DIT410/TIN174, Artificial Intelligence
21 March, 2017
”It is not my aim to surprise or shock you – but the simplest way I can summarize is to say that there are now in the world machines that can think, that learn, and that create.
Moreover, their ability to do these things is going to increase rapidly until — in a visible future — the range of problems they can handle will be coextensive with the range to which human mind has been applied.”
by Herbert A Simon (1957)
Weak AI — acting intelligently
Strong AI — being intelligent
Most AI researchers don’t care
Do we want a system that…
|1943||McCulloch & Pitts: Boolean circuit model of brain|
|1950||Alan Turing’s “Computing Machinery and Intelligence”|
|1951||Marvin Minsky develops a neural network machine|
|1950s||Early AI programs: e.g., Samuel’s checkers program,
Gelernter’s Geometry Engine,
Newell & Simon’s Logic Theorist and General Problem Solver
|1956||Dartmouth meeting: “Artificial Intelligence” adopted|
|1965||Robinson’s complete algorithm for logical reasoning|
|1966||Joseph Weizenbaum creates Eliza|
|1969||Minsky & Papert show limitations of the perceptron
Neural network research almost disappears
|1971||Terry Winograd’s Shrdlu dialogue system|
|1972||Alain Colmerauer invents Prolog programming language|
|1976||MYCIN, an expert system for disease diagnosis|
|1980s||Era of expert systems|
|1990s||Neural networks, probability theory, AI agents|
|1993||RoboCup initiative to build soccer-playing robots|
|1997||IBM Deep Blue beats the World Chess Champion|
|2003||Very large datasets: genomic sequences|
|2007||Very large datasets: WAC (web as corpus)|
|2011||IBM Watson wins Jeopardy|
|2012||US state of Nevada permits driverless cars|
|2014||“Deep learning”: recommendation systems, image tagging,
board games, speech translation, pattern recognition
|2016||Google AlphaGo beats the world’s 2nd best Go player, Lee Se-dol|
|Teachers||Peter Ljunglöf, John J. Camilleri, Jonatan Kilhamn, Inari Listenmaa, Claes Strannegård|
|Student representatives||Caterina Curta (N2COS), Claudia Castillo (MPALG), Ibrahim Fayaz (MPALG), Johan Ek (MPCAS), Tarun Nandakumar (MPCAS), Yan Wang (MPALG) (updated 22nd March)|
|Course book||Russell & Norvig (2002/10/14)
Read it online at Chalmers library: http://goo.gl/6EMRZr
Note for GU students:
Don’t forget to register, today!
This is what you (hopefully) will learn during this course:
Introduction to AI history, philosophy and ethics.
Basic algorithms for searching and solving AI problems:
This course is an introduction to AI, giving a broad overview
of the area and some basic algorithms.
We do not have the time to dig into the most recent algorithms
and techniques that are so hyped in current media.
Therefore, you will not learn how these things work:
Group work: Form a group
Group work: Write an essay
Group work: Shrdlite programming project
Written and oral examination
Obligatory group supervision
The exam is 2nd May (in the middle of the course)
The exam is only pass/fail
The exam is peer-corrected
Your project group will write a 6-page essay about the historical,
ethical and/or philosophical aspects of an AI topic.
After submitting your essay, you will get two other essays to read and review.
You will also get reviews on your essay, which you update and submit
a final version.
Claes Strannegård is responsible for the essay. He will organise
supervision sessions for all of you, regarding the essay.
Your group will implement a dialogue system for controlling a robot that lives
in a virtual block world and whose purpose in life is to move around objects
of different forms, colors and sizes.
Every group will get a personal supervisor, which you meet once every week.
There are two intermediate labs, which you submit by showing them to
To design a rational agent,
we must specify the task environment,
which consists of the following four things:
The task environment for an autonomous car:
|Observable?||full vs. partial|
|Deterministic?||deterministic vs. stochastic|
|Episodic?||episodic vs. sequential|
|Static?||static vs. dynamic (semidynamic)|
|Discrete?||discrete vs. continuous|
|Number of agents||single vs. multiple (competetive/cooperative)|
The environment type largely determines the agent design
|Chess (w. clock)||
|N:o agents||multiple (compet.)||multiple (compet.)||multiple (cooper.)||single|
The real world is (of course):
partially observable, stochastic, sequential, dynamic, continuous, multi-agent
Given an informal description of a problem, what is a solution?
Typically, much is left unspecified, but the unspecified parts
cannot be filled in arbitrarily.
Much work in AI is motivated by common-sense reasoning.
The computer needs to make common-sense conclusions
about the unstated assumptions.
Does it matter if the answer is wrong or answers are missing?
Classes of solutions:
An optimal solution is a best solution according to some
measure of solution quality.
A satisficing solution is one that is good enough, according
to some description of which solutions are adequate.
An approximately optimal solution is one whose measure
of quality is close to the best theoretically possible.
A probable solution is one that is likely to be a solution.
|Simple reflex agent||selects actions based on current percept
— ignores history
|Model-based reflex agent||maintains an internal state that depends
on the percept history
|Goal-based agent||has a goal that describes situations
that are desirable
|Utility-based agent||has a utility function that measures
|Learning agent||any of the above agents can be a learning agent
— learning can be online or offline
There are different opinions…
It’s all in the definitions:
The most important paper in AI, of all times:
(and I’m not the only one who thinks that…)
“Computing Machinery and Intelligence” (Turing, 1950)
introduced the “imitation game” (Turing test)
discussed objections against intelligent machines, including
almost every objection that has been raised since then
it’s also easy to read… so you really have to read it!
(1) The Theological Objection
(2) The “Heads in the Sand” Objection
(3) The Mathematical Objection
(4) The Argument from Consciousness
(5) Arguments from Various Disabilities
“you can make machines do all the things you have mentioned
but you will never be able to make one to do X.”
where X can… “be kind, resourceful, beautiful, friendly, […],
have a sense of humour, tell right from wrong, make mistakes,
fall in love, enjoy strawberries and cream, […], use words properly,
be the subject of its own thought, […], do something really new.”
(6) Lady Lovelace’s Objection
(7) Argument from Continuity in the Nervous System
(8) The Argument from Informality of Behaviour
(9) The Argument from Extrasensory Perception
this was the strongest argument according to Turing…
“the statistical evidence […] is overwhelming”
“Let us play the imitation game, using as witnesses a man who is good as a telepathic receiver, and a digital computer. The interrogator can ask such questions as ‘What suit does the card in my right hand belong to?’ The man by telepathy or clairvoyance gives the right answer 130 times out of 400 cards. The machine can only guess at random, and perhaps gets 104 right, so the interrogator makes the right identification.”
The brain replacement experiment
by Searle (1980) and Moravec (1988)
suppose we gradually replace each neuron in your head with
an electronic copy…
…what will happen to your mind, your consciousness?
Searle argues that you will gradually feel dislocated from your body
Moravec argues you won’t notice anything
The Chinese room experiment (Searle, 1980)
an English-speaking person takes input and generates answers in Chinese
he/she has a rule book, and stacks of paper
the person gets input, follows the rules and produces output
i.e., the person is the CPU, the rule book is the program and
the papers is the storage device
Does the system understand Chinese?
Will AI lead to superintelligence?
“…ever accelerating progress of technology and changes in the mode of human life, which gives the appearance of approaching some essential singularity in the history of the race beyond which human affairs, as we know them, could not continue” (von Neumann, mid-1950s)
“We will successfully reverse-engineer the human brain by the mid-2020s. By the end of that decade, computers will be capable of human-level intelligence.” (Kurzweil, 2011)
“There is not the slightest reason to believe in a coming singularity.” (Pinker, 2008)
What are the possible risks of using AI technology?
AI might be used towards undesirable ends
AI might result in a loss of accountability
what’s the legal status of a self-driving car?
or a medical expert system?
AI might mean the end of the human race