Study questions Artificial Intelligence part
Lecture 1
• Explain the difference between weak AI and strong AI. What are their assumptions
and what does that mean for AI?
Weak AI: No matter how intelligent machine behavior may seem, it does not reflect
true intelligence or sentience. Machines can simulate human intelligence using
clever tricks. Thus, we can recreate human brain but it doesn’t mean that it’s
conscious or intelligent. It’s just clever tricks, not true intelligence ( Turing Test).
Strong AI: intelligent systems can actually think. The human mind is an information
processing system, and that thinking is a form of computing. One day a computer
can do anything a human can do. Are we asking the right questions in order to check
whether a machine is intelligent or not? A well-programmed mchine that exactly
emulates the human brain is a mind, and thereby intelligent. The substrate doesn’t
matter, whether it’s biological brain or computer chip. computationalist approach
of cognition.
• Explain the difference between symbolic and connectionist AI.
Symbolic AI (GOFAI): does not concern itself with neurophysiology. Has to do with
the manipulation of representations in the brain (e.g. a bird has a symbolic
representation). Human thinking is a kind of symbol manipulation. Can use if-then
rules. You can solve chess and checkers with this easily.
But then, language, translation, navigation symbolic AI sucks at these tasks where
humans are really good at neural networks.
Connectionist (PDP) AI: Maybe we can solve AI problems with neural networks.
Based on structure in human brain. Computation is massively parallel. Mental states
are represented as N-dimensional vectors of numeric activation values over neural
network units. Memory is created by modifying the connection strength (weight)
between units. No a priori assumption about problem space or statistical distribution.
By manipulating the connections we can create intelligence.
,• What was Turing’s imitation game about and how does it relate to intelligent
systems?
Turing thought it would be too hard to determine if computers can think. The question
is if it is possible to design a computer as described in the experiment and what the
biggest problems would be. Imitation game: it’s played with 3 people, a man (A), a
woman (B) and an interrogator (C). C must get to know which one is the man and
which one is the woman. It’s A’s object in the game to try and cause C to make the
wrong decision. B must help te interrogator. What will happen if a machine takes the
part of A? Will the interrogator decide wrongly as often when the game is played like
this as he does when the game is played between a man and woman? Turing: a
machine is intelligent if we cannot distinguish it from a human in conversation. (No
claims about underlying mechanisms).
, • How does Searle’s Chinese room argument relate to Turing’s imitation game? What
conclusions would Searle draw out of a computer passing the imitation game?
Chinese room argument: gedachte-experiment. Een computer kan zich gedragen als
een mens, maar dit hoef niet te betekenen dat hij ook kan denken als een mens. Dat
iemand de juiste antwoorden geeft, betekent dat dat we met een intelligent iets te
maken hebben? Volgens Searle kan een machine heel intelligent ogen, maar geeft
het geen echte intelligentie of begrip weer. Het zijn slimme trucjes ontworpen door
mensen. Het experiment: als je iemand in een kamer zet met een chinees
woordenboek en chinese mensen gaan vragen stellen, dan kun je antwoorden door
het chinese boek met regels maar snap je dan ook chinees? Volgens Searle heb je
dan geen echte understanding van Chinees en zo werkt dat ook voor computers?
Rule-based manipulation of symbols does not constitue intelligence. No matter how
intelligent machine behavior may seem, it does not reflect true intelligence or
sentience.
• Explain the criticism on symbolic AI. Give arguments against and for.
Knowledge-based or expert systems were hugely successful.
Against: why do we need symbolic representation? It doesn’t seem necessary for
many behaviors. It is unclear how processes like pattern recognition would work in a
purely symbolic way. It can’t handle noise very well. It doesn’t work for language,
translation, navigation etc. Also, it isn’t related to neurophysiology of the brain.
For: Expert systems can make human-like decisions, like MYCIN (diagnosis of
treatments). The system is more accurate than doctors.
• What are the similarities between connectionist AI architectures and the human
brain?
Connectionism:
- Is based on the structure of the human brain.
- Lesioned or damaged networks still can process information.
- ANNs are capable of learning, and are able to generalize rules to novel input.
- Parallel processing way faster than serial processing.
Lecture 1
• Explain the difference between weak AI and strong AI. What are their assumptions
and what does that mean for AI?
Weak AI: No matter how intelligent machine behavior may seem, it does not reflect
true intelligence or sentience. Machines can simulate human intelligence using
clever tricks. Thus, we can recreate human brain but it doesn’t mean that it’s
conscious or intelligent. It’s just clever tricks, not true intelligence ( Turing Test).
Strong AI: intelligent systems can actually think. The human mind is an information
processing system, and that thinking is a form of computing. One day a computer
can do anything a human can do. Are we asking the right questions in order to check
whether a machine is intelligent or not? A well-programmed mchine that exactly
emulates the human brain is a mind, and thereby intelligent. The substrate doesn’t
matter, whether it’s biological brain or computer chip. computationalist approach
of cognition.
• Explain the difference between symbolic and connectionist AI.
Symbolic AI (GOFAI): does not concern itself with neurophysiology. Has to do with
the manipulation of representations in the brain (e.g. a bird has a symbolic
representation). Human thinking is a kind of symbol manipulation. Can use if-then
rules. You can solve chess and checkers with this easily.
But then, language, translation, navigation symbolic AI sucks at these tasks where
humans are really good at neural networks.
Connectionist (PDP) AI: Maybe we can solve AI problems with neural networks.
Based on structure in human brain. Computation is massively parallel. Mental states
are represented as N-dimensional vectors of numeric activation values over neural
network units. Memory is created by modifying the connection strength (weight)
between units. No a priori assumption about problem space or statistical distribution.
By manipulating the connections we can create intelligence.
,• What was Turing’s imitation game about and how does it relate to intelligent
systems?
Turing thought it would be too hard to determine if computers can think. The question
is if it is possible to design a computer as described in the experiment and what the
biggest problems would be. Imitation game: it’s played with 3 people, a man (A), a
woman (B) and an interrogator (C). C must get to know which one is the man and
which one is the woman. It’s A’s object in the game to try and cause C to make the
wrong decision. B must help te interrogator. What will happen if a machine takes the
part of A? Will the interrogator decide wrongly as often when the game is played like
this as he does when the game is played between a man and woman? Turing: a
machine is intelligent if we cannot distinguish it from a human in conversation. (No
claims about underlying mechanisms).
, • How does Searle’s Chinese room argument relate to Turing’s imitation game? What
conclusions would Searle draw out of a computer passing the imitation game?
Chinese room argument: gedachte-experiment. Een computer kan zich gedragen als
een mens, maar dit hoef niet te betekenen dat hij ook kan denken als een mens. Dat
iemand de juiste antwoorden geeft, betekent dat dat we met een intelligent iets te
maken hebben? Volgens Searle kan een machine heel intelligent ogen, maar geeft
het geen echte intelligentie of begrip weer. Het zijn slimme trucjes ontworpen door
mensen. Het experiment: als je iemand in een kamer zet met een chinees
woordenboek en chinese mensen gaan vragen stellen, dan kun je antwoorden door
het chinese boek met regels maar snap je dan ook chinees? Volgens Searle heb je
dan geen echte understanding van Chinees en zo werkt dat ook voor computers?
Rule-based manipulation of symbols does not constitue intelligence. No matter how
intelligent machine behavior may seem, it does not reflect true intelligence or
sentience.
• Explain the criticism on symbolic AI. Give arguments against and for.
Knowledge-based or expert systems were hugely successful.
Against: why do we need symbolic representation? It doesn’t seem necessary for
many behaviors. It is unclear how processes like pattern recognition would work in a
purely symbolic way. It can’t handle noise very well. It doesn’t work for language,
translation, navigation etc. Also, it isn’t related to neurophysiology of the brain.
For: Expert systems can make human-like decisions, like MYCIN (diagnosis of
treatments). The system is more accurate than doctors.
• What are the similarities between connectionist AI architectures and the human
brain?
Connectionism:
- Is based on the structure of the human brain.
- Lesioned or damaged networks still can process information.
- ANNs are capable of learning, and are able to generalize rules to novel input.
- Parallel processing way faster than serial processing.
