I85PREP
⌘ K
EXIN BCS AI Essentials study materials
EXIN BCS AI Essentials study resources

EXIN BCS AI Essentials practice questions

One question at a time. Build your EXIN BCS AI Essentials understanding.

First 50 of 200 questions. Each one lists the choices, the correct answer, and an explanation. The checker at the top of this page saves progress in this browser when it is available.

Continue in the practice tool above to load more questions, 50 at a time.

Question 1Artificial and Human Intelligence: An Introduction and History

In the history of artificial intelligence, which researcher is most closely associated with proposing the Turing Test as a way to judge machine intelligence?

  1. John McCarthy
  2. Marvin Minsky
  3. Claude Shannon
  4. Alan Turing Correct answer.

Explanation

Alan Turing proposed the imitation game, later called the Turing Test, in his 1950 paper on computing machinery and intelligence. The test asks whether a machine's text responses can be distinguished from a human's, making it a behavioural criterion rather than a definition of consciousness.

Question 2Artificial and Human Intelligence: An Introduction and History

The term 'artificial intelligence' was coined in connection with which event?

  1. The 1969 first International Joint Conference on AI
  2. The 1980 launch of the first expert system shell
  3. The 1956 Dartmouth workshop Correct answer.
  4. The 1943 McCulloch-Pitts paper

Explanation

John McCarthy used the phrase artificial intelligence in the 1955 proposal for the 1956 Dartmouth summer research project, which is conventionally treated as the birth of the field. The workshop gathered researchers to explore how machines could simulate learning and intelligence.

Question 3Artificial and Human Intelligence: An Introduction and History

Which statement best describes the relationship between artificial intelligence and human intelligence as framed in this domain?

  1. AI systems reproduce some human cognitive functions by computational means rather than by biological processes Correct answer.
  2. AI is defined as any computer program that performs arithmetic faster than a person
  3. AI and human intelligence are identical because both process information
  4. AI systems must replicate the human brain's neural chemistry to be considered intelligent

Explanation

AI is generally characterised as engineering computational systems that perform tasks associated with human intelligence, such as reasoning or perception, without claiming to copy biological mechanisms. Human intelligence is embodied and biological, while AI is implemented in machines, so the two overlap functionally but are not the same thing.

Question 4Artificial and Human Intelligence: An Introduction and History

Which capability is a classic example of something humans do easily but that has proved hardest for AI systems?

  1. Multiplying two eight-digit numbers
  2. Recognising a familiar face in an unusual pose Correct answer.
  3. Searching a database for an exact match
  4. Sorting a list of a million records

Explanation

Moravec's paradox notes that perception and sensorimotor skills that feel effortless to humans are computationally very hard, while tasks humans find difficult, such as large arithmetic, are easy for machines. Face recognition under varied lighting, pose and expression illustrates this perceptual difficulty.

Question 5Artificial and Human Intelligence: An Introduction and History

The 1943 model by McCulloch and Pitts contributed to AI history primarily by showing what?

  1. That networks of simple threshold units can compute logical functions Correct answer.
  2. That natural language can be parsed with context-free grammars
  3. That reinforcement learning can beat human chess players
  4. That machine translation between languages is feasible

Explanation

McCulloch and Pitts described an idealised neuron as a threshold logic unit and demonstrated that connected networks of such units can realise Boolean functions. This established the theoretical foundation later developed into artificial neural networks.

Question 6Artificial and Human Intelligence: An Introduction and History

Which pair correctly matches an AI subfield with the question it primarily addresses?

  1. Robotics with proving mathematical theorems
  2. Planning with translating spoken language
  3. Computer vision with interpreting images and video Correct answer.
  4. Knowledge representation with optimising graphics rendering

Explanation

Computer vision concerns extracting meaning from visual data such as images and video. Each other pairing mismatches a subfield with a task belonging to a different area, so only the first pairing is coherent.

Question 7Artificial and Human Intelligence: An Introduction and History

A team builds a system that recommends films by finding users with similar rating histories. Which classic AI approach does this best illustrate?

  1. Case-based reasoning over explicit rules
  2. Symbolic rule-based reasoning
  3. Statistical similarity and pattern matching Correct answer.
  4. Genetic algorithm optimisation

Explanation

Collaborative filtering is a data-driven statistical technique that measures similarity between users or items and predicts preferences from observed patterns. It does not rely on hand-written rules or on evolving candidate solutions as a genetic algorithm would.

Question 8Artificial and Human Intelligence: An Introduction and History

Which description best captures the 'strong AI' position as debated in the history of the field?

  1. A suitably programmed computer can have a mind in the full sense Correct answer.
  2. AI research should prioritise speed over understanding
  3. AI should only be applied to narrow commercial problems
  4. AI systems should mimic human biology exactly

Explanation

Strong AI holds that an appropriately programmed computer with the right inputs and outputs would literally possess a mind and understanding, not merely simulate it. Weak AI, by contrast, treats the computer as a useful model of mental processes without claiming genuine mentality.

Question 9Artificial and Human Intelligence: An Introduction and History

The philosopher John Searle introduced which thought experiment to argue that symbol manipulation alone does not produce understanding?

  1. The Chinese room Correct answer.
  2. The trolley problem
  3. The ship of Theseus
  4. The brain in a vat

Explanation

In the Chinese room, a person follows rules to manipulate Chinese symbols without understanding the language, which Searle uses to argue that syntax is not sufficient for semantics. The argument targets the claim that a running program by itself yields comprehension.

Question 10Artificial and Human Intelligence: An Introduction and History

Which historical period is known as the first 'AI winter'?

  1. The late 1990s, after the dot-com crash
  2. The 2010s, after deep learning became dominant
  3. The early 1970s, after funding cuts followed critical reports on machine translation and general problem solving Correct answer.
  4. The late 1950s, after the Dartmouth workshop

Explanation

The first AI winter refers to the funding collapse of the early to mid 1970s, driven by disappointed results, the ALPAC report on machine translation and the Lighthill report in the UK. Research continued but at much reduced scale until expert systems revived interest.

Question 11Artificial and Human Intelligence: An Introduction and History

An expert system used in medical diagnosis stores symptoms, diseases and rules linking them, and asks the user questions to reach a conclusion. Which components are essential to such a system?

  1. A knowledge base and an inference engine Correct answer.
  2. A convolutional neural network and a GPU cluster
  3. A reinforcement learning agent and a reward signal
  4. A genetic algorithm and a fitness function

Explanation

Classic expert systems separate domain knowledge, held in the knowledge base as facts and rules, from the inference engine that applies those rules to the case at hand. This separation is what distinguishes them from purely learned models.

Question 12Artificial and Human Intelligence: An Introduction and History

What was the central claim of the physical symbol system hypothesis advanced by Newell and Simon?

  1. Intelligence is equivalent to statistical pattern classification
  2. A physical symbol system has the necessary and sufficient means for general intelligent action Correct answer.
  3. Intelligence emerges only from embodied interaction with the world
  4. Intelligence requires massively parallel subsymbolic processing

Explanation

Newell and Simon argued that symbol manipulation of the kind performed by digital computers is both necessary and sufficient for general intelligence. This hypothesis underpinned the symbolic AI tradition and was later challenged by connectionist and embodied approaches.

Question 13Artificial and Human Intelligence: An Introduction and History

Which development most directly enabled the deep learning boom of the 2010s?

  1. A new formal proof that search algorithms are complete
  2. The standardisation of the COBOL programming language
  3. Large labelled datasets combined with GPU-accelerated neural network training Correct answer.
  4. The invention of the relational database

Explanation

Deep learning advanced rapidly when very large datasets, such as ImageNet, met affordable GPU computing, allowing deep networks to be trained effectively. Algorithmic refinements helped, but data and compute were the decisive enablers.

Question 14Artificial and Human Intelligence: An Introduction and History

Which statement correctly distinguishes machine learning from classical rule-based programming?

  1. Machine learning requires no training data because it reasons from first principles
  2. Machine learning always produces explanations that humans can read directly
  3. Machine learning is limited to problems with a single numerical input
  4. Machine learning derives its behaviour from data rather than from hand-coded decision rules Correct answer.

Explanation

In classical programming the developer writes the rules explicitly, whereas in machine learning the model infers a mapping from examples. This data-driven nature is what allows learning systems to handle problems where rules are hard to specify.

Question 15Artificial and Human Intelligence: An Introduction and History

A chatbot answers questions by matching the user's wording against stored templates and retrieving the closest phrase. Which limitation most clearly shows this is not genuine natural language understanding?

  1. It cannot store more than one hundred templates
  2. It fails when the same meaning is expressed with unfamiliar wording Correct answer.
  3. It cannot produce output in more than one language
  4. It requires a graphics processor to run

Explanation

Template matching depends on surface form rather than meaning, so paraphrases that share no keywords break the retrieval. Genuine understanding would generalise across different phrasings of the same intent.

Question 16Artificial and Human Intelligence: An Introduction and History

Which historical milestone is correctly attributed?

  1. The Turing Test was passed by a program in 1966
  2. The term machine learning was coined at the 1956 Dartmouth workshop
  3. AlphaGo defeated a world champion at the game of Go in 1985
  4. Deep Blue defeated the reigning world chess champion in 1997 Correct answer.

Explanation

IBM's Deep Blue beat Garry Kasparov in a six-game match in 1997, a landmark for search-based game playing. Go proved far harder and was only mastered by AlphaGo in 2016.

Question 17Artificial and Human Intelligence: An Introduction and History

The 'ELIZA effect' refers to which tendency observed in early human-computer interaction?

  1. Developers overfitting a model to its training set
  2. Users preferring graphical interfaces to command lines
  3. Users attributing understanding and empathy to a program that merely manipulates text patterns Correct answer.
  4. Programs becoming slower as their rule bases grow

Explanation

ELIZA, a 1960s pattern-matching chatbot, led some users to confide in it as if it understood them. The ELIZA effect names this human tendency to read intelligence and feeling into shallow conversational behaviour.

Question 18Artificial and Human Intelligence: An Introduction and History

Which task best illustrates the difference between narrow AI and general intelligence?

  1. A system that uses a larger screen for output
  2. A system that stores more data than its predecessor
  3. A system that plays Go at superhuman level but cannot answer simple questions about the game's rules in prose Correct answer.
  4. A system that runs faster on newer hardware

Explanation

Narrow AI excels within a defined task boundary, whereas general intelligence would transfer competence across domains. A Go engine that cannot discuss the game in language demonstrates high performance without broad, flexible understanding.

Question 19Artificial and Human Intelligence: An Introduction and History

Which philosophical question is raised most directly by the possibility of machines that appear to feel emotions?

  1. Whether machines can be manufactured at low cost
  2. Whether programming languages should be standardised
  3. Whether networks can transmit data quickly enough
  4. Whether functional behaviour is sufficient evidence for subjective experience Correct answer.

Explanation

The problem of other minds asks how we can know that any entity, machine or human, has subjective experience rather than merely behaving as if it does. Expressive machine behaviour sharpens this because the evidence available is purely functional.

Question 20Artificial and Human Intelligence: An Introduction and History

Which statement about the relationship between AI and cognitive science is most accurate?

  1. Cognitive science proves that all AI must be symbolic
  2. AI models can serve as testable hypotheses about how human cognition might work Correct answer.
  3. AI is only concerned with building faster hardware for psychologists
  4. Cognitive science has no influence on the design of AI systems

Explanation

Computational models in AI can be compared against human behaviour and brain data, making them candidate explanations of cognition. The exchange runs both ways, with psychological findings also shaping AI architectures such as neural networks.

Question 21Artificial and Human Intelligence: An Introduction and History

Which of these is the clearest example of a task requiring common-sense knowledge that early symbolic systems struggled with?

  1. Converting a temperature from Celsius to Fahrenheit
  2. Inferring that a wet umbrella implies it has been raining Correct answer.
  3. Sorting names into alphabetical order
  4. Computing the square root of a large number

Explanation

Common-sense reasoning requires vast background knowledge that is rarely stated explicitly, such as the link between wet umbrellas and rain. Symbolic systems with limited hand-coded facts could not capture this breadth, a difficulty known as the knowledge acquisition bottleneck.

Question 22Artificial and Human Intelligence: An Introduction and History

What does the term 'GOFAI' denote in the history of artificial intelligence?

  1. The use of graphics processing units for training
  2. The general-purpose optimisation of financial algorithms
  3. The generation of facial images by neural networks
  4. The classical symbolic approach using explicit representations and search Correct answer.

Explanation

GOFAI stands for good old-fashioned AI, the symbolic paradigm built on logic, explicit knowledge representation and heuristic search. It is contrasted with later statistical and connectionist methods that learn from data.

Question 23Artificial and Human Intelligence: An Introduction and History

A researcher claims a system is intelligent because it passes the Turing Test. Which criticism is most relevant?

  1. The test can only be applied to chess-playing programs
  2. The test measures conversational imitation rather than genuine understanding or broad competence Correct answer.
  3. The test demands that the machine be conscious by definition
  4. The test requires the machine to be physically embodied

Explanation

The Turing Test assesses whether a machine's responses are indistinguishable from a human's in text conversation, which can be achieved through imitation and does not establish understanding or general ability. Critics note that it is a behavioural and anthropocentric criterion.

Question 24Artificial and Human Intelligence: An Introduction and History

Which sequence correctly orders these milestones from earliest to latest?

  1. First AI winter, Deep Blue, McCulloch-Pitts neuron model, Dartmouth workshop
  2. Deep Blue, first AI winter, Dartmouth workshop, McCulloch-Pitts neuron model
  3. Dartmouth workshop, McCulloch-Pitts neuron model, Deep Blue, first AI winter
  4. McCulloch-Pitts neuron model, Dartmouth workshop, first AI winter, Deep Blue Correct answer.

Explanation

The McCulloch-Pitts model appeared in 1943, the Dartmouth workshop in 1956, the first AI winter in the early 1970s and Deep Blue's victory in 1997. Only the first option respects this chronological order.

Question 25Artificial and Human Intelligence: An Introduction and History

Which statement best explains why the connectionist approach differs from symbolic AI?

  1. It requires every decision to be traceable to a single explicit rule
  2. It forbids the use of numerical computation
  3. It can only be applied to problems in formal logic
  4. It distributes knowledge across many weighted connections rather than storing discrete rules Correct answer.

Explanation

Connectionist models represent knowledge as patterns of weights across many simple units, allowing gradual learning and generalisation. Symbolic AI instead manipulates explicit symbols and rules, which makes reasoning transparent but brittle.

Question 26Artificial and Human Intelligence: An Introduction and History

Which discipline contributed the neuron-inspired processing unit that later became the artificial neuron in connectionist AI?

  1. Civil engineering, through structural load analysis
  2. Accountancy, through double-entry bookkeeping
  3. Neurophysiology, through models of biological nerve cells Correct answer.
  4. Cartography, through map projection methods

Explanation

Early connectionism drew directly on neurophysiological research into how biological neurons receive and transmit signals. McCulloch and Pitts translated this into a simplified logical unit that could compute, forming the basis of artificial neural networks.

Question 27Artificial and Human Intelligence: An Introduction and History

In the history of AI, what did Alan Turing propose as a practical substitute for the question 'Can machines think?'

  1. A measurement of the machine's processing speed
  2. An operational test based on indistinguishability of responses in text conversation Correct answer.
  3. A formal proof that thinking requires consciousness
  4. A survey of public opinion about machine intelligence

Explanation

Turing replaced the vague philosophical question with an operational one: whether a machine's conversational responses could be distinguished from a human's. This reframing made machine intelligence testable in principle rather than dependent on defining thought.

Question 28Artificial and Human Intelligence: An Introduction and History

Which statement best describes the Dartmouth workshop of 1956 in the history of artificial intelligence?

  1. It produced the first commercially successful industrial robot
  2. It established the first international regulation of autonomous weapons
  3. It gathered researchers who coined the field's name and set an ambitious research agenda Correct answer.
  4. It introduced the backpropagation learning algorithm to the wider community

Explanation

The Dartmouth summer project is conventionally treated as the founding event of AI as a named discipline, bringing together researchers such as McCarthy, Minsky, Shannon and Rochester. Its proposal framed the goal of making machines simulate intelligence.

Question 29Artificial and Human Intelligence: An Introduction and History

Which capability is characteristic of a general-purpose intelligent agent but not of today's narrow AI systems?

  1. Executing a single well-defined classification task accurately
  2. Processing input faster than a human operator
  3. Operating continuously without human supervision
  4. Transferring competence to unfamiliar tasks without task-specific retraining Correct answer.

Explanation

Generality means an agent can apply its intelligence across many domains, including ones it was not specifically built for. Current narrow systems are typically specialised and require retraining or redesign for new tasks.

Question 30Artificial and Human Intelligence: An Introduction and History

Which pair correctly links an early AI programme with the problem domain it addressed?

  1. SHRDLU with diagnosing blood infections
  2. GPS with translating Russian text into English
  3. DENDRAL with inferring molecular structures from mass-spectrometry data Correct answer.
  4. MYCIN with playing tournament chess

Explanation

DENDRAL was an early expert system in chemistry, using spectral data and heuristic rules to propose molecular structures. Each of the other pairings misattributes a well-known system to a domain it did not serve.

Question 31Artificial and Human Intelligence: An Introduction and History

The Lighthill report of 1973 is remembered in AI history primarily because it did what?

  1. Proved that neural networks could not learn
  2. Announced the first successful general-purpose robot
  3. Founded the first academic department of artificial intelligence
  4. Criticised the field's failure to deliver on its promises and contributed to funding cuts in the UK Correct answer.

Explanation

Lighthill's assessment questioned the practical achievements of AI and highlighted the combinatorial explosion problem, which damaged the field's credibility with UK funders. It is commonly linked to the funding decline of the 1970s.

Question 32Artificial and Human Intelligence: An Introduction and History

Which statement best captures the difference between a knowledge base and an inference engine in a rule-based expert system?

  1. The knowledge base learns from data, while the inference engine labels training examples
  2. The knowledge base holds facts and rules, while the inference engine applies them to derive conclusions Correct answer.
  3. The knowledge base performs the reasoning, while the inference engine stores the rules
  4. The knowledge base is written in the user interface, while the inference engine is the database

Explanation

Classical expert systems separate declarative knowledge, stored as facts and if-then rules, from the reasoning procedure that chains those rules to reach a conclusion. This separation allows the knowledge to be edited without rewriting the reasoning code.

Question 33Artificial and Human Intelligence: An Introduction and History

What did the term 'cybernetics', as used by Norbert Wiener, primarily concern?

  1. Control and communication in animals and machines Correct answer.
  2. The simulation of language understanding by rule sets
  3. The design of very large integrated circuits
  4. The statistical analysis of large text corpora

Explanation

Wiener's cybernetics studied feedback, control and communication across biological and engineered systems. It overlapped with early AI but emphasised regulation and feedback rather than symbolic reasoning.

Question 34Artificial and Human Intelligence: An Introduction and History

A warehouse robot navigates by maintaining an internal map and planning routes through it. Which classical AI tradition does this most directly reflect?

  1. The statistical tradition of fitting parameters to labelled sensor data
  2. The behaviourist tradition of stimulus-response without representation
  3. The deliberative, symbolic tradition of representing a world model and reasoning over it Correct answer.
  4. The evolutionary tradition of breeding solutions through selection

Explanation

Maintaining an explicit map and planning over it exemplifies the deliberative approach, in which knowledge is represented symbolically and manipulated by search or planning algorithms. The robot's behaviour is derived from reasoning about the model rather than from reactive reflexes alone.

Question 35Artificial and Human Intelligence: An Introduction and History

Why did the combinatorial explosion of search spaces become a central problem for early general problem-solving programs?

  1. Human users refused to provide any problem descriptions
  2. The programming languages of the era lacked arithmetic operations
  3. The programs could not store any data in memory
  4. The number of states to examine grew so fast that exhaustive search became computationally infeasible Correct answer.

Explanation

General search methods such as GPS attempted to explore possible states, but branching factors cause the number of states to grow exponentially with depth. Without domain-specific heuristics to prune the space, performance collapsed on realistic problems.

Question 36Artificial and Human Intelligence: An Introduction and History

Which statement correctly describes the shift from the first to the second AI winter in terms of technology expectations?

  1. Hopes placed in speech synthesis were followed by a ban on voice interfaces
  2. Hopes placed in the internet were followed by the collapse of e-commerce
  3. Hopes placed in expert systems and specialised hardware were followed by disappointment at their brittleness and cost Correct answer.
  4. Hopes placed in quantum computing were followed by the discovery that qubits cannot be controlled

Explanation

The second winter is associated with the collapse of the expert-system market, where systems proved costly to maintain and unable to handle novel cases. Overstated expectations about commercial expert systems and Lisp machines preceded the decline.

Question 37Artificial and Human Intelligence: An Introduction and History

Which feature most clearly distinguishes the Chinese Room argument from the Turing Test as a way of assessing machine intelligence?

  1. It depends on measuring the physical size of the computer
  2. It focuses on whether the system understands rather than on whether its behaviour is indistinguishable from a human's Correct answer.
  3. It judges intelligence by the number of rules stored in memory
  4. It requires the machine to be faster than a human at arithmetic

Explanation

The Turing Test uses observable behaviour as the criterion, whereas Searle's argument asks whether the underlying process involves genuine understanding. Searle claims that symbol manipulation alone yields syntax without semantics.

Question 38Artificial and Human Intelligence: An Introduction and History

A historian notes that many early AI demonstrations worked only in simplified 'toy' domains. Which limitation does this most directly illustrate?

  1. Excessive reliance on graphics processing units
  2. The absence of any programming languages at the time
  3. A legal prohibition on using real-world data
  4. Poor scaling from constrained micro-worlds to open-ended real environments Correct answer.

Explanation

Micro-world programs such as SHRDLU succeeded because the vocabulary and state space were small and closed. When transferred to messy real-world settings, the assumptions behind those demonstrations failed to hold.

Question 39Artificial and Human Intelligence: An Introduction and History

Which contribution is correctly attributed to Warren McCulloch and Walter Pitts?

  1. The invention of the relational database for knowledge storage
  2. A logical model of the neuron showing that networks of simple threshold units can compute Correct answer.
  3. The formulation of the frame problem in robotics
  4. The design of the first stored-program computer

Explanation

Their 1943 work modelled neurons as threshold logic units and demonstrated that networks of them could realise logical functions. This provided a mathematical foundation for later neural network research.

Question 40Artificial and Human Intelligence: An Introduction and History

A system is described as using heuristic search. What does the term 'heuristic' contribute in this context?

  1. A rule of thumb that guides the search toward promising states without guaranteeing optimality Correct answer.
  2. A method for encrypting the search tree
  3. A proof that the search will always find the shortest path
  4. A requirement that all states be examined exhaustively

Explanation

Heuristics estimate promise and prune the search, trading guarantees of completeness or optimality for tractability. They were central to making search feasible in early AI programs.

Question 41Artificial and Human Intelligence: An Introduction and History

Why is the physical symbol system hypothesis often described as an empirical claim rather than a mathematical theorem?

  1. It has been formally proved using set theory
  2. It asserts that symbol manipulation is necessary and sufficient for intelligence, which is a testable but unproven proposition Correct answer.
  3. It applies only to hardware, not to software
  4. It was derived from measuring human brain activity

Explanation

Newell and Simon proposed that a physical symbol system has the necessary and sufficient means for general intelligent action. This is a broad empirical hypothesis about what intelligence requires, not a result proved from axioms.

Question 42Artificial and Human Intelligence: An Introduction and History

What role did the Logic Theorist programme play in the early history of AI?

  1. It proved theorems in symbolic logic, demonstrating machine reasoning on a formal task Correct answer.
  2. It controlled a self-driving vehicle
  3. It generated photorealistic images
  4. It translated spoken language in real time

Explanation

Logic Theorist, developed by Newell, Shaw and Simon, searched for proofs of theorems in Principia Mathematica. It is often cited as one of the first programs to perform a task associated with human reasoning.

Question 43Artificial and Human Intelligence: An Introduction and History

Which observation best explains why the connectionist revival of the 1980s depended on new learning algorithms?

  1. Networks required larger screens to display their outputs
  2. Networks with hidden layers needed a way to assign credit to internal weights, which backpropagation provided Correct answer.
  3. Networks could only be built with quantum hardware
  4. Networks needed formal verification before they could compute

Explanation

Multi-layer networks contain hidden units whose errors are not directly observable, creating a credit-assignment problem. Backpropagation's use of gradient information made training such networks practical and fuelled renewed interest.

Question 44Artificial and Human Intelligence: An Introduction and History

A critic argues that a system which answers every question correctly by consulting a huge lookup table still lacks intelligence. Which historical debate does this argument most resemble?

  1. The debate over the correct size of computer memory
  2. The debate over which programming language is fastest
  3. The debate over whether computers should use binary or decimal arithmetic
  4. The debate over whether symbol manipulation without understanding counts as intelligence Correct answer.

Explanation

The objection targets the gap between producing correct outputs and possessing understanding, echoing Searle's Chinese Room and related critiques of strong AI. Lookup does not by itself supply semantics.

Question 45Artificial and Human Intelligence: An Introduction and History

Why is the Turing Test sometimes criticised as measuring imitation rather than intelligence?

  1. It requires the machine to be physically identical to a human
  2. It demands that the machine solve arithmetic instantly
  3. A system could be designed to mimic human conversational quirks without possessing general reasoning ability Correct answer.
  4. It cannot be administered without specialised hardware

Explanation

Because the test judges only the text produced, a program tuned to imitate human phrasing could pass without broader cognitive competence. This is the basis of the 'imitation game' objection.

Question 46Artificial and Human Intelligence: An Introduction and History

Which statement best explains why the 'frame problem' is a challenge for symbolic agents acting in a changing world?

  1. They must translate their rules into machine code at runtime
  2. They must represent which facts remain unchanged after an action without listing every unaffected fact explicitly Correct answer.
  3. They must compress video streams before reasoning
  4. They must obtain a licence before acting autonomously

Explanation

The frame problem concerns how to specify the effects of actions economically, since most facts persist but enumerating all of them is impractical. It exposes a difficulty in scaling symbolic representation to dynamic environments.

Question 47Artificial and Human Intelligence: An Introduction and History

Consider a system that passes a text-based imitation test but fails when asked to use everyday physical knowledge. What does this combination suggest about the original test?

  1. It shows that physical knowledge is irrelevant to intelligence
  2. It proves that the system has general intelligence despite the failure
  3. It may be insensitive to competences that are not exercised in the restricted conversational setting Correct answer.
  4. It demonstrates that the system cannot produce grammatical sentences

Explanation

The test samples only a narrow channel of behaviour, so success there does not establish broader competence. Failure on common-sense physical reasoning reveals the gap between passing the test and possessing general understanding.

Question 48Artificial and Human Intelligence: An Introduction and History

Which comparison most accurately contrasts the symbolic and connectionist traditions with respect to knowledge representation?

  1. Both traditions require hand-coded rules written by domain experts
  2. Neither tradition represents knowledge in any form
  3. Symbolic systems store knowledge in weights, while connectionist systems store it in explicit rules
  4. Symbolic systems encode explicit structures such as rules and graphs, while connectionist systems distribute knowledge across numeric weights Correct answer.

Explanation

Symbolic AI manipulates discrete, inspectable structures, whereas connectionist models encode regularities in the strengths of connections learned from data. The distinction concerns the form of representation, not merely the implementation language.

Question 49Artificial and Human Intelligence: An Introduction and History

A researcher claims that because a deep network's weights changed after training, the system now 'knows' the concept it learned. Which philosophical issue does this claim most directly raise?

  1. Whether the training data were stored on a solid-state drive
  2. Whether learned statistical regularities constitute knowledge in the sense required for understanding Correct answer.
  3. Whether the programming language supported recursion
  4. Whether the network used too much electricity during training

Explanation

The claim moves from a description of parameter change to an epistemic ascription, raising the question of whether statistical learning yields understanding or merely reliable discrimination. This echoes long-standing debates about strong AI and semantics.

Question 50Artificial and Human Intelligence: An Introduction and History

Why might the history of AI be better described as a series of overlapping research programmes than as a single linear progression?

  1. Symbolic, connectionist and statistical approaches coexisted and influenced one another across the same decades Correct answer.
  2. Because the field has no documented history before 2010
  3. Because all AI research ended during the first winter
  4. Because only one algorithm has ever been used in the field

Explanation

Different approaches developed concurrently, with symbolic methods dominant in some periods while connectionist and statistical work continued and later resurged. Funding cycles and media narratives create an illusion of simple linear progress.