially heuristic search)
-Planning
-Reasoning under uncertainty (probabilistic reasoning)
-Learning
-Agent architectures
-Robotics and perception
-Natural language processing
What are examples of successes of AI(3)?
-Optical Character recognition(OCR)
-Speech recognition
-Spell checker and grammar checker
What’s a more pragmatic def of AI today?
AI research is that which computing scientists do not know how to do cost-effectively today.
What are the years of creation of AI?
1940-1956:
1943: early work in neural netowrks
1950: alan turing describes the turing test
1956: the darmouth workshop
What is the darmouth workshop?
-Get-together of the big guys: McCarthy, Minsky, Shannon,etc.
-The term AI is first adopted
What does the rise of AI(1956-70s) consist of?
-GOFAI-> Good OId Fashioned AI
-LISP
-PROLOG
What is GOFAI(2)?
-Symbolic computation rather than numeric computation
-Higher level of abstraction
What is LISP(3)?
-Functional language (recursive functions)
-McCarthy 1958
-Gave rise to SCHEME, ML,…
What is PROLOG(3)?
-Logical language (first-order predicate logic)
-Colmeraurer 1972 for NL understanding
-Based on Robinson’s resolution method for theorem proving
What are AI programming languages today?
-General-purpose programming languages like C,C++,Java,…
-Current #1 for AI: Python
Why are we using general-purpose programming languages today?
Because of the decline in symbolic computations in AI
When was the first major of AI winter?
Late 60s-early 70s
When did expert systems start and end?
-Start: 1970s-1980s
-End : mid 80s-mid 90s
What are expert systems?
-Knowledge-intensive, rule-based techniques
-Commercial expert systems
-Decision-support systems
-Humans need to write the rules by hand
Why did expert systems end?
-Too tedious to write rules by hand
-Too expensive to maintain
When did Machine Learning rise?
-1980s-2010
-Rules are now learned automatically
Why Machine Learning?
-More powerful CPUs->usable implementation of neural networks
-Big data->huge data sets are available
When did Deep Learning rise?
-2010-today
Why Deep Learning?
-Trained on massive data sets
-Use of GPU for computations
-Use of generic networks for many applications(image recognition, self driving cars…)
What is Eliza?
-Joseph Weizenbaum 1960s
-Simulation of a dialogue with a psychotherapist
-Great success at the time
-Example of a simple production system
