U1T3.1 - Applications of DT

Question 1

What is meant by artificial intelligence?

Answer 1

Study of machines to model types of intelligence/creativity of people. Makes machines more capable + helps improve human intelligence + cognitive behaviour.

Question 2

What is cognitive science?

Answer 2

Understanding human mind + its thought processes. Can use computer models of info processing to explain how mind functions.

Question 3

What are some of the applications of AI?

Answer 3

Expert systems, image processing + visions, speech recognition + natural language processing + machine learning.

Question 4

What is the Turing Test?

Answer 4

Used to assess ability of machine to exhibit intelligent behaviour which can’t be distinguished from that of a human. Human simultaneously asks questions of computer + human. If person doing test can’t tell which is which, the machine has passed and can be classified as intelligent. Convo is through interface, text only. Test doesn’t assess correctness, just similarity to human response so criticised.

Question 5

What was the controversy in 2014 about the Turing Test?

Answer 5

Reliability questioned as chatbot called Eugene Goostman passed, designed to imitate 13 year old boy from Ukraine. Some experts discounted the results due to limited vocab of 2nd language young boy.

Question 6

What is a neural network model?

Answer 6

Computational model (software simulation) used in comp science to model primitive brain. Simulates densely interconnected brain cells to support learning, pattern recognition + decision making process as human would.

Question 7

What are the main features of a neural network?

Answer 7

Consists of artificial neurons (units) (50 - millions) depending on complexity of network. Units arranged in layers. Input units receive info from outside world. Series of hidden units (brain) process input from input units. Output units provide response from network, repping info it has learned after input processing. Usually fully connected meaning each unit connected to every other unit in the layers either side. Each connection weighted + algorithms calc weighted sum of any inputs into node to gen output value. Learns by itself.

Question 8

How does a neural network learn?

Answer 8

Patterns of info enter via input units, triggering layers of hidden units until arrive at output units. Not all units fire all the time. Inputs into unit multiplied by weightings of connections they travel along. Weighted outputs added together + if sum above threshold value, unit fires units connected to it. Known as feedforward network.

Question 9

What is backpropogation?

Answer 9

Feedback process used to support learning in neural network. Output produced compared to theoretical output. Diff used to modify connections between various units (go backward from output units, through hidden layers to input units). Over time, this reduces diff between actual + intended output, causing learning to occur.

Question 10

What are expert systems?

Answer 10

Knowledge-based systems. Application of AI. Knowledge of human expert made available through comp package. Narrow knowledge field (knowledge domain of system) Knowledge of human expert programmed into expert system + represented using rules + facts. Creates knowledge base which holds knowledge about domain. (Can be stored as series of IF-THEN-ELSE rules.

Question 11

What is needed to create an expert system?

Answer 11

Subject expert + team of programmers + technical experts (knowledge engineers). Extra rules + facts can be added to the knowledge base as time passes + users provide feedback about answer quality.

Question 12

What is an expert?

Answer 12

Experienced practitioner in particular field.

Question 13

What are the 2 types of knowledge that the knowledge base stores?

Answer 13

Factual, Heuristic.

Question 14

What is factual knowledge?

Answer 14

Factual info acquired by knowledge engineers from human experts, relating to subject domain. Widely agreed type of info.

Question 15

What is heuristic knowledge?

Answer 15

Captures info about accurate judgement + ability to estimate + evaluate. Not just from logic but person’s experience. ‘Rules of thumb’. Designed to work with uncertainty + simulate producing decisions based on experience.

Question 16

Gives examples of facts in an expert system.

Answer 16

Water boils at 100c, any person over 17 can apply for a driver’s licence.

Question 17

Give examples of rules in an expert system.

Answer 17

If water reaches its boiling point, steam will be produced.

Insurance policies for young people tend to be more expensive.

Question 18

What is done once the knowledge base has been created?

Answer 18

A system for consulting it is put into place. (Inference Engine)

Question 19

What is an Inference Engine?

Answer 19

Software which interrogates knowledge base + draws inferences + conclusions based on rules stored about subject domain. Poses questions to user + uses answers provided by user to determine suitable response.

Question 20

Give an example of how an inference engine works?

Answer 20

Qs -> Do you have a cough? Do you have sore ears>

Rules -> IF patient has cough AND sore ears, THEN diagnosis is severe head cold.

Question 21

What does the user interface do in an expert system?

Answer 21

Allows user to communicate with system. Requests for info/advice passed from UI to IE. Processed by IE which applies rules to knowledge base + returns response to user.

Question 22

What is the shell in an expert system?

Answer 22

Piece of software containing structure for creating expert system. ‘Empty’ expert system w/out knowledge base. Creator can enter appropriate rules + facts to gen expert system.

Question 23

What is fuzzy logic?

Answer 23

Generalisation of standard logic. Used for reasoning about inherently vague concepts like ‘tallness, richness, famousness, darkness’ Allows computers to manipulate this info.
e.g. X is tall, with a degree of truth of 0.9

Question 24

What is propositional (binary) logic?

Answer 24

Applies only to concepts which are true or false (1 or 0)

e.g. If it’s raining, I’ll take my umbrella.

Question 25

What is standard logic?

Answer 25

Applies to concepts which can possess a degree of truth between 0.0 and 1.0 e.g. Chance of getting an odd number when I throw a dice is 0.5

Question 26

What are some of the applications of fuzzy logic?

Answer 26

Modeling, evaluation, optimisation, decision making, control, diagnosis + info. Best suited for control systems + applied in areas like breakdown prediction of nuclear reactors, earthquake forecasting + subway control. Hoovers, microwaves + camcorders e.g. hoover applies more suction in dirty area.

Question 27

How is ABS controlled by fuzzy logic?

Answer 27

Consists of parameters like speed, brake pressure, brake temp, interval between applications of brakes + angle of car's lateral motion to forward motion. Range of values continuous, open to interpretation.

Question 28

What fields might expert systems be used in?

Answer 28

Medicine, car engine fault diagnosis + life insurance.

Question 29

How are expert systems used in medicine?

Answer 29

Who treated first, modern drug knowledge, rare condition diagnosis, specialists in different parts of world + checking diagnosis. Knowledge base = medical info, query = symptoms, advice = diagnosis.

Question 30

How are expert systems used in car engine fault diagnosis?

Answer 30

Car manufacturers create own expert systems about own cars. Stored on comp which can be plugged into car to determine fault + system recommends how to fix problem.

Question 31

How are expert systems used in life insurance?

Answer 31

Involves insurance company taking risk to cover individual's life for set monthly fee (insurance premium) Consider age, gender (older men pay more), smoking status, health.

Question 32

What are the benefits of expert systems?

Answer 32

Expert advice always available, expert knowledge recorded + used before they move on, training aid, rational decisions with no emotional interference, not tired or overworked, faster than humans, quickly identify equipment faults, accurate, cheaper than v.professional human, recommendations consistent + impartial + up to date.

Question 33

What are some of the limitations of expert systems?

Answer 33

Can make mistakes + don't learn from then, risk of over reliance so experts become deskilled, less suitable to less predictable decisions where interpretation required, human advisor takes special circumstances into account, can give reasons for decision but can't be further questioned + not every scenario can be programmed so errors possible.

Question 34

What is NLP?

Answer 34

Natural Language Processing. Branch of AI. 'Ability of machine to analyse, understand and generate human speech'. Makes human-computer interaction as human like as possible. ASR part of it.

Question 35

What is ASR?

Answer 35

Automated Speech Recognition. Converts input sound to text before NLP.

Question 36

What are some examples of how natural language processing is used today?

Answer 36

Spam filters, answering questions, extracting info + summarising info.

Question 37

How is NLP used in spam filters?

Answer 37

Extract meaning from strings of text to help identify unwanted email + prevent from entering clients in box in email apps.

Question 38

How is NLP used in answering questions?

Answer 38

Search engines provide lots of info but rely on users being specific. Google focuses on NLP so meaning can be extracted + appropriate answers provided.

Question 39

How is NLP used in extracting info?

Answer 39

Financial organisations use algorithmic trading to manage investments. Controlled by technology which evaluate news articles + extracts relevant info to evaluate stock market patterns before determining whether client should buy, sell or hold onto stocks.

Question 40

How is NLP used in summarising info?

Answer 40

Info overload is an issue. Try to analyse users with social media data to determine which articles + ads should be higher in news feed. Some access user's microphone to collect data... questionable.

Question 41

How does NLP work?

Answer 41

Computers often try to understand meaning of individual words rather than sentences. Apps must know context.

Question 42

What is morphology?

Answer 42

How words are formed + relationship with other words. (How context changes meaning)

Question 43

What is syntax?

Answer 43

How words + sentences are put together.

Question 44

What is semantics?

Answer 44

Meaning of words + groups of words.

Question 45

What is pragmatics?

Answer 45

Context of spoken expressions.

Question 46

What is phonology?

Answer 46

Sound associated with spoken language. (How words + phrases sound when spoken)

Question 47

What is PoS?

Answer 47

Part-of-Speech-Tagging. First step of NLP is morphology. Use self-learning algorithm, tagging words with multiple meanings. Determines highest occurring meaning + uses to try to understand functions/meanings of words around it.

Question 48

What are parse trees/diagrams?

Answer 48

Using knowledge derived from syntax to understand sentence structure. Algorithm breaks sentence into noun + verb phrases to aid sentence understanding. Outcome is parse tree.

Question 49

What is the order of the steps of NLP?

Answer 49

PoS, parse trees, semantics.

Question 50

What is the third step of NLP?

Answer 50

Considers semantics of sentence. If word has same meaning, processes words that appear before and after tagged word to apply meaning to sentence. (Cut bread on board, member of board) Must learn difference. Preceded by bread, cutting board. Preceded by member, board of directors.

Question 51

Where are there areas for improvement in NLP?

Answer 51

Pragmatics. Most sentences conform to context that requires gen understanding of human world + emotions which is difficult to teach computer. e.g. sarcasm.

Question 52

How do voice recognition systems work?

Answer 52

Input + digitisation of spoken word into VR software using computer with sound card + mic/headset. Specialised software must be downloaded to support voice recognition.

Question 53

What is a VR system?

Answer 53

Voice Recognition system. Combo of hardware + software systems which can decode spoken command. Used to operate devices/execute commands without using keyboards/mice/tracker pads.

Question 54

What are the different ways computers can recognise speech? (4)

Answer 54

Pattern matching, pattern + feature analysis, statistical analysis or artificial neural networks.

Question 55

Describe pattern matching for VR.

Answer 55

Words spoken by user recognised in entirety. Used by businesses with automated switchboards. User presented with questions with limited response (yes/no). Computer analyses input + matches with list of potential sound patterns representing available answers.

Question 56

Describe pattern + feature analysis for VR.

Answer 56

Spoken input recorded by mic + digitised using ADC. Digital data analysed + compared to stored dictionary + words identified. More complex input analysed this way + less limited responses.

Question 57

What is an ADC?

Answer 57

Analogue to Digital Converter.

Question 58

Describe statistical analysis for VR.

Answer 58

More complex systems take more statistical approach to analysis of speech input. Apply grammar rules to predict words for speech recog, especially when spoken word not clear.

Question 59

Describe artificial neural networks for VR.

Answer 59

Still being explored. Looked at how train through examples to recognise speech input. Recent studies look at combo of ANN + stat analysis to improve accuracy of voice recognition applications.

Question 60

Give the advantages of natural language. (3)

Answer 60

No training required, increases accessibility of application + handsfree if combined with speech recognition.

Question 61

Give the advantages of voice recognition. (3)

Answer 61

No training required, accurate spelling for users with literacy issues + effective input method for those with limited mobility.

Question 62

Give the disadvantages of natural language. (2)

Answer 62

Not all commands recognised/identified correctly all the time + applications complex + expensive to create.

Question 63

Give the disadvantages of voice recognition. (3)

Answer 63

Prone to interference, must speak slower, louder + more clearly + may only recognise limited range of voices/trained to recognise specific user's voice.

Question 64

What is robotics?

Answer 64

Combo of range of disciplines, including comp sci, mechanical + electronic engineering to support design, production operation + application of robots. Incorporates use of comp control systems in info processing + feedback application from sensors used as input elements. Use of control systems + info technologies to reduce need for human work in goods production + services.

Question 65

What does robotics involve?

Answer 65

Design, construction, programming + testing of machines using combo oh physics, mechanical, electronic + structural engineering, maths + computing to produce robots which can solve issues. Bio, medicine + chem may also be involved.

Question 66

What is a robot?

Answer 66

Machine that imitates a human. Can't give robot enough common sense to reliably interact in dynamic world but humanoid robots can do dangerous, boring or nasty work. Found in auto, medical, manufacturing + space industries.

Question 67

What is an industrial robot?

Answer 67

Automatically controlled, reprogrammable, multipurpose manipulator programmable in 3 or more axes.

Question 68

What are some of the essential characteristics of robots?

Answer 68

Sensing, movement, energy + intelligence.

Question 69

What should robots be able to do in terms of sensing?

Answer 69

Sense surroundings like human. Sensors (light, touch + pressure, chemical (nose), hearing, sonar + taste)

Question 70

What should robots be able to do in terms of movement?

Answer 70

Should be able to move around environment. Wheels, legs, thrusters.

Question 71

What should robots be able to do in terms of energy?

Answer 71

Power itself (solar, electric, battery)

Question 72

What should robots be able to do in terms of intelligence?

Answer 72

Needs to be smart. Programming allows it to know what it's supposed to do.

Question 73

What are the component parts of an industrial robot?

Answer 73

Controller, arm, end effector, drive + sensors.

Question 74

Describe the controller in an industrial robot.

Answer 74

Reps brain of robot arm + co-ordinates operation of all other parts. Controlled using comp program, allows robot to connect to external systems. Program used input using teach pendant.

Question 75

What is a teach pendant?

Answer 75

Specially designed hand held device used to instruct robot by specifying movements it takes to complete task.

Question 76

Describe the arm in an industrial robot.

Answer 76

Vary in size + shape in accordance with their designated task. Positions end effector. Moves + rotates end effector into place using range of joints. Each joint adds 3 degrees freedom to range of motion. If 3 points, 3 degrees of motion so up + down, left + right, forward + backward. Most have 6 degrees motion.

Question 77

Describe the end effector in an industrial robot.

Answer 77

Item of hardware, connects to arm + performs specific function. Can be rippers, vacuum pumps, magnets, welding torches, paint sprays etc. Can be changed to support completion of diff tasks by arm.

Question 78

Describe the drive in an industrial robot.

Answer 78

Engine/motor used to move various robot parts into appropriate position when completing programmed task. May be hydraulic, electric or pneumatic.

Question 79

Describe the sensors in an industrial robot.

Answer 79

Provide feedback to controller about environment arm interacts with. Processed by controller to determine further action arm must do to complete task.

Question 80

What is online programming for robots?

Answer 80

For basic tasks like paint spraying, + pickup + place operations. Moves in unison with commands used by programmer. Incorporates record play-back where robot performs sequence of moves. 2 types.

Question 81

What are the 2 types of online programming?

Answer 81

Lead through + drive through.

Question 82

What is lead through online programming?

Answer 82

Teach robots to carry out tasks like paint spraying, applying adhesive to irregular service. Arms have wrist motion + 3 degrees freedom + revolute joints. Manually taken through operating cycle + movements of each axis automatically logged at frequent intervals. Joint position sensors provide position info as robot moved through cycle. Position info sampled periodically + stored in comp memory.

Question 83

What is drive through online programming?

Answer 83

Used for industrial tasks like spot welding + machine (un)loading. Movements controlled by inputs from keypad. Programmer specifies movement + speed of each limb. Robot cycle is sequence of movements which can be observed during programming. In play-back period, can modify sequence to get optimum cycle time + accuracy. Operator may need to be close to driven robot which creates safety issues.

Question 84

What is offline programming for robots?

Answer 84

Program uses high level language like VAL, gives decision making power. Needs lots of computing power + uses sensors to provide system status info. Input sensors e.g. position, vision, tactile. Sensor info allow robot to take alt action within task cycle. Robot can make decisions like counting num operations/perform one task until another ready to be started. Useful in production as can be reprogrammed with min interference to production process.

Question 85

What are some of the ways robotics is used in commercial situations?

Answer 85

Welding, spray painting, product assembly, product unpacking, aerospace robots + healthcare provision (surgery).

Question 86

What are the predictions in terms of robots vs humans in the future?

Answer 86

Cheaper, better robots may replace humans, reducing manufacturing costs. Investment may grow 10% a year. Can only operate in predictable situations.

Question 87

What are the advantages of industrial robots?

Answer 87

Increased efficiency, higher quality + consistency, reducing need for quality control, improved working environment (dangerous situations, never tired, no RSI), increased profitability (due to efficiency, quality, time + resources) + productivity (can work 24/7, never tired)

Question 88

What are the disadvantages of industrial robots?

Answer 88

Capital cost (v.expensive but positive ROI), expertise (require training + expertise to set up) + limitations (depend on how well surrounding systems integrated)

Question 89

What is ROI?

Answer 89

Return on Investment.

Question 90

How does a mobile phone work?

Answer 90

Uses low intensity microwave signals to transmit + receive voice + data info. Transmits radio signals to broadcast radio location so calls diverted to it. Use network of cells to transmit data. Listens for SID on control panel. If can't find control channels, out of range + displays 'no service' message. When receives SID, compares to programmed SID + if match, connected to home network. Sends registration request + MTSO registers phone + tracks location.

Question 91

What is an SID?

Answer 91

System Identification Code. Unique code assigned to each carrier/network provider.

Question 92

What is the MTSO?

Answer 92

Mobile Telephone Switching Office. Mobile equiv to PSTN. Routes calls between diff regions of network. Diff telecom companies + network providers interconnected to form single virtual network to span entire country.

Question 93

How do you make a phone call on a mobile phone?

Answer 93

Microphone in phone picks up voice signal + microchip converts into radio wave so transmitted over mobile network. Antennae broadcasts signal to nearest mobile phone mast that picks up strongest signal. Receives signal, transmits to base station. Calls from 1 mobile to other on same network routed through base stations until dest. phone. On diff network, longer route. Radio transmitter + receiver in mobile can't transmit over long distances so numbers sent to nearest base station + passed through neighbouring stations to dest. phone. Mobiles comm. with nearest mast + base, allowing calls to be routed from or to phone.

Question 94

What is the cell?

Answer 94

Area over which base station has control. Base station coordinates transmission of signal across cell. In hexagon shape, masts + base stations placed strategically for max cell coverage. As user moves, transmitted smoothly to diff cells. More cells, more calls at once.

Question 95

What are masts?

Answer 95

Large high powered antennae capable of sending radio signals over larger distances.

Question 96

What is handoff?

Answer 96

How network automatically switches coverage responsibility from one base station to another while user moves. Use of cells means system can handle multiple calls at once as each cell uses same frequency set as cells near it.

Question 97

What is an MSC?

Answer 97

Mobile Switching Centre. Carries out switching functions like call set-up, release + routing. Also routes SMS messages + conference calls + interfaces with PSTN. Allows base stations to connect to it while connects to PSTN. Enables all forms of comms, whether between 2 mobiles or mobile + landline.

Question 98

What is PSTN?

Answer 98

Public Switched Telephone Network.

Question 99

Give a basic rundown of the main features of a neural network?

Answer 99

Many nodes interconnected by 1 or 2 way connections. Some connected to outside world providing input + output. Applies weightings calculated from inputs to provide output. Made to function like human brain.

Question 100

What is a base station?

Answer 100

Controls num of phone masks + routes call to mobile switching centre.