What is Artificial Intellegence and what is it not?
The science and engineering of making intelligent machines, especially intelligent computer programs
– John McCarthy
Artificial Intelligence and (Deep) Machine Learning are not the same thing
Artificial Intelligence have developed into different research fields
AI can be referred to, as the science, that uses computers to understand human intelligence but it does not have to limit itself to that topic
What is Machine Learning?
The science "concerned with the question of how to construct computer programs that automatically improve with experience
What is Deep Learning?
A form of machine learning that makes use of artificial neural networks
What four different criteria can be combined to define and categorize AI?
Thought processes/reasoning vs. behavior/action
Success according to human standards vs. success according to an ideal concept of intelligence: rationality
What is needed for AI-based systems which should think humanlike?
Understanding on how humans think
Cognitive Science –> combines computational models from AI and experimental techniques from psychology
Cognitive Neuroscience –> how does the brain work on a neuronal level
What is needed for AI-based systems which should think rational?
Laws of thought –> resoning and logic
What are the problems of systems that think rational?
Logic is hard to formalize –> Feigenbaum Bottleneck
Practical constraints –> lack of computational power
Not every form of intelligence is mediated to logic behavior
How can it be tested if a machine acts humanlike?
Turing Test/Imitation Game
A judge can communicate with two anonimous inputs –> one input is a human the other a computer
If the computer can successfully
fool the judge, it has passed the test and
can be considered intelligent
Advantage: Instead of defining a long and controversial list of necessary prerequisites for intelligence. Only compare to undeniably intelligent beings → humans
Disadvantage: not reproducible, constructive or amenable to mathematical analysis
What is needed for systems that should act rational?
Systems should do the “right thing”, which is implemented as the goal of a task
The problem is, that perfect rationality is only achievable in a perfect environment
What is an algorithm?
Algorithms are often used in the field of AI
An algorithm is a step-by-step procedure for finding a solution to a class of problems
deterministic –> the sequence of steps taken for the same problem is always the same
terminates after finite time –> you should get a result after some time
correct –> should always get you the right result
In which applications is AI already used?
AI supports people in need
AI does the household
AI drives cars
AI knows a lot –> Jeopardy
AI is an Artist
AI plays chess and GO
What are the most important subfields of AI?
Problem Solving and Search
Game Playing
Knowledge Representation
Planning
Machine Learning
Deep Learning
Neural Networks
Natural Language Understanding
Computer Vision
Robotics
Deep Learning is a sub-discipline of Machine Learning
According to John McCarthy, AI has to confine itself to biologically observable methods since it is related to using computers to understand human intelligence?
The ”Imitation Game” by Alan Turing focusses on which of the following definitions of AI?
What is the difference (from the perspective of AI research) between rational thinking and high-level human-like thinking (i.e. that researched by symbolic AI field)
Deep learning has been particularly successful in pattern recognition
Three of the six people were 2019 Turing Awardees
There are some definitions of artificial intelligence, organized into four categories. Match them correctly!
Defintion
Category
”The study of the computations that make ist possible to perceive, reason, and act.” (Winston, 1992)
Systems that think like humans
”AI ... is concerned with intelligent be- havier in artifacts.” (Nilsson, 1998)
Systems that think rationally
”The art of creating machines that perform functions that require intel- ligence when performed by people.” (Kurzweil, 1990)
Systems that act like humans
”The exciting new effort to make com- puters think ... machines with minds, in the full and literal sense.” (Hauge- land, 1985)
Systems that act rationally
Map the following important contributions to the topic of what Artificial Intelligence is.
Person
Contribution
Tom Mitchell
Described what Machine Learning is
Al-Chwarizmi
The Turing Award was named after him
John McCarthy
One of the first people who defined what Artificial Intelligence is
Alan Turing
Was the first person who formalized algorithms
What is NOT a problem of Systems that think rationally?
The basic idea of the Turing Test is to compare computers with undeniably intelligent beings (humans), instead of defining a long and controversial list of necessary prerequisites for intelligence
Which statements about the ”Turing test” are correct?
Deep Learning is a form of machine learning that makes use of artificial neural networks?
Machine Learning is the science ”concerned with the question of how to construct computer programs that automatically improve with experience”. (Tom Mitchell)
Which statements about ”Algorithms” are correct?
The Turing Test is meant to evaluate if a model can think rationally
An algorithm has only two main characteristics. These are that it is deterministic and that it terminates in finite time.
Which programming language did John McCarthy invent?
Complete the sentences to the topic ”Current Trend in AI”.
”Intelligent prostheses“ are tools ...
... were developed in the 1980s
Use strong scientific standards ...
... and not on solving AI as a whole
Exploit the strengths of the computer ...
... and do not try to model human thought processes.
Focus on particular tasks ...
... that support us in tasks that would otherwise require human.
.... that support us in tasks that would otherwise require human.
Strong scientific standards ...
Map the following definitions with their own system
”Computational Intelligence is the study of the design of intelligent agents” (Poole et al., 1998)
”The art of creating machines that perform functions that require intel- ligence when performed by people” (Kurzweil, 1990)
”The exiting new effort to make com- puters think ... machines with minds, in the full and literal sense” (Hauge- land, 1985)
”The study of mental faculties though the use of computational models” (Charniak and McDermott, 1985)
“The study of how to make computers do things which, at the moment, people are better”. (Rich and Knight, 1991)
“[The automation of] activities hat we associate ith human thinking, activities such as decision-making, problem solving, learning…” (Bellman, 1978)
Please map the concepts with their definition
”It is the science and engineering of making intelligent machines ... and does not have to confine itself to me- thods that are biologically observa- ble.” -John McCarthy, Stanford(1956)
It is a field that bring together com- putational models from AI and experi- mental techniques from psychology to consutruct precise and testable theo- ries of the mind
Cognitive Science
It is the science ”concerned with the question of how to construct compu- ter programs that automatically im- prove with experience.” -Tom Mitchell (1997) CMU
Artificial Intelligence
What must be considered when saying that AI is about building clever machines that show rational behavior?
It does not mean that the methods which are used have to be particularly clever
It also does not mean that machines must use the same methods as humans do
Last changeda year ago