Đề Thi FE ITE303c - SP26 - FE - RE

admin · 3/5/26

ITE303c SP26 FE RE

Câu 1: Choose 1 answer Multiple ChoiceIf an Al system causes damages to third persons...

A. nobody will be held civilly liable as Al systems do not have a will and therefore a conscience for their actions

B. the responsibility of Al's action is always transferred to the human user, who always must take reasonable attention in using this type of technology

C. by virtue of the concept of strict liability even in absence of a will and consciousness, the Al system itself will be held civilly liable, to the extent it is a self-learning one acting on the bases of automated decision-making processes

D. by virtue of the concept of strict liability, the manufacturer or programmer or user might be held liable, also in consideration of the risk connected to the usage of the Al system at issue

Câu 2: Choose 1 answer Multiple ChoiceWhat is a tradeoff involved in deciding who to protect from algorithmic bias?

A. Protecting all demographic groups equally, sacrificing overall model accuracy

B. Prioritizing the majority group to maximize accuracy, potentially perpetuating bias against minority groups

C. Ignoring fairness considerations to ensure model interpretability

D. Excluding certain groups from protection to speed up the model training process

Câu 3: Choose 1 answer Multiple ChoicePredictive loops in marketplace models like dating apps are especially susceptible to bias due to

A. fairness quotients

B. marketplace forces

C. short feedback cycles

D. engagement levels

Câu 4: Choose 1 answer Multiple ChoiceA "purposefully biased" algorithm used to identify unfair attributes is known as

A. an adversarial algorithm

B. a predictive model

C. a discriminatory algorithm

D. an aggregate algorithm

Câu 5: Choose 1 answer Multiple ChoiceIntellectual property rights are extremely important for enterprises as they allow to enhance any creative works the companies have invested in, taking full control over their usage, distribution and dissemination

A. True

B. False

Câu 6: Choose 1 answer Multiple ChoiceA model can still be unfair even though it won't explicitly know which groups are being inputted into the system

A. false-negative optimized

B. blind attribute

C. single attribute

D. biased training

Câu 7: Choose 1 answer Multiple ChoiceWhat does it mean to use fairness as a modeling concept in machine learning?

A. Prioritizing accuracy over fairness to achieve optimal outcomes

B. Incorporating fairness considerations into the model development process

C. Ignoring demographic factors to simplify model training

D. Exclusively focusing on protecting historically disadvantaged groups

Câu 8: Choose 1 answer Multiple ChoiceError rate parity means an equal chance of

A. mistakes made for each group

B. prediction rate for each group

C. approval for each group

D. outcomes for each group

Câu 9: Choose 1 answer Multiple ChoiceAs a cognitive bias, humans see lack of context/meaning around a piece of information and tend to

A. disagree with their previous beliefs

B. seek authority figures

C. fill in gaps with existing knowledge

D. use biased sources of research

Câu 10: Choose 1 answer Multiple ChoiceWhich type of fairness would make sense when dividing tickets evenly between groups?

A. error rate parity

B. equality of prediction rate

C. statistical parity

D. equality of false positives

Câu 11: Choose 1 answer Multiple ChoiceIn a real-world scenario, what is a potential solution to address fairness issues related to disparate impact?

A. Ignoring fairness considerations to maintain high accuracy

B. Balancing false positives and false negatives across demographic categories

C. Prioritizing certain demographic groups over others for better outcomes

D. Continuing with the deployment without addressing fairness concerns

Câu 12: Choose 1 answer Multiple ChoiceThe moral dilemma...

A. Regards the criteria for determining who, between man and machine, should be sanctioned when Al commits an offence

B. Concerns how to establish and set criteria for Al to choose how to act when the choice impacts on human life

C. Is a false problem, as despite many cultural differences, there is a universal morality worldwide

Câu 13: Choose Multiple answersTo be patentable an invention must meet the following requirements:

A. It must be an intellectual property, be new (ie, not already available to the public), be created by a human being, have patrimonial rights in it

B. It must be new (i.e. not already available to the public), involve an inventive step, have an industrial application

C. It must be new (i.e. not already available to the public), be created by a human being, have an industrial application

D. It must be novel, it must not be obvious or easily deducible by a professional of the sector, it must be susceptible to industrial application

Câu 14: Choose 1 answer Multiple ChoiceAn example of a public dataset at risk of an algorithmic privacy violation is the

A. Google search view dataset

B. Fidelity Bank lending dataset

C. Google Maps satellite view dataset

D. Stanford hospital dataset

Câu 15: Choose 1 answer Multiple ChoiceThe machine learning algorithms need to have the inbuilt capability to explain their logic, define their _____ and specify a clear understanding of their future behaviour.

A. Strengths/Weaknesses

B. Drawbacks

Câu 16: Choose 1 answer Multiple ChoiceIn the context of privacy, what role does data encryption play in machine learning?

A. Encrypting data is irrelevant to privacy concerns in machine learning.

B. Data encryption ensures that all model predictions are fully transparent.

C. Encrypting data helps protect sensitive information during data transmission and storage.

D. Privacy considerations do not involve securing data through encryption methods.

Câu 17: Choose Multiple answersWhich of the following are the goals of Al?

A. To Create Express Systems

B. To Implement Human Intelligence in Machines

C. To Create machine with emotions

D. To solve highly complex problems

Câu 18: Choose 1 answer Multiple ChoiceWhat is a potential drawback of black box models concerning trust and accountability?

A. Black box models are always more trustworthy due to their complexity.

B. Understanding the decision-making process in black box models can be challenging, impacting trust.

C. Black box models are inherently more accountable for their predictions

D. Trust and accountability are not relevant considerations for black box models.

Câu 19: Choose 1 answer Multiple Choicea non-sensitive column can endanger user privacy

A. True

B. false

Câu 20: Choose 1 answer Multiple ChoiceA set of governing morals and principles is called

A. Ethics

B. Augmented System

C. Bias

D. Customised system

E. State-of-theart

Câu 21: Choose 1 answer Multiple ChoiceWhat is a potential consequence of algorithmic privacy violations?

A. Increased transparency in data handling

B. Improved data accuracy in the model

C. Unauthorized access to sensitive information

D. Enhanced model interpretability

Câu 22: Choose 1 answer Multiple ChoiceThe following are the risks in developing Al's EXCEPT

A. Loss of job

B. Fraud detection

C. Distribution of power

D. Cost of maintenance and repair

Câu 23: Choose 1 answer Multiple ChoiceThe goal of differential privacy is ensure that the output of a predictive model should not change if the _____ is either included or excluded in the training set.

A. model fairness score

B. individual's data

C. dataset filter

D. biased dataset

Câu 24: Choose 1 answer Multiple ChoiceAn adversarial model relies on using _____ to observe different outputs

A. pressure inputs

B. algorithm detection

C. perturbed inputs

D. sensitive columns

Câu 25: Choose 1 answer Multiple ChoiceA glass box model is traditionally preferred by businesses

A. True

B. False

Câu 26: Choose 1 answer Multiple ChoiceHow are predictive models used in hedge funds?

A. They provide predictions to shareholders to estimate returns

B. They predict future movement of stocks and find points to exploit the market moving in either direction

C. They aid researchers by forecasting financial collapse

D. They predict whether people will be able to pay off loans, and then provide loans

Câu 27: Choose 1 answer Multiple ChoiceA good example of cultural reflection in training data is

A. a predictive model incorporates training data from a variety of sources

B. a model fails to recognize cultural differences due to incorrect attributes

C. an image recognition model selects one face over another based on sample data

D. a model selects for one demographic less often because of their historical representation

Câu 28: Choose 1 answer Multiple ChoiceIn which of the following everyday applications is machine learning commonly employed?

A. Cooking recipes

B. Social media content moderation

C. Car maintenance

D. Planting a garden

Câu 29: Choose 1 answer Multiple ChoiceWe can assume that Al-related technology is driving a kind of new industrial revolution

A. True

B. False

Câu 30: Choose 1 answer Multiple ChoiceWhat is the purpose of dropout regularization in neural networks?

A. To randomly drop input features during training

B. To prevent overfitting by randomly deactivating neurons during training

C. To introduce noise into the dataset for better generalization

D. To speed up the training process by skipping certain iterations

Câu 31: Choose 1 answer Multiple ChoiceA fact of learning algorithms is that

A. even if you haven't shared an direct datapoint about yourself, with enough related datapoints the algorithm can make an educated guess with alarming accuracy

B. just because they are capable of improving outputs, they don't need more inputs

C. they only learn when given small amounts of data, and without the proper training sample, the results can be wildly inaccurate

D. they cannot make predictions with current technology

Câu 32: Choose 1 answer Multiple ChoiceIn what way do navigation apps like Google Maps incorporate machine learning?

A. Providing directions based on a fixed schedule

B. Offering the same route regardless of traffic conditions

C. Analyzing real-time traffic data to suggest the fastest route

D. Ignoring alternate routes and shortcuts

Câu 33: Choose 1 answer Multiple ChoiceA type of artificial intelligence that outperforms humans in all tasks is known as

A. Encompassing Al

B. Outwit Al

C. Specific Al

D. General Al

Câu 34: Choose 1 answer Multiple ChoiceTo measure a predictive model's accuracy, you

A. divide the number of correct predictions by the total number of predictions

B. divide the number of predictions by the total dataset

C. multiply the number of total predictions by the percentage of correct predictions

D. measure the ratio of the model's error curve

Câu 35: Choose Multiple answersFrom a civil liability perspective, if an Al system act as an agent:

A. The existing legal framework on product liability applies

B. There is a regulatory gap that can in no way be filled (even partially) with the current legal framework

C. Pursuant to the European Parliament Resolution on Al issued in 2020, any national legal framework on product liability will no longer be applicable as two measures will be applied in all the EU member States, that are (i) the insurance on any Al machine and (ii) the provision of economic caps for the compensation of damages

D. Pursuant to the European Parliament Resolution on Al issued in 2020, legislation on the matter must be uniformed allocating the responsibility of damages caused by Al systems on those who can exercise control over the them, without however totally repealing the national regulations on product liability currently in force

Câu 36: Choose 1 answer Multiple ChoiceWhat side effect of learning algorithms creates an ethical dilemma for its users?

A. Learning algorithms require large amounts of computing power, which is bad for the environment

B. Learning algorithms require government regulation, which is bad for software developers

C. Learning algorithms are costly to run, which drives up prices for consumer services

D. Learning algorithms require large datasets, which means storing identifying information about users

Câu 37: Choose 1 answer Multiple ChoiceOur goals for building an ethical predictive model include making sure the results are

A. accurate, precise, fair

B. precise, methodical, ethical

C. precise, explainable, predictable

D. accurate, fair and explainable

Câu 38: Choose 1 answer Multiple ChoiceWhat is a recommended practice for addressing bias in a real-world scenario when deploying a machine learning model?

A. Ignoring bias concerns to expedite model deployment

B. Continuously monitoring and updating the model to address biases

C. Assuming that bias is unavoidable and does not need mitigation

D. Exclusively focusing on bias during the model development phase

Câu 39: Choose 1 answer Multiple ChoiceIn the context of real-world fairness analysis, what is the significance of intersectional fairness?

A. Intersectional fairness is irrelevant in real-world scenarios.

B. It involves considering the impact of multiple intersecting demographic attributes.

C. Intersectional fairness only applies to binary classification problems.

D. Ignoring the interplay of different demographic factors improves model simplicity.

Câu 40: Choose 1 answer Multiple ChoiceWhat is foreseeability?

A. The ability to predict the future

B. To describe responsible action

C. To have freedom of action

D. To be able to know the consequences of action

Câu 41: Choose 1 answer Multiple ChoiceWhich of the following frameworks primarily promotes human rights?

A. The Asilomar Al Principles

B. The Beijing Al Principles

C. The Toronto Declaration

D. The Montreal Declaration

Câu 42: Choose 2 answers Multiple ChoiceWhich of the following are requirements set forth by the Biometric Information Privacy Act (BIPA)?

A. Organizations must destroy biometric data in a timely fashion.

B. Organizations must not transmit biometric data across an unsecured network like the Internet.

C. Organizations must store biometric data in local, on-premises databases.

D. Organizations must obtain consent from individuals regarding the collection and use of biometric data.

Câu 43: Choose 1 answer Multiple ChoiceWhat is moral responsibility?

A. It corresponds to liability

B. It means to be accountable for our actions and the effects of these actions

C. It means to adopt a moral theory describing responsibility

D. It is an ethical approach in professional life

Câu 44: Choose 1 answer Multiple ChoiceHow does the "virtuous cycle" that benefits Big Tech operate?

A. By acting virtuous, the public respects Big Tech more and more.

B. Organizations write algorithms with fewer biases, which leads to fairer outcomes.

C. Data-driven algorithms improve solutions, leading to new customers, and better data.

D. Better classes of customers lead to richer and more refined data for algorithmic systems.

Câu 45: Choose 1 answer Multiple ChoiceIn a fictional scenario using game theory fairness analysis, what is the significance of "fair division" strategies?

A. Strategies that prioritize unequal distribution for competitive advantage

B. Strategies that ensure an equal and fair distribution of resources or outcomes

C. Ignoring any division strategies for simplicity

D. Strategies that favor a single player over others for maximum profit

Câu 46: Choose 1 answer Multiple ChoiceHow do Al and other data-driven technologies use probability?

A. By determining the objective likelihood of some event happening

B. By estimating the likelihood of some event happening without input data

C. By guaranteeing that some event will occur with 100% likelihood

D. By providing a model of belief about the likelihood of some event happening

Câu 47: Choose 1 answer Multiple ChoiceHow can the concept of "Nash equilibrium" be applied in a fictional scenario involving game theory fairness analysis?

A. By ensuring that players make decisions independently without considering each other

B. By promoting collaboration among players to achieve the best possible outcome

C. By finding a stable state where no player has an incentive to deviate from their chosen strategy

D. By ignoring strategic interactions among players for simplicity

Câu 48: Choose 1 answer Multiple ChoiceWhat are the jobs that Al is most likely to displace over the next several years?

A. All jobs will be displaced

B. Jobs that comprise primarily of non-routine, non-repetitive work

C. Jobs that comprise primarily of routine, repetitive work

D. Most jobs involving office work (white collar jobs)

Câu 49: Choose 1 answer Multiple ChoiceWhich type of entity are the OECD Principles on Artificial Intelligence mostly geared towards?

A. Municipal governments

B. National governments

C. Private corporations

D. Individuals

Câu 50: Choose 1 answer Multiple ChoiceIn a fictional scenario involving game theory fairness analysis, how might the concept of "iterated dominance" be used to refine strategic choices?

A. By consistently choosing the same dominant strategy throughout the game

B. By iteratively eliminating strategies that are dominated by other available choices

C. Iterated dominance has no relevance to strategic decision-making

D. By alternating between dominant and dominated strategies to confuse opponents

Đề Thi FE ITE303c - SP26 - FE - RE

admin

Member

Đính kèm

Đề Thi FE ITE303c - SP26 - FE - RE

adminadmin is verified member.

Member

Đính kèm

admin