Artificial Intelligence Concepts Worksheet

A perceptron is a simple model of an artificial neuron used for binary classification.

• A single-layer neural network unit with a threshold function

• A deep neural network with multiple layers

• A clustering algorithm

• A regression model used for prediction

Explain what it means to call an AI system "opaque" or "transparent". Select the option that best states the meaning.

• Opaque means the system's decision-making process is not easily understood or visible, while transparent means it is comprehensible and open.

• Opaque means the system is faster, while transparent means it is slower.

• Opaque and transparent refer to the system's color interface design.

• Opaque means the system is cheaper, while transparent means it is more expensive.

In artificial intelligence, the symbolic approach relies on explicit symbols and rule-based logic, whereas the subsymbolic approach involves distributed representations and statistical learning.

• Symbolic AI uses explicit symbols and rules; subsymbolic AI uses distributed representations and statistical methods.

• Symbolic AI uses distributed representations and statistical methods; subsymbolic AI uses explicit symbols and rules.

• Both approaches use explicit symbols and rule-based procedures.

• Both approaches use distributed representations and statistical methods.

Subsymbolic approaches to artificial intelligence create difficulties with explanation because they operate on distributed representations that are not easily mapped to symbolic, interpretable reasoning.

• They operate on distributed representations that are hard to trace back

• They are built on clear, rule-based structures

• They always provide explicit symbolic explanations

• They function with transparent algorithms

According to Kate Vredenburgh, the right to explanation arises when algorithmic decisions significantly impact individuals, ensuring accountability, fairness, and transparency. Which of the following best summarizes why and when this right arises?

• It arises only when decisions are automated without any human review.

• It arises when algorithmic decisions impact individuals and need to be explained for accountability and fairness.

• It arises to ensure that every automated decision is entirely free from errors.

• It arises only under strict data protection regulations without relation to decision impact.

Vredenburgh explains informed self-advocacy as the ability to use both forward-looking abilities (planning for future needs) and backward-looking abilities (reflecting on past experiences); an example is a person planning their career by learning from past job challenges.

• It means a focus on both future planning and past reflection in advocating for oneself.

• It refers exclusively to developing forward-looking abilities.

• It only involves reflecting on past experiences without planning for the future.

• It is solely about using external advice rather than personal reflection.

Alex London responds by stating that the necessity for modern medical professionals to explain the reasons behind their recommendations does not preclude the use of opaque AI in medicine.

• Opaque AI can be valid and reliable even if its internal reasoning is not fully explainable.

• Opaque AI should be discarded because it doesn't allow for clear explanations.

• The transparency of AI is irrelevant because all medical decisions are ultimately subjective.

• Opaque AI is less effective than traditional methods because it hides its reasoning.

Alex London argues in favor of sometimes using opaque AI in medicine because it can offer improved diagnostic accuracy when validated.

• Because it always provides clear and interpretable results.

• Because it can offer improved diagnostic accuracy when validated.

• Because transparency is not necessary in any medical context.

• Because AI should replace human judgement entirely.

According to Carissa Véliz, the modern data economy tends to put privacy in danger due to excessive data commodification and lack of individual control over personal information.

• Because of excessive data commodification and lack of control over personal information.

• Because technology companies invest too heavily in privacy protection.

• Because governments strictly regulate data collection and usage.

• Because data is primarily used for beneficial public services.

Define how hard power refers to coercive methods such as military force and economic pressure, while soft power refers to the ability to attract and co-opt through culture and values.

• Hard power refers to coercive methods such as military force and economic pressure, while soft power refers to the ability to attract and co-opt through culture and values.

• Hard power refers to non-coercive measures and influence via culture, while soft power refers to military strength and economic sanctions.

• Hard power and soft power both rely on diplomatic negotiations.

• Hard power is used strictly in economic contexts while soft power is used strictly in military contexts.