Week 2 – Things that think Flashcards
this is a rule that has been executed or used
describe a
fired rule
name the subsytems required for a
think-sense-act model robot
a robot of this model requires:
- perception subsytem
- cognition subsystem
- actuation subsystem
give two observations of
Valentino Braitenberg light seeking machine
two observations for this machine are:
- there is no programming such as if and else it s simply hard-wired and appears to show a specific behaviour
- The behaviour is spawned from how the sensors are connected to the actuators
who is
John Searle
an american philosipher who created concepts such as
- the chinese room
- strong A.I vs Weak A.I
the structure of this is:
- Condition/antecedent - a fact that may or may not be true
- Consequent - a fact that is true if the antecedent was true
example
Condition/antecedent - if my dog is eating her dinner is true
Consequent - then my dog is wagging her tail is true
describe the structure of a
rule
describe the
sense-act model
this model describes how an action is carried out only through sensory information
- robots - will act depending on what its sensors are detecting
- humans - will act/react according to what their senses tell them
important note
no processing or thinking is required it is an action carried out only through what the senses are reading
strategies to overcome this are:
- A rule can be given a higher priority
- The order in which instructions are written
- Which rues have not recently fired
name 3 methods that a robot can take to
resolve conflict in rules
describe
Learnt reactive behaviour
this is behaviour that has been learnt and is now an automatic reflex
it can be said that this fits the sense-act model
example
proper use of brakes and what action to take when a dangerous situation arises whilst driving
describe the
input layer of an artificial neural network
this layer of an artificial neural network receives data, such as readings from a sensor
the first robots programmed by Grey walter between 1948 and 1949 that were capable of perception and actuation and so fit the sense-act model
who are
elmer and elsie
this problem solving method can be described as:
using a fact solely because it remains working for the specific problem
in the context of problem solving desribe
conforming to habit
these include:
- Mechanical assembly - this includes everything that holds the robot together
- Power subsystem - this includes anything that is the source of the robots power
- Sensory subsystem - this includes any sensors that supply the robot with input
- Mobility/actuation subsystem - includes anything that allows the robot to move or cary out any orther useful tasks
- Control subsystem - anything that allows the robot to think receiving input from the sensory subsystem and providing input to the mobility/actuation subsystem
name the 5 major
parts that make up a robots architecture
this type of A.I would have human like intelligence and would have emotions and conciousness.
This is on the assumption that there is nothing mystical about human intelligence and that we really are simply a biological computer
describe
strong A.I
describe the structure of a
rule
the structure of this is:
- Condition/antecedent - a fact that may or may not be true
- Consequent - a fact that is true if the antecedent was true
example
Condition/antecedent - if my dog is eating her dinner is true
Consequent - then my dog is wagging her tail is true
this layer of an artificial neural network receives data from the input layer or other hidden layers.
Its job is to compute the data received
the more layers and artificial neurons in these can increase the neural networks performance
describe the
hidden layer of an artificial neural network
the layers for this include:
- input layer
- hidden layer
- output layer
what are the 3
layers of an artificial neural network
in the context of problem solving desribe
conforming to habit
this problem solving method can be described as:
using a fact solely because it remains working for the specific problem
this can be described as:
a machine that can learn by using previous data / data structures
example
learning what music someone likes by recognising what other people like and applying that knowledge to the individual
describe
machine learning
the cognition subsystem is often referred to as this
this is a problem solver that will take into account any information it has before deciding what action should be taken and subsequently which actuators should be used.
Information will be held in memory and could be new from sensors or existing from past data
describe the
Information processing system of a robot
these robots worked so that the light sesor and front wheel would turn and upon the light sensor detcting light it would turn its wheels towards that light source.
However if the light became to bright it would become dazzled and move away from the light
for the touch sensor if a touch was detected then it would move away from the touch
how did
grey walters robots elmer and elsie actually work
this problem solving method can be described as:
where there is information available that will help in achieving the goal then the logical decision is to take/use this fact
in the context of problem solving desribe
logical deduction
in the context of neural nets describe
generalisation
this is a term to describes how neural nets (once trained) can give a correct output for data they have not seen before.
This is in contrast to a rule based system that would struggle with this situation
read the relationship between
alan turings turning test / imitation game
and
john searles chinese room
this would appear to have human intelligence however would only ever be an illusion and would be acting from instructions instead of authentic understanding
describe
weak A.I
describe the
Information processing system of a robot
the cognition subsystem is often referred to as this
this is a problem solver that will take into account any information it has before deciding what action should be taken and subsequently which actuators should be used.
Information will be held in memory and could be new from sensors or existing from past data
read the relationship between
alan turings turning test / imitation game
and
john searles chinese room
this american philosipher proposed:
- strong ai
- weak ai
what are the
two types of A.I that were suggested by john searle
describe
Valentino Braitenberg light avoider vehicle
this machine is a simple design that implements the sense-act model and conveys behaviour
It works by having two light sensors each connected to a wheel each. The stronger the light reading the faster its wheel will turn and so it will turn away from the light. Similarly switching the cables will create a light seeking robot
this is a rule that could have been executed/fired but did not because of a circumstance such as another rule taking priority
describe a
triggered rule
in the context of facts describe a
phrase
in the context of facts this can be described as a collection of words
in the context of problem solving desribe
logical deduction
this problem solving method can be described as:
where there is information available that will help in achieving the goal then the logical decision is to take/use this fact
these include:
- Natural selection (genetic algorithm) - where a successful behaviour is passed on to the next generation
- Artificial neural networks - a way of training robots and mimics the connectivity found in animal brains
- Reinforcement learning - teaching by giving a reward when correct this reinforces the correct behaviour
name 3
approaches to training a robot
describe the
hidden layer of an artificial neural network
this layer of an artificial neural network receives data from the input layer or other hidden layers.
Its job is to compute the data received
the more layers and artificial neurons in these can increase the neural networks performance
these robots were wired so that the touch and light sensors would output to intermediate circuitry before being in-putted to the actuators.
This approach showed that the behaviour of a machine can depend on how its sensors and actuators are connected
how were
grey walters robots elmer and elsie wired
how did
grey walters robots elmer and elsie actually work
these robots worked so that the light sesor and front wheel would turn and upon the light sensor detcting light it would turn its wheels towards that light source.
However if the light became to bright it would become dazzled and move away from the light
for the touch sensor if a touch was detected then it would move away from the touch
in the context of facts describe a
fact
in the context of facts this can be described as a phrase that can be recognised and can have a truth value given to it
what are the
two types of A.I that were suggested by john searle
this american philosipher proposed:
- strong ai
- weak ai
describe
deep learning
this is a subset of machine learning but instead of using previous data structures it makes use of neural networks in its learning
example
finding what music someone likes instead of using past data we can let the user select like / hate this in turn allows the A.I to learn what the user likes
who is
Valentino Braitenberg
an Italian researcher who in the 1980s was unaware of the sense-act model work that Grey Walter had produced 30 years prior but did investigate the same model.
he never built any of his designs but they do demonstrate behaviour using the sense-act model
name the 5 major
parts that make up a robots architecture
these include:
- Mechanical assembly - this includes everything that holds the robot together
- Power subsystem - this includes anything that is the source of the robots power
- Sensory subsystem - this includes any sensors that supply the robot with input
- Mobility/actuation subsystem - includes anything that allows the robot to move or cary out any orther useful tasks
- Control subsystem - anything that allows the robot to think receiving input from the sensory subsystem and providing input to the mobility/actuation subsystem
in the context of problem solving desribe
associations
this problem solving method can be described as:
using past knowledge perhaps it has learnt new facts and so will have had associated a problem with a specific result either good or bad
the parts of this include:
- phrase
- truth value
example
The Queen of England is a man is false
Where:
Phrase = The Queen of England is a man
Truth value = is false
what are the 2
parts of a fact
an american philosipher who created concepts such as
- the chinese room
- strong A.I vs Weak A.I
who is
John Searle
how are
neural networks trained
- these will first take a set of known inputs and outputs this is referred to as training data
- Initially this will be incorrect but by adjusting its weights when it is incorrect it can eventually be trained to give the correct output for the input
note
Training these can take a long time and can require high amounts of energy and data
describe the term
reasoning
this is using cognition (gained knowledge and understanding) in order to reflect on facts and come to new facts
describe the
output layer of an artificial neural network
this layer of an artificial neural network is the last layer and will produce some kind of output such as moving a motor depending on the input it has received
describe
weak A.I
this would appear to have human intelligence however would only ever be an illusion and would be acting from instructions instead of authentic understanding
this model allows for more complex behaviour in which input can be assesed and various outputs performed from the ‘thinking’ processing
describe the
sense-think-act model
this man was a neuroscientist who was the first to build robots that fit the sense-act model
who was
grey walter
an Italian researcher who in the 1980s was unaware of the sense-act model work that Grey Walter had produced 30 years prior but did investigate the same model.
he never built any of his designs but they do demonstrate behaviour using the sense-act model
who is
Valentino Braitenberg
a robot of this model requires:
- perception subsytem
- cognition subsystem
- actuation subsystem
name the subsytems required for a
think-sense-act model robot
what are the 3
layers of an artificial neural network
the layers for this include:
- input layer
- hidden layer
- output layer
who are
elmer and elsie
the first robots programmed by Grey walter between 1948 and 1949 that were capable of perception and actuation and so fit the sense-act model
in the context of facts this can be described as a phrase that can be recognised and can have a truth value given to it
in the context of facts describe a
fact
describe a
fired rule
this is a rule that has been executed or used
this is the ability to acquire knowledge and understanding
describe the term
cognition
describe the
cognition subsystem of a robot
this is a subsystem of a robot that will implement cognition and is coined as the brains of the robot
however this is a misleading term (covered in later topic)
describe
strong A.I
this type of A.I would have human like intelligence and would have emotions and conciousness.
This is on the assumption that there is nothing mystical about human intelligence and that we really are simply a biological computer
this machine is a simple design that implements the sense-act model and conveys behaviour
It works by having two light sensors each connected to a wheel each. The stronger the light reading the faster its wheel will turn and so it will turn away from the light. Similarly switching the cables will create a light seeking robot
describe
Valentino Braitenberg light avoider vehicle
this model describes how an action is carried out only through sensory information
- robots - will act depending on what its sensors are detecting
- humans - will act/react according to what their senses tell them
important note
no processing or thinking is required it is an action carried out only through what the senses are reading
describe the
sense-act model
these include:
- logical deduction
- associations
- conforming to habit
name 3
problem solving methods
- these will first take a set of known inputs and outputs this is referred to as training data
- Initially this will be incorrect but by adjusting its weights when it is incorrect it can eventually be trained to give the correct output for the input
note
Training these can take a long time and can require high amounts of energy and data
how are
neural networks trained
this is behaviour that has been learnt and is now an automatic reflex
it can be said that this fits the sense-act model
example
proper use of brakes and what action to take when a dangerous situation arises whilst driving
describe
Learnt reactive behaviour
who was
grey walter
this man was a neuroscientist who was the first to build robots that fit the sense-act model
this problem solving method can be described as:
using past knowledge perhaps it has learnt new facts and so will have had associated a problem with a specific result either good or bad
in the context of problem solving desribe
associations
in the context of facts this can be described as a collection of words
in the context of facts describe a
phrase
this is a subset of machine learning but instead of using previous data structures it makes use of neural networks in its learning
example
finding what music someone likes instead of using past data we can let the user select like / hate this in turn allows the A.I to learn what the user likes
describe
deep learning
what are the 2
parts of a fact
the parts of this include:
- phrase
- truth value
example
The Queen of England is a man is false
Where:
Phrase = The Queen of England is a man
Truth value = is false
this is a term to describes how neural nets (once trained) can give a correct output for data they have not seen before.
This is in contrast to a rule based system that would struggle with this situation
in the context of neural nets describe
generalisation
name 3 methods that a robot can take to
resolve conflict in rules
strategies to overcome this are:
- A rule can be given a higher priority
- The order in which instructions are written
- Which rues have not recently fired
this is using cognition (gained knowledge and understanding) in order to reflect on facts and come to new facts
describe the term
reasoning
two observations for this machine are:
- there is no programming such as if and else it s simply hard-wired and appears to show a specific behaviour
- The behaviour is spawned from how the sensors are connected to the actuators
give two observations of
Valentino Braitenberg light seeking machine
name 3
approaches to training a robot
these include:
- Natural selection (genetic algorithm) - where a successful behaviour is passed on to the next generation
- Artificial neural networks - a way of training robots and mimics the connectivity found in animal brains
- Reinforcement learning - teaching by giving a reward when correct this reinforces the correct behaviour
name 3
problem solving methods
these include:
- logical deduction
- associations
- conforming to habit
describe the term
cognition
this is the ability to acquire knowledge and understanding
describe the
sense-think-act model
this model allows for more complex behaviour in which input can be assesed and various outputs performed from the ‘thinking’ processing
this layer of an artificial neural network is the last layer and will produce some kind of output such as moving a motor depending on the input it has received
describe the
output layer of an artificial neural network
how were
grey walters robots elmer and elsie wired
these robots were wired so that the touch and light sensors would output to intermediate circuitry before being in-putted to the actuators.
This approach showed that the behaviour of a machine can depend on how its sensors and actuators are connected
describe a
triggered rule
this is a rule that could have been executed/fired but did not because of a circumstance such as another rule taking priority
this will receive data such as numbers
from this it will determine whether those numbers exceed a threshold when added. if so then the artificial neuron is fired and produces an output
Each output link of the artificial neuron may also have a weight applied so each link will have different data depending on its weight.
The weight can help determine the importance of that particular link and thus determine the input value for the next neuron
describe a
single artificial neuron
describe a
single artificial neuron
this will receive data such as numbers
from this it will determine whether those numbers exceed a threshold when added. if so then the artificial neuron is fired and produces an output
Each output link of the artificial neuron may also have a weight applied so each link will have different data depending on its weight.
The weight can help determine the importance of that particular link and thus determine the input value for the next neuron
describe
machine learning
this can be described as:
a machine that can learn by using previous data / data structures
example
learning what music someone likes by recognising what other people like and applying that knowledge to the individual
this layer of an artificial neural network receives data, such as readings from a sensor
describe the
input layer of an artificial neural network
this is a subsystem of a robot that will implement cognition and is coined as the brains of the robot
however this is a misleading term (covered in later topic)
describe the
cognition subsystem of a robot
